阅读说明：本技术 缩放工具 (Scaling tool ) 是由 艾德里安·斯塔赛 约亨·舒尔茨 于 2019-01-31 设计创作，主要内容包括：本申请总体上涉及缩放生产化学、药物和/或生物技术产品的生产过程和/或缩放相应生产设备的生产状态。特别地,提供了一种缩放生产化学、药物和/或生物技术产品的生产过程的计算机实现的方法,该缩放是从源规模到目标规模,其中生产过程由控制生产过程的执行的一个或多个过程参数所指定的多个步骤限定,该方法包括：(a)检索：描述过程参数的时间演化的参数演化信息；多个配方模板,其中配方包括限定生产过程的多个步骤,并且其中配方模板是指定多个步骤的过程参数中的至少一个是可变的且一开始没有预定值的参数的配方；(b)接收：用于在源规模执行生产过程的源机构的源机构规格,源机构规格包括源规模值；用于在目标规模执行生产过程的目标机构的目标机构规格,目标机构规格包括目标规模值；限定源规模下的生产过程的源配方；限定源规模和/或目标规模下的所述过程参数的值的条件的至少一个可接受性函数；(c)使用源机构规格、源配方和参数演化信息来模拟生产过程在源规模下的执行；(d)根据模拟来确定过程参数的一个或多个源轨迹,其中轨迹对应于在模拟生产过程的执行期间可记录的值的基于时间的曲线；(e)执行目标确定步骤,该目标确定步骤包括：从多个配方模板中选择与生产过程相关的配方模板；为所选择的配方模板中的至少一个可变参数提供输入值；使用目标机构规格、所选择的配方模板、至少一个可变参数的输入值和参数演化信息来模拟生产过程在目标规模下的执行；根据模拟来确定过程参数的一个或多个目标轨迹；将源轨迹与目标轨迹进行比较；基于比较和至少一个可接受性函数来计算所选择的配方模板的可接受性得分；通过优化可接受性得分和/或计算至少一个可变参数的可接受范围来计算所选择的配方模板中的至少一个可变参数的最佳值；(f)如果存在至少另一个相关配方模板,则针对至少另一个相关配方模板重复目标确定步骤；(g)基于针对一个或多个配方模板计算的可接受性得分来选择多个配方模板中的至少一个和可变参数的对应计算值作为目标配方。(The present application relates generally to scaling a production process for producing chemical, pharmaceutical and/or biotechnological products and/or scaling production states of corresponding production equipment. In particular, there is provided a computer-implemented method of scaling a production process for producing a chemical, pharmaceutical and/or biotechnological product, the scaling being from a source scale to a target scale, wherein the production process is defined by a plurality of steps specified by one or more process parameters controlling the execution of the production process, the method comprising: (a) and (3) retrieval: parameter evolution information describing the time evolution of the process parameters; a plurality of recipe templates, wherein a recipe comprises a plurality of steps that define a production process, and wherein a recipe template is a recipe that specifies a parameter for which at least one of the process parameters for the plurality of steps is variable and which initially has no predetermined value; (b) receiving: a source mechanism specification for a source mechanism for performing a production process at a source scale, the source mechanism specification comprising a source scale value; a target mechanism specification of a target mechanism for performing a production process at a target scale, the target mechanism specification including a target scale value; defining a source recipe for a production process at a source scale; at least one acceptability function defining conditions for values of the process parameter at a source scale and/or a target scale; (c) simulating execution of the production process at the source scale using the source agency specification, the source recipe and the parameter evolution information; (d) determining one or more source trajectories of process parameters from the simulation, wherein the trajectories correspond to time-based curves of values recordable during execution of the simulated production process; (e) performing a goal determining step, the goal determining step comprising: selecting a recipe template associated with the production process from a plurality of recipe templates; providing an input value for at least one variable parameter in the selected recipe template; simulating execution of the production process at the target scale using the target agency specification, the selected recipe template, the input value for the at least one variable parameter, and the parameter evolution information; determining one or more target trajectories for the process parameter based on the simulation; comparing the source trajectory with the target trajectory; calculating an acceptability score for the selected recipe template based on the comparison and the at least one acceptability function; calculating an optimal value for at least one variable parameter in the selected recipe template by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter; (f) repeating the targeting step for at least one other related recipe template if there is at least one other related recipe template; (g) at least one of the plurality of recipe templates and the corresponding calculated value of the variable parameter are selected as the target recipe based on the acceptability scores calculated for the one or more recipe templates.)

1. A computer-implemented method of scaling a production process for producing a chemical, pharmaceutical and/or biotechnological product, the scaling being from a source scale to a target scale, wherein the production process is defined by a plurality of steps specified by one or more process parameters that control execution of the production process, the method comprising:

-retrieving:

-parameter evolution information describing the time evolution of said process parameters;

-a plurality of recipe templates, wherein:

the recipe includes the plurality of steps defining the production process, and

the recipe template is a recipe specifying a parameter for which at least one of the process parameters for the plurality of steps is variable and initially has no predetermined value;

-receiving:

-a source authority specification for a source authority for performing the production process at the source scale, the source authority specification comprising a source scale value;

-a target mechanism specification for a target mechanism for performing the production process at the target scale, the target mechanism specification comprising a target specification value;

-defining a source recipe for said production process at said source scale;

-at least one acceptability function defining conditions for values of said process parameters at said source scale and/or at said target scale;

-simulating the execution of the production process at the source scale using the source agency specification, the source recipe and the parameter evolution information;

-determining one or more source trajectories of the process parameters from the simulation, wherein a trajectory corresponds to a time-based curve of values recordable during the simulation of the execution of the production process;

-performing a targeting step, the targeting step comprising:

-selecting a recipe template associated with said production process from said plurality of recipe templates;

-providing input values for at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the target scale using the target agency specifications, the selected recipe template, the input values of the at least one variable parameter and the parameter evolution information;

-determining one or more target trajectories of said process parameters from said simulation;

-comparing said source trajectory with said target trajectory;

-calculating an acceptability score for the selected recipe template based on the comparison and the at least one acceptability function;

-calculating an optimal value for the at least one variable parameter in the selected recipe template by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, wherein values within the acceptable range yield an acceptability score above a specific threshold;

-if there is at least one other related recipe template, repeating the targeting step for the at least one other related recipe template;

-selecting at least one of the plurality of recipe templates and a corresponding calculated value of a variable parameter as a target recipe based on the acceptability score calculated for one or more recipe templates.

2. The computer-implemented method of claim 1, further comprising outputting the target recipe.

3. The computer-implemented method of claim 1 or 2, further comprising:

providing the target recipe to a target control system;

executing, by the target control system, the production process at the target scale based on the target recipe.

4. The computer-implemented method of claim 3, further comprising:

evaluating the performance of the production process at the target scale; and

modifying the at least one acceptability function based on the evaluation.

5. The computer-implemented method of any of the preceding claims, wherein a plurality of acceptability functions are received and a plurality of target trajectories are calculated, and wherein calculating the acceptability score comprises:

obtaining, for each target trajectory of the plurality of target trajectories, a second partial acceptability score by:

-selecting one or more applicable acceptability functions;

-for each applicable acceptability function, performing the following calculation steps:

-calculating an acceptability value based on the acceptability function for each time point in the target trajectory;

-accumulating said acceptability values for different time points to obtain a first partial acceptability score;

-if there is a single applicable acceptability function, setting the second part acceptability score to the first part acceptability score;

-if there are a plurality of applicable acceptability functions, accumulating the first part acceptability scores of all applicable acceptability functions to obtain the second part acceptability score; and

accumulating the second portion acceptability scores for all target trajectories to obtain the acceptability score.

6. The computer-implemented method of any of the preceding claims, further comprising:

defining a target quantity characterizing the production process at the source scale;

performing, by a source control system, the production process a plurality of times at the source scale while changing the process parameters and/or process steps;

selecting a process based on the results of the target quantities given by the process parameters and the process steps in order to define the source recipe.

7. A computer-implemented method of scaling a production process for producing a chemical, pharmaceutical and/or biotechnological product, the scaling being from a source scale to an intermediate target scale to a final target scale, wherein the production process is defined by a plurality of steps specified by one or more process parameters controlling execution of the production process, the method comprising:

-retrieving:

-parameter evolution information describing the time evolution of said process parameters;

-a plurality of recipe templates, wherein:

the recipe includes the plurality of steps defining the production process, and

the recipe template is a recipe specifying a parameter for which at least one of the process parameters for the plurality of steps is variable and initially has no predetermined value;

-receiving:

-a source authority specification for a source authority for performing said production process at said source scale, said source authority specification comprising a source scale value;

-an intermediate target mechanism specification for an intermediate target mechanism for performing the production process at the intermediate target scale, the intermediate target mechanism specification comprising an intermediate target gauge value;

-a final target mechanism specification for a final target mechanism for performing the production process at the final target scale, the final target mechanism specification comprising a final target mechanism specification value;

-defining a source recipe for said production process at said source scale;

-at least one acceptability function that defines conditions for a value of the process parameter at any one of the source scale, the intermediate target scale and the final target scale when said value is considered singularly and/or when said value at any one scale is considered relative to a corresponding value at any other scale or scales;

-simulating the execution of the production process at the source scale using the source agency specification, the source recipe and the parameter evolution information;

-determining one or more source trajectories of the process parameters from the simulation, wherein a trajectory corresponds to a time-based curve of values recordable during the simulation of the execution of the production process;

-performing a targeting step, the targeting step comprising:

-selecting a recipe template associated with said production process from said plurality of recipe templates;

-providing a first input value for said at least one variable parameter in the selected recipe template;

-providing a second input value for said at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the intermediate target scale using the intermediate target agency specification, the selected recipe template, the first input value of the at least one variable parameter and the parameter evolution information;

-determining one or more intermediate target trajectories of said process parameters from said simulation;

-simulating the execution of the production process at the final target scale using the final target agency specification, the selected recipe template, the second input value of the at least one variable parameter and the parameter evolution information;

-determining one or more final target trajectories of the process parameters from the simulation;

-performing a first, second and third pair-wise comparison between any two of the source trajectory, the intermediate target trajectory and the final target trajectory, and performing a three-way comparison between the source trajectory, the intermediate target trajectory and the final target trajectory;

-calculating an acceptability score based on at least two comparisons and the at least one acceptability function;

-calculating a first and a second optimum of said at least one variable parameter by optimizing said acceptability score, and/or

Calculating a first acceptable range and a second acceptable range for the at least one variable parameter, wherein values within the first acceptable range and values within the second acceptable range yield an acceptability score above a particular threshold;

-if there is at least one other related recipe template, repeating the targeting step for the at least one other related recipe template;

-selecting at least one of the plurality of recipe templates and a corresponding calculated value of a variable parameter as a target recipe based on the acceptability score calculated for one or more recipe templates.

8. A computer-implemented method of scaling a state of a production facility of a production process for producing a chemical, pharmaceutical and/or biotechnological product, the scaling being from a source scale to a target scale, wherein the state is defined by a set of state parameters describing a condition and/or behavior of the production facility, the method comprising:

-retrieving mapping information describing how the state parameters relate to a set of derived parameters;

-receiving:

-a source authority specification for a source authority for performing said production process at said source scale, said source authority specification comprising a source scale value;

-a target mechanism specification for a target mechanism for performing the production process at the target scale, the target mechanism specification comprising a target specification value;

-a first set of state parameters at the source scale;

-a second set of state parameters at said target scale, wherein at least one of said state parameters at said target scale is a variable parameter having initially no predetermined value;

-at least one acceptability function defining conditions on the value of the state parameter and/or the value of the derived parameter at the source scale and/or the target scale;

-calculating a first set of derived parameters at the source scale using the first set of state parameters, the source authority specifications and the mapping information;

-providing input values for said at least one variable parameter of said second set of state parameters;

-calculating a second set of derived parameters at the target scale using the second set of state parameters, the input values, the target agency specifications and the mapping information;

-comparing the first set of state parameters with the second set of state parameters and/or comparing the first set of derived parameters with the second set of derived parameters;

-calculating acceptability scores for the second set of state parameters based on the comparison and the at least one acceptability function;

-calculating an optimal value of the at least one variable parameter by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, wherein values within the acceptable range yield an acceptability score above a specific threshold.

9. A computer-implemented method of scaling a state of a production facility of a production process for producing a chemical, pharmaceutical and/or biotechnological product, the scaling being from a source scale to an intermediate target scale to a final target scale, wherein the state is defined by a set of state parameters describing a condition and/or behavior of the production facility, the method comprising:

-retrieving mapping information describing how the state parameters relate to a set of derived parameters;

-receiving:

-a source authority specification for a source authority for performing said production process at said source scale, said source authority specification comprising a source scale value;

an intermediate target mechanism specification for an intermediate target mechanism for performing the production process at the intermediate target scale, the intermediate target mechanism specification comprising an intermediate target mechanism specification value;

-a final target mechanism specification for a final target mechanism for performing the production process at the final target scale, the final target mechanism specification comprising a final target mechanism specification value;

-a first set of state parameters at the source scale;

-a second set of state parameters at said intermediate target scale, wherein at least one of said state parameters at said intermediate target scale is a first parameter that is variable and initially has no predetermined value;

-a third set of state parameters at said final target scale, wherein at least one of said state parameters at said final target scale is a second parameter which is variable and initially has no predetermined value;

-at least one acceptability function defining conditions with respect to the value of the state parameter and/or the value of the derived parameter at the source scale and/or the intermediate target scale and/or the final target scale;

-calculating a first set of derived parameters at the source scale using the first set of state parameters, the source authority specifications and the mapping information;

-providing a first input value for at least one first variable parameter of the second set of state parameters;

-calculating a second set of derived parameters at the intermediate target scale using the second set of state parameters, the first input values, the intermediate target agency specifications and the mapping information;

-providing a second input value for at least one second variable parameter of the third set of state parameters;

-calculating a third set of derived parameters at the final target scale using the third set of state parameters, the second input values, the final target mechanism specification and the mapping information;

-performing a plurality of pair-wise comparisons within any two of all pairs of the first, second and third set of state parameters and/or within any two of all pairs of the first, second and third set of derived parameters, and performing at least one three-way comparison between the first, second and third set of state parameters and/or between the first, second and third set of derived parameters;

-calculating an acceptability score based on at least two comparisons and the at least one acceptability function;

-calculating a first optimal value of said at least one first variable parameter and a second optimal value of said at least one second variable parameter by optimizing said acceptability score, and/or

Calculating a first acceptable range for the at least one first variable parameter and a second acceptable range for the at least one second variable parameter, wherein values within the first acceptable range and values within the second acceptable range yield an acceptability score above a particular threshold.

10. A computer program product comprising computer readable instructions which, when loaded and executed on a computer system, cause the computer system to perform the operations of any preceding claim.

11. A computer system operable to scale a production process for producing a chemical, pharmaceutical and/or biotech product from a source scale to a target scale, wherein the production process is defined by a plurality of steps specified by one or more process parameters that control execution of the production process, the computer system comprising:

-a retrieval module configured to retrieve:

-parameter evolution information describing the time evolution of said process parameters;

-a plurality of recipe templates, wherein:

the recipe includes the plurality of steps defining the production process, and

the recipe template is a recipe specifying a parameter for which at least one of the process parameters for the plurality of steps is variable and initially has no predetermined value;

-a receiving module configured to receive:

-a source authority specification for a source authority for performing said production process at said source scale, said source authority specification comprising a source scale value;

-a target mechanism specification for a target mechanism for performing the production process at the target scale, the target mechanism specification comprising a target specification value;

-defining a source recipe for said production process at said source scale;

-at least one acceptability function defining conditions for values of said process parameters at said source scale and/or at said target scale; and

-a computing module configured to:

-simulating the execution of the production process at the source scale using the source agency specification, the source recipe and the parameter evolution information;

-determining one or more source trajectories of the process parameters from the simulation, wherein a trajectory corresponds to a time-based curve of values recordable during the simulation of the execution of the production process;

-performing a targeting step, the targeting step comprising:

-selecting a recipe template associated with said production process from said plurality of recipe templates;

-providing input values for at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the target scale using the target agency specifications, the selected recipe template, the input values of the at least one variable parameter and the parameter evolution information;

-determining one or more target trajectories of said process parameters from said simulation;

-comparing said source trajectory with said target trajectory;

-calculating an acceptability score for the selected recipe template based on the comparison and the at least one acceptability function;

-calculating an optimal value for the at least one variable parameter in the selected recipe template by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, wherein values within the acceptable range yield an acceptability score above a specific threshold;

-if there is at least one other related recipe template, repeating the targeting step for the at least one other related recipe template;

-selecting at least one of the plurality of recipe templates and a corresponding calculated value of a variable parameter as a target recipe based on the acceptability score calculated for one or more recipe templates.

12. The computer system of claim 11, further comprising an output module configured to output the target recipe.

13. The computer system of claim 11 or 12, further configured to interface with a target control system for controlling a target process plant, wherein:

the computing module is further configured to provide the target recipe to the target control system; and

the target control system is configured to execute the production process at the target scale based on the target recipe.

14. The computer system of claim 13, wherein the computing module is further configured to:

evaluating the performance of the production process at the target scale; and

modifying the at least one acceptability function based on the evaluation.

15. The computer system of any one of claims 11 to 14, further configured to interface with a source control system for controlling a source process device, wherein:

the source control system is configured to execute the production process a plurality of times at the source scale while changing the process parameters and/or the process steps; and the calculation module is configured to define a target quantity characterizing the process at the source scale and to select a process based on results of the target quantity given by the process parameters and the process steps in order to define the source recipe.

Technical Field

The following description relates to processes for producing chemical, pharmaceutical and/or biotechnological products. In particular, aspects of the present application relate to scaling processes on two or more scales.

Background

Processes for the production of chemical, pharmaceutical and/or biotechnological products are scale-dependent; in other words, the process behaves at least partially differently on a small scale (e.g., in a laboratory) as compared to a large scale (e.g., in production). Typically, the process is performed first on a small scale and then on a successively larger scale.

However, at each scale switch there is a risk that process performance will be lost. This loss may be a catastrophic failure on a larger scale or simply a reduction in product quality or titer. Process performance is problematic because it is not possible to keep all process and physical parameters constant on a scale. For example, mixing time tends to increase with scale as an important driver for the microenvironment seen by the cells in the bioreactor. To compensate for this, a larger stirring speed may be selected on a larger scale. However, this would significantly increase the specific power input, which may be detrimental to the cell or product per se. Similarly, on a smaller scale, evaporation and sampling involve a larger proportion of the bioreactor volume than on a larger scale.

Thus, there is a problem associated with how best to scale up a process that has been found to be optimal on a small scale, or in other words how best to shift the process from a source scale to a target scale, possibly through an intermediate target scale.

In current methods for transitioning between scales, scale-independent parameters are used as intermediaries or links between different scales. Starting from a set of known parameters for the source mechanism configuration, such as the agitation speed, fill volume and gas exit rate of the bioreactor, scale independent parameters (e.g. power input per volume) are derived by combining the known parameters. Then, a given parameter (e.g., a desired fill volume) is set for the target mechanism configuration. Finally, the remaining parameters of the target mechanism configuration (e.g., stirring speed and gassing rate) are calculated to match the previously obtained values of the scale-independent parameters.

However, the above-described scale-up method suffers from several problems:

use only a single scale independent variable as an intermediary between scales, however these processes require careful compromises between multiple variables.

Handling scale transitions as if they were only single step transitions, i.e. a direct transition from one source scale to a single target scale, whereas scale transitions often occur at multiple scales (so-called "scaling queues"). In other words, in a typical process, only two scales are considered at a time for each transition, without regard to possible subsequent transitions. Thus, if a first transition results in a "dead-end mustache" because the final goal is not considered, subsequent transitions (e.g., to an even larger scale) may be problematic. A "dead-end" is a configuration that is risky for subsequent scaling in terms of the prospect of a failed transition. For example, a transition from a small scale to a medium scale may favor very low agitation speeds on the medium scale (e.g., to maintain low shear forces in the cellular environment), but such low agitation speeds may then not be able to transition to the large scale as a result of the mixing time.

Failure to take into account the fact that the variable differences between scales may be asymmetric as a result. For example, mixing time considerations are highly asymmetric: reducing the mixing time is generally not a problem, but increasing the mixing time can be a problem. Similarly, the process can be spared an increase in kLa, but not a decrease.

No way is provided for a priori information on what constitutes the optimal process (e.g. the desire to reduce energy input on a large scale, the desire to sample frequently on a small scale) to be integrated with the need to match certain conditions between scales. This a priori information includes two aspects. First, a goal in terms of constraints on how the bioreactor is expected to operate. For example, a user may want to run a bioreactor with certain constraints (e.g., 5% to 95% of maximum stirring speed). Second, knowledge about the sensitivity of an organism or process, for example, a particular organism or process may be highly oxygen-requiring, highly pH-sensitive, etc. and the transition needs to take this into account. Similarly, certain organisms, products or processes may be shear sensitive and therefore need to be taken into account in terms of tip speed/swirl size/etc. at the transition. In a typical approach, no a priori process knowledge is taken into account when scaling. Instead, it is introduced afterwards, resulting in unrealistic process parameter values. -providing a binary result, i.e. good/acceptable or bad/unacceptable, with respect to one or more parameters of the target scale. In fact, there is a range of possible values associated with lower or higher risk of reducing process performance. In other words, the prior art methods do not provide a way to investigate the sensitivity of the process to one or more parameters.

No means is provided for scaling the process from one scale to another by considering the process as a whole, but instead focusing on a single point in time within the process. Optimization at some points in time within the process may be disadvantageous for other points in time if the process is not considered as a whole.

Therefore, there is a need for a scaling method that reduces the risk associated with scaling. In particular, it is desirable to identify appropriate process conditions at large scale to reduce the risk associated with poor process performance at subsequent larger scales.

Disclosure of Invention

It is an object of the present invention to move a process (for the production of chemical, pharmaceutical and/or biotechnological products) from a source scale to a target scale in such a way that the similarity of the process is maximized between scales in terms of the success of the process. A process is similar at different scales if the important predictions of performance (e.g., in terms of productivity or titer) determined from a priori knowledge are themselves similar between the source and target processes.

For example, if a process at the source scale results in a low percentage of Dissolved Oxygen (DO) throughout the process, a good transition at the target scale will typically maintain a low percentage of DO throughout the process, for example by adjusting the stirring speed or gassing. The percentage of DO known is generally a prediction of the performance of the organism and is also a prediction of productivity and quality. While in some cases, a higher DO percentage may be better in terms of productivity or quality at the target scale, scaling resulting in a higher DO percentage would be considered undesirable because of the lower similarity between processes. However, if the organism is known to be insensitive to the DO percentage, then the similarity between processes will be evaluated based on other aspects that actually affect the product.

In particular, scaling should not only optimize the results of the process at the target scale, e.g., quality and yield of the product. Instead, the process itself is also optimized in terms of similarity between different scales. Thus, the best match between the process at the source scale and the process at the target scale is found from both the process itself and the product perspective.

In other words, it is an object of the present invention to identify how to run a process at a given scale (e.g., a larger scale) to maximize the chances of obtaining the same performance obtained at another scale (e.g., a smaller scale). It is also an object of the present invention to identify how to run a process at a given scale (e.g., smaller scale) in view of an expected deployment at another scale (e.g., larger scale). In particular, it is contemplated to identify a range of process variants that can be run on a smaller scale, where it is reasonable to expect them to behave similarly on one or more larger scales. Thus, when the process is optimized at small scale (by performing experiments within a defined range), the optimized process then switches well back to larger scale.

It is another object of the present invention to identify problems with processes in computers prior to deployment on hardware and to find process alternatives to overcome these problems.

The objectives according to the invention are achieved as set forth in the independent claims. Further developments of the invention are the subject matter of the dependent claims.

According to an aspect, a computer-implemented method of scaling a production process for producing a chemical, pharmaceutical and/or biotech product is provided.

An example of a process according to the present application is an industrial process, in particular a biopharmaceutical process. Other examples include research and development processes or scientific research.

Examples of inputs or raw materials for a production process according to the present application may include biological materials, i.e. materials comprising biological systems, such as cells, cellular components, cellular products and other molecules, and materials derived from biological systems, such as proteins, antibodies and growth factors. Other raw materials may include chemical compounds and various substrates.

Examples of inputs may include gases and liquids. The gas is any or all of air, oxygen, nitrogen, oxygen-enriched air and carbon dioxide. The liquid is typically:

medium (nutrient mixture in buffer, e.g. glucose + amino acids + salts + water)

Inoculum (relatively high density of organisms in the medium)

Bases (for adjusting pH, e.g. ammonium hydroxide solution)

Acids (for adjusting pH, e.g. HCl solution)

Nutrient feed (high concentration nutrient mixture in buffer)

Inducers (regulating the behavior of the organism).

The production process of the present application may involve chemical or microbial conversion of materials and transfer of mass, heat and momentum. The process may include homogeneous or heterogeneous chemical and/or biochemical reactions. The process may include, but is not limited to, mixing, filtering, clarification, centrifugation, and/or cell culture. The production process may involve chemical or biological reactions that take time to complete, for example, 6 hours for an E.coli microbial batch and 60 days for a mammalian perfusion process.

In particular, "producing" a chemical, pharmaceutical, and/or biotechnological product indicates any treatment of the input, including, but not limited to, modifying the state of any input (e.g., changing its temperature, oxygen content, etc.), combining any input reversibly or irreversibly, using the input to create a new material.

Possible products may include transformed substrates, baker's yeast, lactic ferments, lipases, invertases, rennet. Other exemplary biopharmaceutical products that may be produced according to the techniques described in this application include the following: recombinant and non-recombinant proteins, vaccines, gene vectors, DNA, RNA, antibiotics, secondary metabolites, growth factors, cells for cell therapy or regenerative medicine, semisynthetic products (e.g. artificial organs). Various production systems can be used to facilitate the process, for example, cell-based systems such as animal cells (e.g., CHO, HEK, PerC6, VERO, MDCK), insect cells (e.g., SF9, SF21), microorganisms (e.g., e.coli, s.ceravisae, p.pastoris, etc.), algae, plant cells, cell-free expression systems (cell extracts, recombinant nucleoprotein systems, etc.), progenitor cells, stem cells, natural and genetically manipulated patient-specific cells, matrix-based cell systems.

Illustratively, the production process may be a batch process, wherein a specific amount of feed medium for distributing the organisms is provided as an initial condition, followed by a control period.

The production process may be a fed-batch process. A fed-batch process may involve culturing in which the base medium supports the initial cell culture and once the initial nutrients are depleted, the feed medium is added to support further growth or product production. In other words, a fed-batch process may involve a batch phase followed by a feed phase.

The production process may be a perfusion process, wherein a batch phase is followed by a feed phase, wherein the product is continuously removed, e.g. by filtration.

The techniques described herein may be used in bioreactor processes, and in processes performed at other production levels.

The production process is defined by a plurality of steps specified by one or more process parameters that control the execution of the production process.

The production process is defined by a sequence of steps that are performed in order to arrive at a product. Some steps may occur simultaneously with one another, and other steps may occur one after another in a time series. The steps may correspond to actions performed during the production process, wherein the actions may be passive, such as waiting for an event to occur (such as an increase in oxygen content due to the culture being in an inactive state), or active, such as causing an event to occur (e.g., stirring or adding fluid) or setting a given amount of value and/or profile.

For example, the steps of performing actions within a production process in a bioreactor may be represented by the following non-exhaustive list:

setting set points in terms of stirring, gas supply, gas mixing, temperature

Performing the depicted changes in terms of stirring, gas supply, gas mixing, temperature

-adding a selected liquid to the bioreactor vessel

-removing liquid from the bioreactor vessel

-specifying the connection of a specific fluid to the bioreactor and the composition of such fluid.

Further, a process may include types of steps that describe the flow of execution, e.g., specify how/when events occur, such as: performing one or more steps until a condition is satisfied; and/or select between various options depending on the state of the bioreactor for a specified number of iterations; waiting until conditions become true (e.g., waiting until dissolved oxygen rises to indicate the end of the batch phase before starting the feed); one or a set of steps are performed simultaneously (e.g., pH control is performed simultaneously with temperature control).

The performance of the various steps is controlled by one or more process parameters. Illustratively, the plurality of steps may be specified by a plurality of parameters, wherein each step is defined by one or more process parameters.

In other words, the production process may include (i.e., may be performed in accordance with) at least one process parameter that has an impact on the performance of the process (e.g., product titer, quality attribute) and the product produced by the process.

The process parameters then control the execution of the process in the sense that they influence the course of the process, but at least some of them may also be influenced by the process in return. Furthermore, process parameters may affect each other.

In particular, some process parameters may be controllable, i.e. the values of at least some process parameters may be specifically adjusted before and/or during execution of the process. In other words, at least one of the process parameters may be set by, for example, an operator or a control system. In particular, these parameters may be parameters describing and/or managing the status and/or behavior of a device (e.g., a bioreactor) used in a production process. Hereinafter, these parameters may also be referred to as "recipe parameters" because they may be set in a recipe, as explained below.

Thus, the adjustable process parameters may be a suitable subset of the process parameters. In the case of bioreactors, these may include, but are not limited to:

one or more parameters directly related to the intervention in the bioreactor, such as the amount of liquid removed in the sampling step;

providing one or more parameters of input to the control loop, such as a set point for oxygen in the bioreactor (the control loop in the bioreactor system would then monitor the oxygen and adjust, for example, the agitation or gassing to reach the set point);

-one or more parameters providing input to the curve, e.g. rate index in exponential increase of feed; and/or

One or more parameters specifying the conditions to be met (for example, the dissolved oxygen must reach 90% before the feed phase can start).

Illustratively, the recipe parameters for a bioreactor may be or include agitation speed, temperature, gassing, liquid addition, sampling, relative profiles, and an indication of which liquids to add.

In particular, the adjustable parameter may be given a constant fixed value expressed by a given number or a value expressed as a function of an independent variable, where the independent variable may be another process parameter (having a constant or varying value) or another quantity, such as time. Thus, the values of the adjustable parameters may be set at the beginning of execution of the production process or may be determined dynamically during the production process. For example, the set points and curves may be dynamically determined to account for events occurring during the production process.

The process parameters may also include one or more parameters describing the state of production, which of course is determined at least in part by the behavior of the device as well as the details of the device and the inputs (e.g., organisms) used. The values of these parameters may be inherent to the production process and may not be directly adjustable. However, they can be adjusted indirectly by modifying factors that affect them, such as recipe parameters.

The values of these parameters may change during the production process and therefore, in the following, they may be referred to as "dynamic parameters". The values of the dynamic parameters measured at a given time during a process (performed or simulated in the real world) correspond to what is commonly referred to as a "process value" or "process variable".

For example, dynamic parameters of a production process in a bioreactor can be classified as:

a. calculable chain effects of recipe parameters as a result of the geometry and capabilities of the bioreactor, e.g., tip speed, agitation speed as part of maximum bioreactor agitation speed, superficial gas velocity;

b. linkage effects that can be obtained from previous empirical studies, e.g., kLa, mixing time, power input, minimum vortex size;

c. variables that can be calculated from those in point (b) and the properties of the bioreactor, such as power per unit volume;

d. variables resulting from feedback from simulated bioreactor attributes and aspects of the process due to control loops in the process, such as gas exit rate, gas mixing and agitation speed;

e. variables caused by the dynamics of oxygen or other gases within the bioreactor when influenced by, for example, agitation speed, gassing, etc., e.g., DO, partial pressure of carbon dioxide (ppCO 2);

f. variables caused by the dynamics of the organism, as determined by the organism model and other variables above and below (e.g., cell count, cell activity, cell metabolism);

g. variables resulting from liquid addition or removal that can be calculated from recipe parameters (e.g., fill volume) and possibly evaporation models,

h. variables caused by the combination of liquid addition and removal and also the kinetics of the organism, such as the concentration of the analyte.

Illustratively, the process parameters include both recipe parameters and dynamic parameters.

Based on the above, examples of process parameters may include, but are not limited to: temperature (affecting cell growth), volume, pH (affecting cell growth), specific buffering capacity (affecting rate of pH change), cell density, cell activity state (average), cell metabolic state (average), kLa (affecting oxygen transfer), reynolds number (affecting mixing time and cell growth), froude number, mixing time, power input per volume, agitation speed, tip speed as part of the maximum possible tip speed in the system, gas exit rate, gas exit mixing, minimum vortex size (potentially affecting cell health), surface gas velocity, concentration of extracted nutrients (e.g., primary carbon source, secondary carbon source, waste products, base, acid, primary nitrogen source, secondary nitrogen source), inducer, key analytes (e.g., product quality, cell debris), protein concentration (e.g., can affect foam production), cellular parameters (e.g., cell subpopulations), bioreactor heterogeneity (e.g., changes in temperature within the bioreactor), hydrodynamic properties (e.g., proportion of cells in time in a high shear environment) versus proportion of the bioreactor without agitation (dead zone), proportion of the bioreactor swept by the impeller per unit time, carbon dioxide and carbonation dynamics, antifoam concentration (interacting with kLa and other gas transfers), and foam accumulation parameters (e.g., function of SGV and protein concentration).

Scaling of the process is performed between a source scale and a target scale.

The size being particularly the size of the mechanism assigned to the arrangement, e.g. for performing the production processWherein the configuration determines, among other things, the throughput and the cost of the production process. Illustratively, for a production process performed with a bioreactor, a scale value may refer to the volume of the bioreactor and/or one or more components thereof, such as an impeller (e.g., type and/or size thereof). The range of scales on which production processes are typically performed includes 2mL (e.g., in microfluidic examples), less than 15mL, 250mL, 2L, 10L, 50L, 200L, 1000L, and 2000L. Bioreactors operating at these scales include Sartorius products such asAnd BIOSTAT

The scale can be divided into three groups: small scale, medium scale and large scale. Some scales may belong to more than one group. For example, 2L may be both small and medium scale, while 50L may be both medium and large scale.

Scaling the process indicates that the process designed and/or tested at the source scale is adapted for the target scale so that the process is still successful. The success of the process may be evaluated, for example, based on: the amount of product produced (titer), the quality of the product (e.g., chemical composition, including glycosylation, and protein folding pattern), and the presence/absence of other factors in the medium that lead to downstream difficulties in purification. For example, quality attributes may be used to assess success, where quality attributes may be physical, chemical, biological, or microbiological properties that should be within appropriate limits, ranges, or distributions to ensure desired product quality.

During the scaling process, one or more of the plurality of steps may be modified, in particular one or more of the process parameters of the specified step, in particular recipe parameters, may be changed. Further, in some examples, multiple steps may be modified by adding or removing one or more steps, e.g., by adding dependencies on conditions to modify actions, and/or changing the order of steps.

The reason why the production process is performed at a given scale and then scaled is as follows. It is expensive to perform the production process on a large scale, for example, in excess of 10000 euros per run. Many variables contribute to the success of production, but these are not known a priori by every process. Thus, small scale experiments were performed to identify production organisms (clones) and to optimize the production process before moving to large scale. A typical workflow according to the implementation of the production process is as follows:

very small scale (e.g. 15ml or less): clones in representative procedures were identified in the context of a large number of trials (e.g., 48 to 1000);

small scale (e.g. 250ml or 2L): improving the process with an intermediate number of trials (e.g., 24 to 96), wherein the objective is to modify the process until the success factor is maximized;

medium scale (e.g. 50L): initial process transfer, potentially other process improvements;

manufacturing scale (e.g. 1000L): and (5) repeatedly producing the product.

At each stage there is a degree of optimization, such as selecting clones based on best performing clones or selecting optimal air-out conditions. Not all parameters may match between scales in each phase transition, as the transition is non-linear. Specifically, the method comprises the following steps:

the requirements of each stage may be different, e.g. the required sample size is small at small scale (few tests) and then large (e.g. >1mL), but in practice is small relative to the total bioreactor size, or minimizing energy input is not a problem at small scale, but may be a problem at larger scale;

opportunities may differ, since it is cheaper to perform a large number of runs at small scale, and it is relatively easy to change process parameters automatically at small scale;

the constraints may be different: the accuracy of pH control I gassing may be lower at smaller scales, the availability of analysis increases with scale, tolerance to intervention (e.g., sampling) decreases toward manufacturing scale, aspects of the bioreactor change what can be achieved at a given scale (e.g., at larger scales, it becomes increasingly difficult to remove the appropriate heat from the microorganism culture and/or mixing times tend to increase).

In view of the above, it is desirable to make the small scale as representative of the large scale as possible, with as low risk as possible for each stage in the process transition (i.e., minimizing the risk of changing the success criteria), and also taking into account the risk of subsequent steps.

In some examples, the source size may be smaller than the target size, e.g., the source size may be 250mL and the target size may be 2L. In other examples, the source size may be the same as the target size, but the constraints on the production process may be different, e.g., a shorter process time may be desired. Alternatively, the scale may be the same, but the equipment configuration may be different, e.g., BIOSTAT with 3-bladed impellers and 6-bladed impellers

50 BIOSTAT with two 3-bladed impellers

50. In such cases, one of the goals of scaling may be to minimize the chance of obtaining the same performance obtained at a smaller scale at a larger scale.

In still other examples, the source size may be greater than the target size. This may be the case if some of the operations for the optimization process can be done faster and at a lower cost on a smaller scale. Thus, starting from a process that is actually performed at a larger scale, the goal of scaling may be to determine a range of process variants (in terms of steps and/or parameters) that can be run at a smaller scale and then scaled back to the original larger scale with good performance. For example, a downscaling process may be used to select between different clones, i.e. the goal is to find a clone that will perform well when upscaled. Indeed, several clones may be tested on a small scale (e.g., on the order of thousands at a scale of <15mL, or about 50 at a scale of 15 mL).

The method includes retrieving parameter evolution information describing a temporal evolution of a process parameter.

The parameter evolution information characterizes how one or more process parameters change over time, including initial conditions of the process parameters. In particular, the parameter evolution information may comprise empirically derived relationships from previous executions of the production process and/or equations derived through theoretical models regarding the evolution of the production process. The evolution information may include explicit dependencies of one or more parameters on time and relationships that connect process parameters to each other. The parameter evolution information may also include information about variables that do not directly specify steps of the production process, but indirectly affect the process parameters and thus the execution of the production process.

The parameter evolution information may have many components, for example, parameter evolution information related to cell dynamics, bioreactor dynamics, and/or chemical reactions occurring within the bioreactor.

For example, the parametric evolution information may include empirically derived mappings between recipe parameters (such as stirring speed, gassing rate, and fill volume) and dynamic parameters (such as mixing time, kLa, and power input). Additionally or alternatively, the parameter evolution information may comprise theoretically derived or empirically derived equations and starting points for the cell culture model.

The parameter evolution information may also describe events related to and beyond what may be strictly considered as the production process itself. In particular, the parametric evolution information may also describe conditions occurring outside the bioreactor, e.g. the time evolution of the sample taken, or conditions occurring in the secondary reactor vessel or in a downstream processing facility (e.g. a purification unit) and/or in a piece of analytical equipment.

In one example, cell culture is modeled using a system of hierarchical ordinary differential equations describing "cellular processes," where any individual cellular process describes, among other things, the dependence of the cellular process, i.e., how its rate depends on the pH, DO, temperature and concentration of various nutrients, and the results of the process due to the process being in an active state, i.e., changes occurring in cell count, titer, pH, DO, etc.

Allowing a given cellular process a to depend on one or more driving processes X, Y … … allows the rate of a to be calculated and then multiplied by the sum or product of the rates of X, Y. A given cellular process may depend on a non-driven cellular process or any number of driven cellular processes, and a given cellular process may not drive other cellular processes or drive any number of other cellular processes.

In the very simple case, for example, where there is only one dependence (temperature dependence) and one outcome (cell growth), this amounts to solving a differential equation:

where ρ is cell density, T is time, rg is maximum growth rate, T is temperature, Topt is optimal growth temperature, Tsens indicates the sensitivity of growth to temperature, and N (x, s) indicates the value of a normal distribution with the standard deviation at x being s.

More typically, the situation will have considerable dependence (e.g., dependence on key nutrients) and consequences (e.g., decreased DO due to cellular activity, increased temperature in the case of microbial cellular activity, decreased amounts of nutrients, etc.). In addition, many of these cellular processes can be accompanied by additive effects, such as basal growth, death due to the presence of toxins, product accumulation, and the like.

The parameter evolution information of cell growth is equivalent to describing the "cellular process" parameters and their dependence on each other. This may take several forms:

a. list data (such as may exist in a spreadsheet) whereby each cellular process itself has one row, and within each row, provides parameters for various potential dependencies (hard-coded repositories of functional forms about dependencies)

b. Database tables, for example, where there may be tables for:

DB_CELL_CULTURE_MODEL，

DB_CELL_CULTURE_PROCESS，

DB_CELL_CULTURE_PROCESS_LINK，

DB_CELL_CULTURE_PROCESS_DEPENDENCY，

DB_CELL_CULTURE_PROCESS_DEPENDENCY_PARAMETER，

where DB _ CELL _ CURTURE _ MODEL is codified into a named MODEL that can then be referenced by software, DB _ CELL _ CURTURE _ PROCESSES is codified into a CELL CULTURE PROCESS within either MODEL, DB _ CELL _ CURTURE _ PROCESSES _ LINk ties together entries in DB _ CELL _ CURTURE _ PROCESSES to indicate the fact that some processes drive other processes, DB _ CELL _ CURTURE _ PROCESSES _ DEPENDENCY indicates a particular dependency (e.g., in the form of trace variables and dependencies that indicate dependencies), and so on.

c. Structured data formats such as XML or equally JSON or proprietary forms.

These data can then be stored in several ways:

a. in software responsible for performing the scale conversion, e.g. embedded in DLLs or executable files

b. In a file usable for s/w on the file system (the file may provide either of these forms)

c. Within a database instance

The data may then be further stored and accessed:

a. local to the software performing the conversion, e.g. on the same file system, or accessed by a database engine built into the s/w, possibly within the memory, SD card, hard drive, CDROM, DVDROM, etc

b. Within File sharing on a-network accessible to software

c. On a client/server (e.g., web service) system, where the client is physically separate from the server, that is, located close to (physically) the software, or accessible by the software on the network, in the latter case stored in one location or distributed over multiple locations.

In addition to cell culture, the physical dynamics of the process are described in the parametric evolution information. This includes how the parameters relate to each other at a certain point in time and how each parameter evolves over time.

Just like the cell culture model, the model of the physical dynamics may be stored as XML, database tables, etc., and the basis of the data may be DVD, CDROM, hard disk, etc., and the location of the data is local or different, and the distribution of the data is at one location or distribution.

In summary, data representing parameter evolution information may be retrieved according to a number of different implementations. In particular, the data may have been stored as such prior to retrieval, or they may be dynamically computed as needed.

The method includes retrieving a plurality of recipe templates.

The recipe includes a number of steps that define the production process. In other words, a recipe is a structured representation of a production process (e.g., the activity of a bioreactor). This means that the steps are expressed in a structured way (e.g. in a format that can be interpreted by a machine).

As already explained above, there are steps indicating an action (passive or active) and steps controlling the flow, for example, a sequence (performing the included steps in order), a repetition (repeating the included steps until a condition or a given number of times is satisfied), and a selection (performing one step or another depending on the condition).

As discussed, the actual behavior of the step during (actual or simulated) execution, i.e., what happens, is determined by one or more of the process parameters, where these include both recipe parameters and dynamic parameters. However, the recipe may include only recipe parameters, i.e., those that may actually be set to run the process. The process parameters may be expressed as numbers, algebraic expressions, or as a function of other variables.

Thus, a recipe specifically includes a plurality of steps and one or more values of recipe parameters that specify the steps. In other words, recipes specifically provide a well-defined process that can be directly implemented when performing a production process and dictate how to control process equipment over time. Values may be fed to the recipe dynamically during execution, but in either case, it is predetermined which values will be fed.

In contrast, a recipe template is specifically a recipe that specifies a parameter for which at least one of the recipe process parameters for a plurality of steps is variable and initially has no predetermined value.

Thus, the recipe template involves one or more degrees of freedom as to how to perform the production process. Different values of the at least one variable parameter result in different recipes being generated from a given template. The difference may be merely a value of a process parameter or may also be a sequence of steps, for example if the path within the template depends on a variable parameter. Multiple parameters may be free within one recipe template.

Exemplary parameters (i.e., variable parameters) that will vary freely may include, but are not limited to, one or more of the following:

additions to the bioreactor system

Concentration of the Primary carbon Source in the batch Medium

Concentration of primary nitrogen source in the feed medium

pH of the acid added for initial pH control

Buffer capacity of the batch Medium

Cell density in inoculum

Percentage of oxygen in oxygen-enriched air

-with respect to curves and set points

Constant rate of supply of off-gas in the batch phase

Rate of increase of stirring speed over time in the batch phase

P or I parameters in PI feedback loop for gas outlet control

Maximum air flow rate before oxygen supplementation takes place

Exponential coefficient in exponential feeding curve

Initial feed rate in the exponential feed curve

Duration of the plateau phase in the feed curve (e.g., after the exponential phase)

Rate of temperature decrease during induced temperature change

Temperature set point during the batch phase

-on discrete events in the recipe

Bioreactor fill volume (as part of the total bioreactor volume)

Inoculum volume (as part of, e.g., bioreactor fill volume)

On recipe structure and flow control

Cascaded sequences in DO control (e.g. stirring followed by gassing and then O2 supplementation; or

Give vent then stir and then O2 supplement)

Threshold of primary carbon source and/or Dissolved Oxygen (DO) to start the feeding phase

Frequency of sampling

Volume of sample at each sample sampling

Threshold primary carbon source in the sample to cause feed supplementation

Threshold sample density to start harvesting.

Different formulation templates can be used on the granularity of different organisms, clones or production processes, deployment systems (e.g., disposable versus stainless steel reusable production bioreactors). Further improvements may be made if the feedback scaling works better or worse under a particular recipe template.

The recipe templates may also be shared between organizations or categorized on a network in a warehouse. They may be selected based on user feedback regarding zoom success or failure.

The recipe template may be part of a repository and may be retrieved similar to the parameter evolution information. For example, the recipe template may be stored in a structured data format (e.g., maintained by a sequencer in XML). The recipe templates may likewise be stored in databases, spreadsheets, as well as locally in an organization file system, in a database within an organization, or on the cloud (remote).

The recipe and thus the recipe template may include indicia identifying a specific part of the production process. Illustratively, the start marker and the end marker may enclose a portion of the process. The markers may distinguish parts of the production process that are more relevant or critical, e.g. at scaling, as will be explained below with reference to the acceptability function and trajectory comparison.

Both the parameter evolution information and the recipe template may be used in a number of different scale transitions, provided that the production process in question is covered by the step and process parameters considered in the parameter evolution information and the recipe template.

Other inputs required by the method may instead be provided for each specific transition. In practice, the method further comprises receiving: a source mechanism specification for a source mechanism for performing a production process at a source scale, the source mechanism specification comprising a source scale value; a target mechanism specification of a target mechanism for performing a production process at a target scale, the target mechanism specification including a target scale value; a source recipe for the production process at the source scale is defined.

The institutional specification includes information about the institution of the process plant used to perform the production process, first the scale value of the plant, e.g., the capacity of the bioreactor expressed in liters. Further, the mechanism specification may include at least one of the components of the mechanism and its characteristics, e.g., specifying which equipment includes the impeller and optionally which impeller, and so forth. In addition, the product can be identified by reference to the model number of the product (e.g.,) To indicate other characteristics of the device. In other words, the machine specification describes the equipment used to perform the production process at a given scale, and in particular provides the information necessary to simulate the process, as will be discussed below.

In particular, the agency specification may specify values for one or more process parameters (e.g., maximum fill volume, minimum fill volume, maximum agitation speed, maximum outgassing rate, minimum outgassing rate, lower impeller height, upper impeller height, liquid cross-sectional area) that must be fed to the recipe prior to deployment of the recipe and/or from which specific parameter evolution information or recipe templates may be selected.

Examples of source and target agency specifications are: with standard distributor and mammalian impeller

250 bioreactor, and Sartorius with annular distributor and two 3-blade impellers

50。

Additionally, a source recipe is provided. In view of the definitions of the recipes given above, a source recipe is precisely a recipe that includes a number of steps (and relative values of process parameters) that define a process as performed at the source scale.

An example of a source recipe may correspond to the following process: "fill the bioreactor with 0.2L of a given medium; heating to 35 ℃; seeding with clones to a density of 1e6 cells mL-1; incubating and stirring at 600rpm for 36 hours; control pH to 7.4 with bottom and top control, i.e. add acid or base as needed to push pH back to 7.4; maintaining the temperature; gassing with air at a rate of 0.1 total volume per minute; feeding for 36 hours by using composite feed; continuously monitoring and controlling the pH and the temperature; DO is controlled by stirring and gassing; an inducer is added to trigger production. And harvesting after 36 hours. "

Further, the method includes receiving at least one acceptability function for conditions defining values of process parameters at the source scale and/or the target scale. The acceptability function may define conditions for both recipe parameters and dynamic parameters.

The acceptability function is a parameterized function mapping from a process parameter or a combination of process parameters to a value indicative of acceptability. The value of acceptability may be a real number between 0 and 1, i.e. equal to 0 or 1 or greater than 0 and less than 1. Thus, the conditions defined for the values of the process parameters may be binary conditions in the sense that the values are considered acceptable or unacceptable, but they may also be more subtle conditions in which the degree of acceptability of a given value is expressed. In other words, the acceptability function may express an exact constraint on which values are allowed and/or indicate how well a given value fits.

Illustratively, the acceptability functions may be divided into two categories: absolute and relative.

The relative acceptability function provides an assessment of how acceptable a value of a process parameter (illustratively, a value of the same process parameter at another scale) is when that value is considered relative to other quantities. In particular, the relative acceptability function may take into account the values of the process parameters at the source scale and the target scale. The source and target values may be placed differently with respect to each other, for example, taking into account differences or ratios.

An example of a relative acceptability function maps a difference between a source value and a target value of the mixing time to 0 if the mixing time is less at the source than at the target, and to 1 otherwise. Another example maps the difference between Power Per Volume (PPV) values at the source and target scales to a normal distribution.

In other examples, the acceptability function may be a function of two or more process parameters. For example, the difference in cell density and products of cellular activity between the source scale and the target scale can be mapped to a normal distribution.

In contrast, an absolute acceptability function provides an assessment of the degree of acceptability of a value of a process parameter when that value is considered independently.

An example of an absolute acceptability function maps a stirring speed between 0% and 5% or between 95% and 100% of the maximum possible stirring speed (in view of the target bioreactor) to 0 and between 5% and 95% of the maximum possible stirring speed to 1. Another example maps PPV to a normal distribution around a given maximum.

For example, the at least one acceptability function may be a plurality of acceptability functions including at least one absolute acceptability function and at least one relative acceptability function.

The acceptability functions may be grouped according to the scale they are applicable to, for example, three groups: a small scale module, a medium scale module and a large scale module. As noted above, certain scales may be divided into more than one group. There may be more than three different gauge sets.

The agency specifications, source recipes, and one or more acceptability functions may be received as input from an external source (e.g., a user and/or a control system configured to perform a production process).

The method further comprises the following steps: simulating execution of the production process at the source scale using the source agency specification, the source recipe and the parameter evolution information; and determining one or more source trajectories of the process parameters from the simulation, wherein the trajectories correspond to time-based curves of values recordable during execution of the simulated production process.

The simulation of the execution of the production process is a simulation of the execution of the production process in the real world, performed by means of a computer system. The source agency specifications and source recipes provide initial conditions for the simulation and a description of the process to be simulated, while the parametric evolution information models the evolution of the process over time.

In particular, it is possible to derive values of process parameters at different times during the evolution of the process in order to obtain a trajectory. Thus, a plurality of source trajectories corresponding to a plurality of process parameters, as evolved when performing a process at a source scale, may be obtained. In some examples, trajectories may be determined for both recipe parameters and dynamic parameters. In other examples, the trajectory may be determined for only the dynamic parameters.

Each trajectory may be understood to summarize and provide an overview of the associated process parameters. Each trajectory may be implemented as a curve or graph describing the time evolution of a process parameter during the execution of the simulated production process. In particular, each trajectory may comprise a plurality of points representing values of the parameter corresponding to different moments in time. For example, the time unit between successive points may be one hour.

The method further comprises an execution targeting step, wherein execution of the production process is simulated on a target scale. To perform the simulation, one of a plurality of recipe templates is selected and input values for at least one variable parameter in the selected recipe template are provided.

The combination of the selected recipe template and the input values for the variable parameters provides a target-scale recipe that can be used for the simulation, similar to how the source recipe is used to perform the simulation at the source scale. When there are a plurality of variable parameters, a corresponding plurality of input values, i.e., an input value of each variable parameter, is provided so that the recipe is completely specified.

The recipe templates may be selected among all available recipe templates or within a subset of recipe templates associated with a particular zoom. "associated" herein may mean: based on organizational and/or procedural knowledge, it is appropriate to deploy the procedure or to the organism. The selection may be performed by the user or automatically based on, for example, flags present in the recipe template and indicating their suitability.

The input value may be a physical guess based on process knowledge, and it may be part of a set of possible values associated with a particular recipe template. For example, the recipe template may include one or more candidate values and a test range for each value, such that the input value may be selected as any value within an interval around the candidate value. Thus, candidate values may be retrieved along with the recipe template. Alternatively, the input value may be supplied by the user.

The target facility specification, the selected recipe template, the input value for the at least one variable parameter, and the parameter evolution information are used to simulate the execution of the production process at the target scale. One or more trajectories of process parameters are then determined, similar to what is done for the source scale.

The target determining step further comprises comparing the source trajectory with the target trajectory. If only one source trajectory for a given process parameter is determined, only the corresponding target trajectory for the same process parameter may be determined and the two may be compared. If multiple trajectories of source and target sizes are determined, the trajectories are compared in pairs, i.e., the source trajectory for a given process parameter is compared with the target trajectory for the same process parameter.

Illustratively, the comparison may be performed by comparing points on the target trajectory with corresponding points on the source trajectory. Each trace represents a numerical description of a process parameter over time such that each point on the trace is a value of the process parameter at a particular time.

The comparison may, for example, comprise calculating differences between point values in the target trajectory and corresponding (i.e. simultaneous) point values in the source trajectory for the same parameters. Other quantitative evaluations of the comparison may be performed, such as taking a ratio of values or combining values according to a given relationship.

The targeting step further includes calculating an acceptability score for the selected recipe template based on the comparison and the at least one acceptability function. In particular, an acceptability score is assigned to a combination of the selected recipe template and the provided (initial) value of the at least one variable parameter.

The acceptability score indicates how well the production process at the target scale fits the conditions of the acceptability function. In other words, the acceptability score provides an assessment of the combination of values of the recipe template and the variable parameters according to the acceptability criteria set in the acceptability function.

As explained above, the acceptability function maps the process parameters, or a combination thereof, to acceptability values. Accordingly, one or more applicable acceptability functions are applied to the target trajectories for the corresponding process parameters in order to obtain an acceptability score for the combination of the values of the variable parameters and the recipe template. In other words, the suitability of the recipe template plus the value of the variable parameter for performing the process at the target scale is evaluated via the process parameter trajectory.

By "applicable" it is meant that the acceptability function defines the conditions for the process parameters corresponding to the source trajectory and/or the target trajectory. It should be noted that the process parameters of the trace may or may not coincide with the variable parameters. In other words, the acceptability score may express the evaluation of the value of the variable parameter in an indirect way, provided that there is a relationship between the trajectory process parameter and the variable parameter, i.e. they are not completely independent of each other.

In particular, when applying an absolute acceptability function to the target trajectory, a comparison with the source trajectory may not be required, i.e. "comparison based" may be an optional feature of calculating (at least part of) the acceptability score. In other words, the absolute acceptability function may be applied to the values corresponding to the points in the trajectory. When applying the relative acceptability function to the target trajectory, the acceptability function may be applied to the comparison between the target trajectory and the corresponding source trajectory. In other words, a relative acceptability function may be applied to comparisons between values corresponding to points in the trajectory, such as pairwise differences. Furthermore, there may be a higher dimensional acceptability function that requires a comparison between multiple source trajectories (e.g., pH and DO) and corresponding (i.e., also pH and DO) target trajectories.

In particular, if there is only one acceptable acceptability function, this is applied to the target trajectory to obtain an acceptability score. In some implementations, this includes calculating an acceptability value for each time point in the target trajectory and accumulating the acceptability values for different time points to obtain an acceptability score. F for example, accumulation is done by considering the mean (algebraic or geometric) or median or product of the acceptability scores, which may be, for example, S1 ═ S (t) > t. Other combinations may be possible.

When there is more than one applicable acceptability function, the acceptability score obtained when applying the single acceptability function to the target trajectory may be a partial acceptability score. All applicable acceptability functions are applied to the target trajectory and the partial acceptability scores obtained by each of them are accumulated, where the accumulation may be done in any of the ways discussed above. For example, the acceptability score may be.

If only one target trajectory exists, the acceptability score obtained by applying all applicable acceptability functions represents the acceptability score of the overall target recipe.

When more than one target trajectory exists, the acceptability scores may be further accumulated. In other words, in the case of multiple trajectories, an acceptability score is calculated for each target trajectory, and these are then accumulated to obtain an acceptability score for the provided values of the selected recipe template and variable parameters.

Illustratively, the scores obtained for different tracks are accumulated by calculating an index of the average of the logarithms of the scores on the tracks. In other words, ". This ensures that any process that did not perform (acceptability score ═ 0) for any part of the culture will have a cumulative acceptability score of 0, and any process that performed perfectly (acceptability score ═ 1) for all cultures will have a cumulative acceptability score of 1. There are alternative ways of calculating the cumulative acceptability score for a recipe template, such as those discussed above.

In some examples, an acceptability value may be calculated for each time point in the trajectory, except that the total cumulative acceptability score for the recipe template is, for example, a real number x (where 0 ≦ x ≦ 1), and the accumulation may be done separately for each time point. The result will be the function x (t).

In summary, when a plurality of acceptability functions are received and a plurality of target trajectories are calculated, calculating an acceptability score may include: obtaining, for each target trajectory of the plurality of target trajectories, a second partial acceptability score by:

-selecting one or more applicable acceptability functions;

-for each applicable acceptability function, performing the following calculation steps:

-calculating an acceptability value based on an acceptability function for each time point in the target trajectory;

-accumulating acceptability values for different time points to obtain a first partial acceptability score;

-if there is a single applicable acceptability function, setting the second part acceptability score to the first part acceptability score;

-if there are a plurality of applicable acceptability functions, accumulating the first part of acceptability scores of all applicable acceptability functions to obtain a second part of acceptability scores;

and accumulating the second portion acceptability scores of all target trajectories to obtain an acceptability score.

As explained above, the recipe template can include indicia identifying a particular part of the production process. Illustratively, the start marker and the end marker may enclose a portion of the process. The markers allow the acceptability function to be applied to some or all of the traces.

For example, in a recipe template, the start of a batch phase and the end of a batch phase may be marked. When a marker is present in the recipe template, all or a subset of the acceptability function should be applied, for example, only to the interval between starting and ending batches to the target trajectory. In other words, some acceptability functions may be applied to only one phase, e.g. the batch phase, and other acceptability functions may be applied to only another phase, e.g. the feed phase. For example, during the batch phase, outgassing may be set to be constant in the recipe template, and an acceptability function that relies on the percentage of dissolved oxygen for acceptability scores is only applied to batch phase intervals in the trajectory, as the goal is to ensure constant outgassing results in a similar dissolved oxygen environment during the batch phase.

When a (relative) acceptability function is applied to the comparison between the source and target trajectories between two markers, the time is optionally scaled between the source and target trajectories so that the intervals between the markers are the same.

For example, in a simulation at source scale, the batch phase starts at t-2 and ends at t-12; at the target scale, the batch phase starts at t-2 and ends at t-22. In the track comparison between these markers, either t-5 in the source track is compared to t-5 in the target track (i.e., no scaling), or t-5 in the source track is compared to t-8 in the target track (i.e., scaling is performed, 2 in each unit value source in the target, 3 time units have passed between the start of the batch).

Finally, the targeting step includes calculating an optimal value for the at least one variable parameter in the selected recipe template by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, wherein values within the acceptable range result in an acceptability score above a particular threshold.

The acceptability score, calculated as explained above, yields a number whose value depends inter alia on the value of the variable parameter assigned to the selected recipe template. Thus, the acceptability score may be viewed as a function of one or more variable parameters, hereinafter referred to as the "score function". In the case of multiple variable parameters present in the selected recipe template, the scoring function is a multivariate function of the variable parameters.

The scoring function may be optimized to find the best input values for the variable parameters, i.e., those values where the process at the target scale is most similar to the process at the source scale and the criteria specified according to the acceptability function are most successful. Finding the best value of the variable parameter is an optimization problem, where the scoring function is maximized or minimized by systematically selecting input values from within the allowed set and calculating the value of the scoring function. Depending on the number of variable parameters, the optimization problem may be a multi-dimensional problem. The allowed groups may be specified in the recipe template, as described above, or may be otherwise determined. In some cases, there may be more than one optimal value of the optimization score function. If the score function takes a real value between 0 and 1, and 1 is assigned to perfect acceptability, then the score function must be maximized. Examples of optimization algorithms are the Nelder-Mead method and the steepest descent method. The result of the optimization is to find the parameter value or combination of parameter values that gives the best (e.g., highest score) for the selected recipe template. In some cases, it may be necessary to optimize using multiple initial values, since the space being explored is highly non-linear and may present multiple local optima.

In addition to or instead of finding the optimal values of the variable parameters, acceptable ranges may be found by constraining the score function to achieve values above a specific predetermined or predeterminable threshold. There may be more than one acceptable range for a given variable parameter. The threshold may be input by the user or set or derived based on an optimal value available for the acceptability score, e.g. a fraction of the optimal value, such as 70%, 80% or 90%, or based on an absolute criterion of the acceptability score, e.g. must be at least 0.5 or 0.6 or 0.7. In particular, the term "range" should be interpreted broadly to encompass multi-dimensional results.

Thus, the result of the targeting step may be a single point in the variable parameter space and/or a curve or surface in the variable parameter space. Mathematically, in the exemplary case, this would correspond to a combination of a set of arbitrary dimensional manifolds.

However, as explained above, a given acceptability score is not only a result of the input values fed to the selected recipe template, but also a selection of the recipe template itself. Thus, if there is more than one related recipe template, the targeting step may be repeated for at least one other recipe template. Thus, the step of "repeating the targeting step for at least one other related recipe template" may be optional. In particular, the targeting step may be repeated for all recipe templates in the library or only for recipe templates in a subset of the relevant recipe templates discussed above.

The method also includes selecting at least one of the plurality of recipe templates as a target recipe based on the acceptability scores calculated for the one or more recipe templates. In particular, one or more recipe templates selected as target recipes are selected along with the optimal input values and/or acceptable ranges of input values.

Once the targeting step that should be performed has been completed for all recipe templates, each recipe template has a (best) acceptability score associated with the best input value and/or a number of acceptability scores above a given threshold associated with a range of input values.

Thus, one or more candidates for a target recipe that will specify how to perform a process at a target scale for actual deployment may be selected among the combination of recipe templates and input values based on an acceptability score calculated at optimization and/or when setting a minimum threshold. In some examples, a single target recipe may be selected, while in other examples, multiple target recipes may be given. In particular, a single recipe template with multiple acceptable input values (e.g., more than one optimal value or acceptable range) for a single variable parameter may result in multiple target recipes. Similarly, multiple recipe templates with good acceptability scores may result in multiple target recipes.

According to an exemplary implementation, the method may further include outputting the one or more target recipes in the form of a recipe template plus input values. Additionally, a corresponding "best" predicted trajectory and/or acceptability score at the target scale may be output. In particular, if a time-dependent acceptability score x (t) has also been calculated, as discussed above, this may also be output.

In some implementations, the target formulation obtained by the selection explained above may be cast in the following manner: the automation system can handle for example

15 or

250 "experimental draft". The method may further comprise: providing a target recipe to a target control system; and executing, by the target control system, the production process at the target scale based on the target recipe. The identified scaling process conditions may be manually or automatically transferred between bioreactor configurations.

In other words, a target recipe with relative input values as selected by the above-described method may be fed directly to a target control system configured to control a target facility device to implement a process at a target scale. If more than one target recipe (see above) is provided to the target control system, the target control system may select only one target recipe, for example based on some a priori, such as the existing configuration of the target facility, or may perform the production process multiple times.

When the production process is performed in the real world at a target scale, the results may be used to provide feedback. In some implementations, the method may further include: evaluating the performance of the production process at a target scale; and modifying at least one acceptability function based on the evaluation.

The performance of the production process may be evaluated, for example, by using the quality attributes defined above. Modifying at least one acceptability function based on the evaluation may include assigning weights to the acceptability function such that they become more or less relevant when determining the acceptability score. For example, if scaling results in good performance, the involved acceptability functions may be given greater weight. Conversely, if scaling results in poor performance, the involved acceptability functions may be given lower weight. In other examples, the form of the acceptability function may be modified, for example, by replacing the normal distribution with a different distribution or by changing the range of values of the process parameter that is mapped to 0 by the acceptability function.

Another example for feedback is to use predicted target and actual trajectories and modify the parameters for the simulation when there is a difference between the predicted and actual trajectories. The method of modifying these parameters is a non-linear fit, i.e. minimizing the difference between observed and expected.

Yet another feedback mechanism may be implemented, such as to improve parameter evaluation information.

In some implementations, a source recipe that defines a production process at a source scale can be obtained by:

defining a target quantity characterizing the production process at the source scale;

performing, by the source control system, the production process a plurality of times at the source scale while changing the process parameters and/or the process steps;

the process is selected based on the results of the target quantities given by the process parameters and process steps in order to define the source recipe.

The target amount may be related to other aspects of the product and/or process. For example, the target amount may be one quality parameter or a combination of quality parameters, or may be otherwise defined. The production process may be repeated multiple times because, for example, on a small scale, it is less expensive to find the best combination of process steps specified by a given process parameter, where the best combination is the combination that optimizes (e.g., maximizes) the target quantity.

Examples of target quantities to be maximized include, but are not limited to, one or more of the following:

total amount or amount per unit volume

Quality (chemical composition of the product, protein structure (primary, secondary and tertiary structure), glycosylation pattern)

Purity (amount of product relative to similar molecule)

Release of product from the cell wall, cytoplasm or other cell compartment

-a release conducive to the lysis of the cells of the product released into the medium

Throughput (i.e., the reciprocal of cycle time, i.e., minimizing incubation period)

Ability to detect process deviations early and correct (e.g. possibly pointing to increased sampling)

Examples of target amounts to be minimized include, but are not limited to, one or more of the following:

cell debris

Presence of molecules, cellular components or cells interfering with decontamination

Shear or chemical damage to the product

Media cost (total media used or expensive components of media)

Energy costs (especially on a larger scale)

Risk of process failure (e.g. the extent to which small fluctuations in e.g. bioreactor or cell performance may cause the process to leave the operating area)

The recipe scaling method described above for scaling from a source scale to a target scale may be generalized for so-called queue scaling, i.e., scaling from a source scale to a final target scale through one or more intermediate target scales. Conventionally, each transition in queue scaling is processed separately. Instead, according to the invention and in particular due to the acceptability function, queue scaling is performed taking into account all transfers from one size to another size at the same time. In other words, queue scaling is considered a single composite process, rather than manually breaking it into pairwise scaling. Thus, the chances of a process being successfully executed at any subsequent scale increase.

The method described above naturally extends to queue scaling scenarios. When applying the scaling method to a queue scaling comprising one or more intermediate target scales, one or more additional simulations need to be run for the one or more intermediate target scales, requiring corresponding intermediate target mechanisms and, in particular, corresponding input values for variable parameters in the recipe template. In fact, one of the goals of the method for queue scaling is to obtain a target recipe for each of the intermediate target size and the final target size. This means that once the relevant recipe template is selected, its variable parameter or parameters are temporarily filled in by one or more sets of input values, respectively. In other words, for each variable parameter, a plurality of input values are provided, each corresponding to a target scale. For example, in the case of two intermediate target scales between the source scale and the final target scale, a first input value, a second input value, and a third input value are provided for the variable parameter; the first input value may provide a physical guess of the value of the variable parameter at a first intermediate target scale, the second input value may provide a physical guess of the value of the variable parameter at a second intermediate target scale, and the third input value may provide a physical guess of the value of the variable parameter at a final target scale. Thus, the simulation run for each scale is based on a combination of the recipe template and the corresponding input values provided for that scale.

Based on the plurality of simulations, trajectories of intermediate target scale are determined in addition to those of the source scale and the final target scale. The traces at different scales are then compared in pairs and also in combinations of higher cardinality (e.g., three traces at three different scales are compared together, or four traces at four different scales are compared together, etc.). For example, the comparison of the three trajectories may, for example, include calculating a difference between the point values in the final target trajectory and a sum of corresponding (i.e., simultaneous) point values in the source trajectory and the intermediate target trajectory of the same parameter. Other quantitative assessments of the comparison may be performed.

The at least one acceptability function for queue scaling may specifically define conditions for values of process parameters at any single scale and/or at any number of scales. In particular, there may be an absolute acceptability function defining the conditions of the process parameters at any single scale alone, and/or there may be a relative acceptability function defining the relationship between values on two or more different scales. In this case, the score will be derived as an accumulation of the output of the acceptability function from a plurality of scale combinations.

Finally, similar to the two-scale transition, the optimization problem is solved. The difference is that the dimension of the optimization problem is higher in the queue scaling. If one considers a simple example where the selected recipe template has only one variable parameter, then the dimension of the optimization problem for that recipe template that accepts the scoring function is one. In the case of queue scaling, the dimensionality increases according to the number of intermediate target scales. If there is an intermediate target size, the optimization process must find the optimum value of the variable parameter for the intermediate target size and the optimum value of the variable parameter for the final target size at the same time, i.e., the presence of an intermediate target size increases the dimensionality of the problem by one. If there are two intermediate target scales, then for the simple example above, the dimension is increased by two. Generally, if the dimension of the optimization problem in the two-scale transition is D1 and the number of intermediate target scales plus final target scales is T, the dimension of the optimization problem in the queue scaling with T-1 intermediate target scales will be D2 ═ D1 × T. Therefore, the number of "constraints" imposed by the acceptability function must be sufficient so that an optimization problem is determined. This may for example be translated into optimizing at least two acceptability score functions simultaneously, which may for example be combined by multiplication. It may also result in a range of intermediate-scale possibilities for the output of the optimization, all of which share the maximum of the cumulative score.

Similar considerations apply to the acceptable ranges in the case of due differences.

Thus, at least one result of the method for queue scaling is to provide one or more target recipes for each of the intermediate target size and the final target size. All other aspects illustrated for the two-scale transition, such as outputting the target recipe and also outputting the acceptability score and predicted trajectory, and feedback mechanisms, where applicable, may be implemented for queue scaling.

A retrofit of the method of queue scaling with one intermediate target size is found below, however, it is apparent that it can be extended to any number of intermediate target sizes.

A computer-implemented method of scaling a production process for producing a chemical, pharmaceutical and/or biotechnological product from a source scale to an intermediate target scale to a final target scale, wherein the production process is defined by a plurality of steps specified by one or more process parameters controlling the execution of the production process, the method comprising:

-retrieving:

-parameter evolution information describing the time evolution of a process parameter;

-a plurality of recipe templates, wherein:

the recipe includes a plurality of steps that define a production process, and

a recipe template is a recipe that specifies parameters for which at least one of the process parameters for a plurality of steps is variable and initially has no predetermined value;

-receiving:

-a source mechanism specification for a source mechanism for performing a production process at a source scale, the source mechanism specification comprising a source scale value;

-an intermediate target mechanism specification for an intermediate target mechanism for performing a production process at an intermediate target scale, the intermediate target mechanism specification comprising an intermediate target mechanism specification value;

-a final target mechanism specification for a final target mechanism for performing the production process at a final target scale, the final target mechanism specification comprising a final target mechanism specification value;

-a source recipe defining a production process at a source scale;

-at least one acceptability function that defines conditions for values when considering values of the process parameter at any one of the source scale, the intermediate target scale and the final target scale singularly and/or when considering values at any one scale relative to corresponding values at any other scale or scales;

-simulating the execution of the production process at the source scale using the source authority specification, the source recipe and the parameter evolution information;

-determining one or more source trajectories of process parameters from the simulation, wherein the trajectories correspond to time-based curves of values recordable during the execution of the simulated production process;

-performing a targeting step, the targeting step comprising:

-selecting a recipe template associated with said production process from said plurality of recipe templates;

-providing a first input value for at least one variable parameter in the selected recipe template;

-providing a second input value for at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the intermediate target scale using the intermediate target mechanism specification, the selected recipe template, the first input value of the at least one variable parameter and the parameter evolution information;

-determining one or more intermediate target trajectories of process parameters from the simulation;

-simulating the execution of the production process at the final target scale using the final target mechanism specification, the selected recipe template, the second input value of the at least one variable parameter and the parameter evolution information;

-determining one or more final target trajectories of the process parameters from the simulation;

-performing a first, second and third pair-wise comparison between any two of the source trajectory, the intermediate target trajectory and the final target trajectory, and performing a three-way comparison between the source trajectory, the intermediate target trajectory and the final target trajectory;

-calculating an acceptability score based on the at least two comparisons and the at least one acceptability function;

-calculating a first optimum and a second optimum of at least one variable parameter by optimizing the acceptability score, and/or

Calculating a first acceptable range and a second acceptable range for the at least one variable parameter, wherein values within the first acceptable range and values within the second acceptable range yield an acceptability score that is above a particular threshold;

-if there is at least one further related recipe template, repeating the targeting step for the at least one further related recipe template;

-selecting at least one of the plurality of recipe templates and the corresponding calculated value of the variable parameter as the target recipe based on the calculated acceptability score for the one or more recipe templates.

In particular, calculating an acceptability score may include calculating any combination of:

-a first acceptability score of the selected recipe template based on the first pairwise comparison and the at least one acceptability function;

-a second acceptability score for the selected recipe template based on the second pairwise comparison and the at least one acceptability function;

-a third acceptability score for the selected recipe template based on the third pairwise comparison and the at least one acceptability function;

-a fourth acceptability score for the selected recipe template based on the three-way comparison and the at least one acceptability function;

and calculating the optimal value/acceptable range may include:

calculating a first optimal value and a second optimal value for at least one variable parameter in the selected recipe template by simultaneously optimizing at least two of the first acceptability score, the second acceptability score, the third acceptability score, and the fourth acceptability score, and/or

Calculating a first acceptable range and a second acceptable range for the at least one variable parameter, wherein values within the first acceptable range yield any one of a first acceptability score, a second acceptability score, a third acceptability score, and a fourth acceptability score that are above the respective first specific threshold, second specific threshold, third specific threshold, or fourth specific threshold, and values within the second acceptable range yield any other one of the first acceptability score, the second acceptability score, the third acceptability score, and the fourth acceptability score that are above the respective first specific threshold, second specific threshold, third specific threshold, or fourth specific threshold.

To better illustrate how the queue scaling method is just an extension of the two-scale method, reference is made to the following wording.

A computer-implemented method for transitioning from a source scale to a target scale, wherein:

-the target size is the final target size;

-there is an intermediate target scale;

-the at least one acceptability function further or alternatively defines the conditions for the values of the process parameter at the intermediate target scale and/or the values of the process parameter at any two scales with respect to each other and/or the values of the process parameter at all scales with respect to each other;

-the input value of the at least one variable parameter is a first input value;

-the comparison between the source trajectory and the final target trajectory is a first pair-wise comparison;

-the optimum value of the at least one variable parameter is a first optimum value and the acceptable range of the at least one variable parameter is a first acceptable range;

wherein the method further comprises:

-retrieving an intermediate target mechanism specification for an intermediate target mechanism for performing the production process at the intermediate target scale, the intermediate target mechanism specification comprising an intermediate target mechanism specification value; and is

The target determining step further comprises:

-providing a second input value for at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the intermediate target scale using the target agency specification, the selected recipe template, the second input value of the at least one variable parameter and the parameter evolution information;

-determining one or more intermediate target trajectories of process parameters from the simulation;

-making a second pair-wise comparison between the source trajectory and the intermediate target trajectory, and making a third pair-wise comparison between the intermediate target trajectory and the final target trajectory, and making a three-way comparison between the source trajectory, the intermediate target trajectory and the final target trajectory;

wherein calculating an acceptability score is performed based on any combination of the first pair-wise comparison, the second pair-wise comparison, the third pair-wise comparison, and the third-party comparison, and

wherein the first optimum is calculated by optimizing the acceptability score simultaneously with/in conjunction with the second optimum for the at least one variable parameter in the selected recipe template, and/or

A first acceptable range is calculated simultaneously with/in conjunction with a second acceptable range for the at least one variable parameter, where values within the first acceptable range and values within the second acceptable range yield an acceptability score that is above a particular threshold.

In another aspect of the invention, a computer-implemented method of scaling a state of a production facility for a production process to produce a chemical, pharmaceutical and/or biotechnological product is provided, the scaling being from a source scale to a target scale, wherein the state is defined by a set of state parameters describing a condition and/or behavior of the production facility. This approach is also referred to as instantaneous or point-in-time scaling and achieves the goal of identifying the best matching parameters for a given point in time between processes at one scale and another.

The status of the production facility may indicate static or dynamic conditions, e.g., how much the bioreactor is filled, and/or the behavior of the production facility, e.g., the agitation speed at which the impeller operates. A state parameter is a quantity that defines such a condition/action, in particular by quantification when a given value is taken. Since the state of a production facility is also affected by its environment, in some cases, the state of a parameter may include a parameter that is indirectly related to the production facility by describing the environment.

The status parameters may be set by an operator or a control system, for example. Illustratively, the state parameters of the bioreactor may include, but are not limited to: stirring speed, outgassing, fill volume, and instructions on which gases to add. In particular, the status parameters may inhibit or be a suitable subset of the recipe parameters defined above.

Although in the following, reference will be made mainly to a plurality of status parameters, the group may also have dimension 1, i.e. there may be only one status parameter.

The method includes retrieving mapping information describing how the state parameters relate to a set of derived parameters.

The mapping information may include theoretical equations and/or fitting relationships that relate one or more state parameters to one or more derived parameters. Although the status parameters are instead related to how the production equipment is set, the derived parameters are more related to the production equipment in the context of performing the production process with the production equipment. In other words, the derived parameter may be a quantity that occurs during the production process and that may not be controlled by external inputs.

Illustratively, the derived parameters of the bioreactor may include, but are not limited to, one or more of: tip speed, kLa, mixing time, power input, reynolds number, froude number, minimum swirl size, and superficial gas velocity. In particular, the derived parameters may be suppressed or be a suitable subset of the dynamic parameters defined above.

It should be noted that the derived parameters provide a static description, i.e. their values are considered at a given point in time rather than evolving during the process. Thus, the export parameter represents a "snapshot" of the process. In scale shifting in this mode of operation, the user can take advantage of a series of time points or "snapshots" within the process and ensure accuracy of the scale shifting in all of these.

In addition, the set of derived parameters may have dimension 1.

The mapping information may illustratively depend on the specific production device such that it may comprise, for example, a plurality of relationships between the same state parameter and the same derived parameter, with each different relationship applying to the specific production device.

The mapping information may be retrieved in any of the ways described with respect to the parameter evolution information.

The method also includes receiving:

-a source authority specification for a source authority for performing the production process at the source scale, the source authority specification comprising a source scale value;

-a target mechanism specification for a target mechanism for performing the production process at the target scale, the target mechanism specification comprising a target scale value;

-a first set of state parameters at source scale;

-a second set of state parameters at a target scale, wherein at least one of the state parameters at the target scale is a variable parameter having initially no predetermined value;

-at least one acceptability function defining conditions with respect to values of the state parameter and/or values of the derived parameter at the source scale and/or the target scale.

The institutional specifications are the same as those discussed above for recipe scaling. In addition, the range and form of the acceptability function is the same as discussed above, in particular there may be relative and absolute acceptability functions. The relative acceptability function may define conditions for values of the state parameter at the source and target scales and/or conditions for values of the derived parameter at the source and target scales.

In addition, a first set of status parameters (i.e., values thereof) is provided, which describes the status of the production equipment at the source scale. Similarly, a second set of state parameters for the target scale is provided. However, the state at the target scale is not completely determined, because one of the goals of the method is to find one or more optimal states at the target scale that correspond to the state at the source scale. Thus, at least one of the state parameters for the target scale in the second set is free in the sense that the value of the state parameter is not fixed but can be modified in order to achieve an optimal solution. The effect of the variable parameters in the set of state parameters is the same as the effect of the variable parameters in the recipe template above.

The method further comprises the following steps: calculating a first set of derived parameters at the source scale using the first set of state parameters, the source authority specification, and the mapping information; providing an input value for at least one variable parameter of the second set of state parameters; a second set of derived parameters at the target scale is calculated using the second set of state parameters, the input values, the target mechanism specification, and the mapping information.

As explained, the mapping information allows derived parameters to be derived from the state parameters. Thus, using the mapping information and the source authority specification, a first set of derived parameters at the source scale can be calculated. The calculation can be performed analytically or by means of numerical simulation.

Source agency specifications may be required to select the appropriate relationships for a given source production facility. Additionally or alternatively, some of the derived parameters may be obtained based on features/quantities contained in the machine specifications, such as with respect to the geometry of the production equipment.

Similarly, a second set of state parameters is used with one or more "guess" input values for one or more variable parameters to derive a second set of derived parameters.

The method further comprises comparing the first set of state parameters with the second set of state parameters and/or comparing the first set of derived parameters with the second set of derived parameters.

The comparison may, for example, comprise calculating the difference between a state parameter in the second set (i.e. at the target scale) and the same parameter in the first set (i.e. at the source scale). The same is true for deriving the parameters. Other quantitative evaluations of the comparison may be performed, such as taking the relative difference, i.e., the absolute value of (value in source-value in target)/(maximum (value in source, value in target)), the ratio of values, or combining the values according to a given relationship.

Finally, the method comprises: calculating acceptability scores for the second set of state parameters based on the comparison and at least one applicable acceptability function; an optimal value for the at least one variable parameter is calculated by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, where values within the acceptable range yield an acceptability score above a particular threshold.

The optimization of the acceptability score function corresponds to the optimization described for the recipe scaling. In this case, the acceptability score function is a function of at least one variable parameter in the second set of state parameters. Since no trajectory exists, the calculation of the acceptability score is simplified.

If there are multiple acceptability functions and multiple variable parameters, calculating an acceptability score includes:

obtaining, for each relevant variable parameter of the plurality of variable parameters, a partial acceptability score by:

-selecting one or more applicable acceptability functions;

-calculating an acceptability value for each applicable acceptability function;

-if there is a single applicable acceptability function, setting the acceptability value as a partial acceptability score;

-if there are a plurality of applicable acceptability functions, accumulating the acceptability values of all applicable acceptability functions to obtain a partial acceptability score; and accumulating the partial acceptability scores for all variable parameters to obtain an acceptability score.

The accumulation may be done as explained previously, e.g. by taking an arithmetic or geometric mean.

Thus, one of the results of this approach is an optimized set of state parameters for the target scale. This can be exported and, for example, automatically fed to a control system for setting up the production facility at the target scale. Other results that may be output include a second set of derived parameters at the target scale and an acceptability score.

It is clear that the time point scaling can be described as a special case of recipe scaling, where some special conditions apply. In particular, the parameter evolution information includes only the relationships between parameters and no equations for time evolution, and the recipes and recipe templates include only parameters and no process steps.

Thus, the same principles explained above apply to extending the point-in-time scaling to the queue scaling. In a particular case of an intermediate target size, the method will comprise:

-retrieving mapping information describing how the state parameters relate to a set of derived parameters;

-receiving:

-a source mechanism specification for a source mechanism for performing a production process at a source scale, the source mechanism specification comprising a source scale value;

an intermediate target mechanism specification for an intermediate target mechanism for performing a production process at an intermediate target scale, the intermediate target mechanism specification comprising an intermediate target gauge value;

-a final target mechanism specification for a final target mechanism for performing the production process at a final target scale, the final target mechanism specification comprising a final target mechanism specification value;

-a first set of state parameters at source scale;

-a second set of state parameters at an intermediate target scale, wherein at least one of the state parameters at the intermediate target scale is a variable first parameter having initially no predetermined value;

-a third set of state parameters at a final target scale, wherein at least one of the state parameters at the final target scale is a second parameter that is variable and initially has no predetermined value;

-at least one acceptability function defining constraints on the values of the state parameters and/or the values of the derived parameters at the source scale and/or at the intermediate target scale and/or at the final target scale;

-calculating a first set of derived parameters at the source scale using the first set of state parameters, the source authority specifications and the mapping information;

-providing a first input value for at least one first variable parameter of the second set of state parameters;

-calculating a second set of derived parameters at the intermediate target scale using the second set of state parameters, the first input values, the intermediate target mechanism specification and the mapping information;

-providing a second input value for at least one second variable parameter of the third set of state parameters;

-calculating a third set of derived parameters at the final target scale using the third set of state parameters, the second input values, the final target mechanism specification and the mapping information;

-performing a plurality of pair-wise comparisons within any two of all pairs of the first, second and third set of state parameters and/or within any two of all pairs of the first, second and third set of derived parameters, and performing at least one three-way comparison between the first, second and third set of state parameters and/or between the first, second and third set of derived parameters;

-calculating an acceptability score based on the at least two comparisons and the at least one acceptability function;

-calculating a first optimal value of at least one first variable parameter and a second optimal value of at least one second variable parameter by optimizing the acceptability score, and/or

A first acceptable range for the at least one first variable parameter and a second acceptable range for the at least one second variable parameter are calculated, wherein values within the first acceptable range and values within the second acceptable range yield an acceptability score above a particular threshold.

It is clear that the above method can be generalized to any arbitrary number of intermediate target scales.

According to another aspect, a computer program product is provided. The computer program product comprises computer readable instructions which, when loaded and executed on a computer system, cause the computer system to perform the operations as described above. The computer program product may be tangibly embodied in a computer-readable medium.

According to yet another aspect of the present invention, a computer system is provided that is operable to scale a production process from a source scale to a target scale to produce a chemical, pharmaceutical and/or biotech product. The production process is defined by a plurality of steps specified by one or more process parameters that control the execution of the production process, and the computer system comprises:

-a retrieval module configured to retrieve:

-parameter evolution information describing the time evolution of a process parameter;

-a plurality of recipe templates, wherein:

the recipe includes a plurality of steps that define a production process, and

a recipe template is a recipe that specifies parameters for which at least one of the process parameters for a plurality of steps is variable and initially has no predetermined value;

-a receiving module configured to receive:

-a source mechanism specification for a source mechanism for performing a production process at a source scale, the source mechanism specification comprising a source scale value;

-a target mechanism specification for a target mechanism for performing a production process at a target scale, the target mechanism specification comprising a target gauge value;

-a source recipe defining a production process at a source scale;

-at least one acceptability function of the conditions defining values of process parameters at the source scale and/or at the target scale; and

-a calculation module configured to:

-simulating the execution of the production process at the source scale using the source authority specification, the source recipe and the parameter evolution information;

-determining one or more source trajectories of process parameters from the simulation, wherein the trajectories correspond to time-based curves of values recordable during the execution of the simulated production process;

-performing a targeting step, the targeting step comprising:

-selecting one of a plurality of recipe templates;

-providing input values for at least one variable parameter in the selected recipe template;

-simulating the execution of the production process at the target scale using the target agency specification, the selected recipe template, the input values of the at least one variable parameter and the parameter evolution information;

-determining one or more target trajectories of process parameters from the simulation;

-comparing the source trajectory with the target trajectory;

-calculating an acceptability score for the selected recipe template based on the comparison and the at least one acceptability function;

-calculating an optimal value for at least one variable parameter in the selected recipe template by optimizing the acceptability score and/or calculating an acceptable range for the at least one variable parameter, wherein values within the acceptable range yield an acceptability score above a specific threshold;

-selecting at least one of the plurality of recipe templates as a target recipe based on the acceptability scores calculated for the one or more recipe templates.

The computer system may particularly comprise a memory and a processor for operating the module. The retrieving module, the receiving module and the calculating module may be separate entities or may at least partly overlap each other.

Further, the computer system may be configured to interface with a target control system and/or a source control system via a network, a shared disk, or a database system, where the control system controls the execution of the production process in the real world at both the source scale and the target scale. The interface may in particular allow the transfer of data and/or commands.

In some examples, the computer system may be consistent with at least one of the target control system and the source control system.

The computer system described above may also be adapted, mutatis mutandis, to implement the recipe queue scaling, the point in time scaling and the point in time queue scaling as described above.

In summary, the present invention specifically provides a scaling way to treat the process as a whole, so that optimization of the match between scales does not favor parts of the process at the expense of the whole. This is the case for both recipe scaling and point in time scaling. Furthermore, a way to assess risk on the potential operating space is provided by the acceptability score, not just the optimal values of the process parameters at the target scale.

The acceptability function in particular allows sensitivity to be expressed, i.e. knowledge about the process parameters that are important and to be taken into account in the scaling process. Furthermore, the parametric evolution information specifically integrates experimental bioreactor data with cell culture models in order to accurately simulate a production process involving several process parameters.

Finally, it is possible to adjust multiple scales simultaneously. In fact, the entire scaling queue can be optimized at once, so that at each stage of deployment onto the hardware during scaling, a good prospect is established that can continue further in case of success.

The subject matter described in this application may be implemented as a method or on an apparatus, possibly in the form of one or more computer program products. The subject matter described in this application may be embodied in a data signal or on a machine-readable medium embodied in one or more information carriers, such as a CD ROM, DVD ROM, semiconductor memory, or hard disk. Such computer program products may cause a data processing apparatus to perform one or more operations described herein.

Additionally, the subject matter described in this application can be implemented as a system that includes a processor and a memory coupled to the processor. The memory may encode one or more programs to cause the processor to perform one or more of the methods described herein. Further, the subject matter described in this application can be implemented using a variety of machines.

Drawings

Details of exemplary embodiments are set forth below with reference to exemplary drawings. Other features will be apparent from the description and drawings, and from the claims. The drawings are to be regarded as illustrative in nature and not as restrictive, since the scope of the invention is defined by the appended claims.

FIG. 1 illustrates a computer system for scaling a production process for producing a chemical, pharmaceutical, or biotech product.

FIG. 2 illustrates a method for recipe scaling for a production process.

FIG. 3 shows a block diagram of inputs and outputs indicating recipe scaling.

FIG. 4 shows portions of an exemplary input for recipe scaling.

Figure 5 shows the acceptability score over time.

FIG. 6 illustrates an exemplary target trajectory.

FIG. 7 illustrates a method for time point scaling of states of a production facility.

Fig. 8 shows a block diagram of inputs and outputs indicating time point scaling.

FIG. 9 shows portions of an exemplary input for point in time scaling.

Detailed Description

Hereinafter, a detailed description of examples will be given with reference to the accompanying drawings. It should be understood that various modifications may be made to the examples. In particular, one or more elements of one example may be combined and used in other examples to form new examples.

FIG. 1 illustrates a computer system 10 for scaling a production process to produce a chemical, pharmaceutical, or biotech product.

Computer system 10 may include a processing unit, a system memory, and a system bus. A system bus couples various system components including the system memory to the processing unit. The processing unit may perform arithmetic, logical, and/or control operations by accessing system memory. The system memory may store information and/or instructions for use in conjunction with the processing unit. The system memory may include volatile and nonvolatile memory such as Random Access Memory (RAM) and Read Only Memory (ROM).

The computer system 10 may also include a hard disk drive for reading from and writing to a hard disk (not shown) and an external unit drive for reading from or writing to a removable unit. The drives and their associated computer-readable media provide nonvolatile storage of computer readable instructions, data structures, program modules and other data for the personal computer 920. The data structure may include relevant data for implementing the methods described above.

A number of program modules may be stored on the hard disk, external disk, ROM, or RAM, including an operating system (not shown), one or more application programs, other program modules (not shown), and program data. The application program may include at least a portion of the functionality as described below.

A user may enter commands and information into the computer system 10 through input devices such as a keyboard and a mouse. A monitor or other type of display device is also connected to the system bus via an interface, such as a video input/output.

The computer system 10 may communicate with other electronic devices. To communicate, computer system 10 may operate in a network environment using connections to one or more electronic devices.

In particular, the computer system may interface and communicate with source control system 20 and target control system 30. The computer system 10 may operate (possibly in conjunction with other devices) to scale the production process.

The source control system 20 may be connected to a bioreactor 40 constituting a source device for performing a production process on a source scale. Similarly, the target system 30 may be connected to a bioreactor 50 constituting a target device for performing a production process on a target scale. Although bioreactor 40 is shown as being smaller than bioreactor 50, this situation may be reversed.

The source control system 20 and the target control system 30 may be located in the same production facility or in different production facilities located in different locations. The source control system 20 and the target control system 30 may be different entities or may be identical, i.e. a single entity (not shown).

The control systems 20, 30 and the computer system 10 may be located in different rooms of the same facility or in different buildings on an enterprise campus. The computer system 10 may be a separate entity from the control systems 20, 30. In other examples (not shown), computer system 10 may be consistent with at least one of source control system 20 and target control system 30.

In some examples, a database 60 may be provided. The database may be connected to a network such that the database is accessible by multiple devices/users. The database may be implemented as a cloud database, i.e., a database running on a cloud computing platform. In other words, the database may be accessed over the internet via a provider that makes the shared data resources and data available to computers and other devices as needed. The database may be implemented using a virtual machine image or a database service. The database may use SQL-based or NoSQL data models.

The database 60 may store any of the following: value sets for process parameters, recipes, recipe templates, parameter assessment information, institutional specifications, acceptability functions. The database 501 may be accessed from the process control device 503 via the internet. Communications between database 60 and computer system 10 may be secured, for example, via Internet Protocol Security (IPSEC) or other security protocols. A Virtual Private Network (VPN) may also be used.

Database 60 may be hosted by a service provider (possibly on a virtual machine) and may be accessed by various users from multiple organizations potentially located in a variety of different geographic locations around the world. Alternatively, database 60 may be hosted locally, for example in computer system 10. Thus, computer system 10 and database 60 may or may not be located in physical proximity. In particular, database 60 may be located at a location that is geographically remote from the computer system (e.g., on another continent).

An example of a production process scaled by the computer system 10 may be a fed-batch process comprising the following phases:

-adding a medium to a bioreactor

Conditions (set temperature, pH)

Addition of inoculum

Allowing growth in the batch phase (control of pH, DO, temperature; sampling at intervals)

Move to the feeding phase when nutrients are depleted

Allowing growth during the feed phase (control of pH, DO, temperature; sampling at intervals; supply of additional nutrients)

-harvesting the product.

In particular, the computer system 10 may perform scaling according to two methods: recipe scaling, where the entire process is switched between scales; and instantaneous or point-in-time scaling, where the setting of a given point in time during the process is switched between scales.

Recipe scaling

FIG. 2 illustrates an exemplary method for recipe scaling for a production process. The method will be described in conjunction with FIG. 3, which shows a block diagram of inputs and outputs indicating recipe scaling.

The production process is defined by a plurality of steps specified by a plurality of process parameters, including recipe parameters and dynamic parameters. In the following, the method will be described for a production process in a bioreactor to be scaled from a source scale to a target scale.

Examples of recipe parameters include, but are not limited to: stirring speed (rpm), fill volume (L), total gassing rate (L/hr), gas percentage (%) O2, gas percentage (%) CO2, and parameters defining gassing, filling, temperature, seeding and induction characteristics, and sampling patterns. The curves may be:

constant rate (e.g. constant rate feed)

Exponential (e.g. feeding increases exponentially, which typically roughly corresponds to the action the organism will perform)

A polynomial (e.g. a 3 rd order polynomial to correspond to a specific organism growth pattern);

and may be parameterized by one or more recipe parameters.

Examples of dynamic parameters include, but are not limited to: tip speed (mps), mixing time(s), k_La(hr^-1) Power input (W), power input per volume (Wm-3), reynolds number, froude number, minimum vortex size (μm), surface gas velocity, cell density, cell metabolic state metric, carbon source availability, nitrogen source availability, secondary nitrogen source availability, inhibitor (toxin) concentration, pH, dissolved oxygen (%), dissolved CO2 (%).

An exemplary scenario for transition between scales may be as follows. In thatA small scale process was set up at 250 scale and the intent was to transfer the process as a single step to 50L to produce large quantities of product to assess downstream processing problems (e.g., cleanup/filtration). In another case, the manufacturing process is built at a given large scale (such as 50L) in an organization and needs to be scaled down, for example to15 or

250 size to perform initial clonal selection; the scaling down process needs to be "representative", otherwise the wrong clone will be selected and production will suffer when scaled back to large scale.

The method starts at S101 and the first step at S103 is to retrieve the parameter evolution information 200 and the recipe template 201. In the current embodiment, by way of example, parameter evolution information is retrieved from a set of hard-coded formulas in software, in conjunction with structured data in XML format stored on a file system accessible to the software.

The parameter evolution information 200 characterizes how a process parameter varies with an event, illustratively including the relationship between the process parameter's initial condition and the process parameter. The parameter evolution information 200 describes the physical modeling of the bioreactor in terms of parameters describing the bioreactor state (e.g., fill volume) or production state (e.g., power per volume) as determined by the bioreactor, as well as the biological modeling of the cell culture in the bioreactor (e.g., in terms of growth, oxygen consumption, and pH reduction).

In particular, the parameter evolution information 200 may include empirically derived relationships from previous executions of the production process as well as equations derived through theoretical models regarding the evolution of the production process.

For example, the parametric evolution information 200 includes an experimental bioreactor data fit, which is an empirically derived mapping between recipe parameters (such as stirring speed, gassing rate, and fill volume) and dynamic parameters (such as mixing time, kLa, and power input). Experimental bioreactor data fitting connects two or more process parameters to each other.

Additionally, the parametric evolution information includes theoretically derived equations and starting points for both the cell culture model and the bioreactor physical model. Examples of starting points for cell culture models may be: growth rate 0.02 hr-1; optimum temperature 36, wherein growth decreases by 80% for each reduction; pH optimum 7.4, wherein growth is reduced by 50% per decreasing unit of 1/10; scaling to a specific 1C consumption rate of units; growth rate saturation function of 1C source. In an alternative embodiment, the cell culture model may be an empirical statistical model.

The bioreactor physical model may encompass fill volume, temperature, analyte concentration, pH, kLa, mixing time, power input, and dissolved oxygen. Details regarding each of these process parameters are provided below:

filling volume

The initial value of the fill volume is zero. The volume accumulates due to the liquid addition and the pill-like liquid addition and decreases due to the sampling. Evaporation is not considered below, but it is possible, for example, to implement a standard evaporation model, whereby it is assumed that the input gas is free of water, and that the exhaust gas is saturated (without a condenser).

For volume v_bThe fill volume is considered to be instantaneously updated such that:

v_new＝v_old+V_b

for volume reduction due to sampling, at a sample volume of v_sThe fill volume is considered to be an instantaneous update, such that:

v_new＝v_old-v_s

during continuous liquid addition at the rate curve r (t), the fill volume is updated according to the following expression:

temperature of

The temperature is precipitated by the following temperatures:

-an external temperature modeling the heating/cooling jacket or the air flow;

-temperature variations due to the supply of liquid.

The continuous variation of the temperature depends on the parameters of the single bioreactor type, which indicate the heat transfer rate between the external temperature and the internal temperature:

wherein T is_intIs the temperature of the liquid, k_TIs a heat transfer coefficient, and T_ext(t) is the external temperature.

The equations for the temperature change due to the supply of liquid (bolus supply or molding supply) are similar to those for the analyte concentration (see below).

Concentration of analyte

The analyte concentration changes due to the liquid addition and the pill-like liquid addition.

For volume v_bFor the concentration c in the pellet_bAny given analyte at which the analyte concentration is considered to be instantaneously renewed such that:

wherein v is_oldIs the volume before the addition of the pill-like liquid.

For continuous liquid addition with rate curve r (t), where the fill volume is expressed as v (t), the concentration in the supply liquid c_aThe concentration of any given analyte c below is updated according to the following expression:

pH

the following is a simplified buffer model intended to give representative results without specifying in detail the detailed buffer properties of the medium. This is achieved by tracking pH and buffering capacity (as two different variables).

Both liquid and pill liquid additions affect pH. The effect of mediated carbon dioxide (or mediated carbonic acid) on pH has not been considered, but there are possible theoretical equations for those in the literature that may be included.

The buffering capacity of the medium is considered to be similar to the analyte and therefore follows the equation for analyte concentration given above.

For a volume of v_bIn a pill-like liquid form (wherein the pH is p)_bAnd a buffering capacity of b_b) Added to a volume of v_mpH of p_mAnd a buffering capacity of b_mThe new medium pH is approximately:

dissolved oxygen

The dissolved oxygen is constantly changing in the bioreactor due to transfer of oxygen from the gas supply, e.g. from k_La is determined such that:

where O is the partial pressure of oxygen in the supply gas relative to oxygen in air.

The biological model of cell culture is intended to enable a brief description of a wide range of biological processes to highlight stirred bioreactor cultures. The following model does not address, for example:

detailed metabolic aspects of the cell: these are only relevant as they influence the behavior of the cells in terms of interaction with the bioreactor or end product;

heterogeneity in the bioreactor: the assumption is that the consideration of heterogeneity is adequately handled by penalizing large mixing times in terms of utility tools;

details and details of any individual process: for example, the goal is to obtain broad but approximate coverage rather than high detail about a particular cell type or product.

The biological model instead focuses on:

bioreactor-related culture effects such as pH drop or rise (which may trigger bioreactor system alkalinization or acidification) and oxygen utilization (which may affect gas flow or agitation speed via the intermediary of the DO control loop);

end product-related culture dynamics, such as modification of cell activity or large-scale changes in cell metabolism, e.g. for production away from growth;

bioreactor-related culture effects such as pH, DO or nutrient concentration effects.

The model is structured in many "culture model processes"; these are additionally combined to produce a system of ordinary differential equations. Each culture model process has many components that govern its overall rate multiplied by a maximum rate that is constant for the process. The components are functions of critical variables, and the outputs of these functions are combined multiplicatively.

For example, the rate may be determined as the temperature T and the primary carbon source concentration c_1CFunction of (c):

r＝f_T(T)f_[c1c](c_1c)

function f_TAnd f_[DO]A biologically highlighted form selected from the (small) repertoire, where the maximum is one unit and the minimum is zero. For example, the temperature dependence can be described by the following formula:

indicating that the optimum temperature is at 37 degrees but that the sensitivity to deviations from this temperature is relatively low. Similarly, the concentration dependence of the primary carbon source can be described by the following formula:

to indicate a saturation maximum, where for 0.5gL^-1The primary carbon source concentration of (a) achieves half the maximum rate.

This rate then determines the rate of change of a set of variables affected by the culture model process. For example, the rate may drive the growth of cell density such that:

a sufficient expression of the cell growth rate will then become:

in the above example, the culture model process has two dependencies (primary carbon ring and temperature dependence) but only produces a single effect (affects cell growth rate). A single culture model process can produce multiple effects, such as affecting cell growth rate and primary carbon source consumption, which immediately results in a system of equations.

The culture model process may depend on one or more driven processes in the hierarchy in terms of its rate. In this case, the rates of the driving processes may be added (e.g., considering the case where the driving processes are related to nutrient dependent growth and production, respectively, and the driven processes are nutrient consuming), or multiplied. The following sections cover details of response and rate.

The culture process responds:

temperature of

The culturing process (such as growth, production or quiescence/death) may depend on the medium temperature. It is assumed that the process will be independent of the external (casing, drive) temperature, except as this regulates the media temperature.

It is envisaged that the temperature dependence will typically include a normal distribution, as described above, or, more generally, an asymmetric normal distribution, i.e.,

for the growth of Cho cells, for example, a normal distribution (where μ -37 and σ -2) would be a good starting point.

pH

The culturing process (such as growth, production or quiescence/death) may depend on the medium pH. It is envisaged that the pH dependence will typically comprise a normal distribution or an asymmetric normal distribution.

For the growth of Cho cells, for example, a normal distribution (where μ ═ 7.4 and σ ═ 0.5) would be a good starting point.

DO

The culturing process (such as growth, production or quiescence/death) may depend on dissolved oxygen saturation within the medium. It is envisaged that DO dependence will reflect saturation or sigmoid kinetics, i.e.,

or

For the growth of Cho cells, for example, saturation kinetics (where k is_crit15% and k_sens5%) will be a good starting point

Response of cellular metabolic status

A single metabolic state variable is used to summarize the relevant properties of the metabolic state of a cell. The meaning of this variable will be culture dependent but is primarily intended to summarize the behavior of the cell in terms of relative energy input for growth and production. For example:

-metabolic state ═ -1, indicating 100% energy put into growth

-metabolic state ═ 1, indicating 100% energy input into production.

The simulation was started with the metabolic state set to zero and the state was maintained over the interval from-1 to 1 (not included).

Obviously, this is a huge simplification of the dynamics in the actual culture. However, modeling the effect of induction and thus the linkage effect is sufficient in terms of further amplification of cell density (or others).

It is envisaged that the response of the culture model process in relation to production and growth takes the form of an S-shape or an inverted S-shape (i.e. 1-S-shape), respectively.

Cellular activity response

The individual cell activity state variables serve to profile relevant properties of the cell activity, in particular quiescence or recovery from retardation, as appropriate for the culture in question.

The simulation was started with cell activity set to zero, and cell activity continued for an interval from-1 to 1 (not included).

In the case of cultures exhibiting a recovery component, it is envisaged that the growth exhibits a sigmoidal response to a state variable of cellular activity, with other processes responsible for increasing cellular activity; in this case, a high cell activity state indicates that the cells have largely recovered from, for example, thawing.

In the case of cultures exhibiting dead or quiescent components, it is envisaged that growth (and possibly also production) will exhibit an inverse sigmoidal response to a state variable of cellular activity, with other processes responsible for increasing cellular activity; in this case, a high cell activity state indicates a large number of quiescent or senescent cells.

Nutrient response

The interaction between the cell and the medium is complex. Regardless of the response to pH that has been covered, the biological response to the medium can be depicted as:

support of growth by the medium, e.g. because sufficient basic nutrients are provided, typically sufficient carbon and nitrogen supply;

support production by media (end product) with the same standards;

inhibition of growth by the medium, for example, due to the presence of toxic medium components;

promotion of cell senescence or death by mediators, for example, because of the presence of toxic mediator components;

-facilitating recovery from lag phase, e.g. because of the presence of a supportive nutrient environment;

-promoting the transition from growth to production.

In many cases, these drivers can be modeled as a saturation power or sigmoid response to the concentration of a particular nutrient, e.g.,

saturation kinetic response of growth to primary carbon source

Growth saturation kinetic response to primary nitrogen source

Sigmoidal dynamic response of metabolic state change to inducer concentration

-reverse sigmoidal or saturation kinetic response to toxic products or toxins.

In some cases, the nutrient response is primarily dependent on the ratio of the two components of the medium. This is especially the case when multiple carbon sources are present, and one carbon source is used in preference to another.

In this case, the cell culture process rate depends on a function of the quotient of the concentrations of the components (such as sigmoid, saturated or anti-sigmoid).

The culture process has the effects:

growth rate

The process of growing or dying cells culture affects the change in cell density ρ:

wherein r is_gIs a growth rate coefficient associated with the cell culture process and its product with the rate of the cell culture process (i.e., R, from the cell process response) indicatesSpecific growth rate. Thus, a constant non-zero R will result in exponential growth of cell density (R)>1) Death (R)<1)。

The overall change in cell density during culture simulation results from growth and death due to the relevant cell culture process, changes in dilution (e.g., due to liquid pill or forming liquid addition), and changes due to supply of inoculum (i.e., because liquid is added with a non-zero inoculum concentration).

Rate of pH inhibition

The cell culture process can inhibit medium pH. The framework provides two ways to model the inhibition of pH:

direct pH suppression by cell culture process

Indirect pH inhibition due to consumption or production of acid or base medium components by the cell culture process.

In cases where cell culture dynamics are considered to be at high levels (e.g., utilizing any carbon source), it may be more appropriate to employ the former approach.

In this case, the pH is adjusted according to the following expression:

wherein r is_[pH]Is the pH inhibition coefficient of the cell culture process, R is the rate of the cell culture process at a given time, and B is the specific buffering capacity of the medium.

Rate of O2 consumption

Growth, maintenance and other metabolic cellular activities consume oxygen from the media. The cell culture process regulating medium DO involved in oxygen uptake was as follows:

wherein r is_[DO]Is the DO uptake coefficient of the cell culture process, and R is the rate of the cell culture process at a given time.

CO₂Rate of generation

Similarly, metabolic activity and in particular oxidative metabolism produces carbon dioxide. Cell culture process regulating medium ppCO involving carbon dioxide evolution₂The following are:

rate of modification of cellular activity

As previously described, within the framework, the state of cell activity describes the movement of cells out of the lag phase, or into quiescence/senescence. Cell activity status is a measure of cell activity between-1 and 1 (not included). To maintain the state within this interval, the cell culture process in the regulatory state does so as follows:

wherein A is the state of cell activity, r_AIs the cell culture process rate coefficient for the state of cell activity, and R is the rate of the cell culture process at a given time. Cell density dependent effects were not assumed, as the activity state is considered to apply equally to all cells in culture.

Rate of modification of cellular metabolic state

The modification of the metabolic state of the cell reflects a modification of the cellular activity, that is to say:

wherein M is the metabolic state of the cell, r_MIs the cell culture process rate coefficient for the metabolic state of the cell, and R is as above.

Rate of nutrient production

The cell culture process may produce or consume components of the medium. For example:

cell growth and maintenance typically consume a carbon source and possibly a different nitrogen source;

cellular metabolism can produce products

Cellular metabolism may be a toxic by-product.

Any given cell culture process may have zero or more nutrient production rate coefficients, each of which indicates for nutrient i in question:

wherein r is_ciIs the rate coefficient.

Illustrative cultivation Process

The following illustrative culture process demonstrates many aspects of the framework described above.

Further, the recipe template 201 is retrieved at step S103. The recipe can be considered as an instruction set that specifies how the bioreactor behaves over time. The recipe template is considered a recipe with free or variable parameters. These variable parameters may result in the generation of very different recipes from a given template (e.g., if the path within the template depends on free variables). The recipe templates may contain calculations based on variable parameters as well as on other process parameters within the process they run. Any of the recipe parameters listed above may be variable parameters. For example, for what is depicted as A + B t + C t²The curve parameters A, B and C of the feed rate of (a), where t is time, can be varied freely.

A library of recipe templates 201 may be retrieved, where different recipe templates may include different steps or instructions and/or may have different variable parameters.

The recipe template 201 may include indicia identifying the stage of the production process that is used with the acceptability function, as explained below.

At step S105, an acceptability function 250 is received. In an exemplary implementation, the software supplies a user interface whereby the acceptability function may be selected from a library and then parameterized or designed graphically. In this case, the library provides an acceptability function that indicates, among other things, (a) a range, (b) a single point, (c) a normal distribution. For other examples, see the canonical form below. The acceptability function defines the conditions for the values of the process parameters at the source scale and/or the target scale, in particular they define how acceptable these values are when viewed individually or relative to each other. The value of acceptability may be a real number between 0 and 1 (including the boundary).

Absolute and relative acceptability functions may exist.

The absolute acceptability function maps from one or more process parameters at the same scale to an estimate. An example of an absolute acceptability function defines the conditions for the following parameters:

-reynolds number (Rn): 0 for low Rn, increasing to 1 as Rn moves into the swing region, then both 1;

-k_La: for low k_La is 0, with k_La increases to 1 as a function of saturation;

-mixing time: 1 for low mixing times and then equal to 20 s/mixing time (i.e. decreasing towards 0) when the mixing time exceeds 20 s;

surface Gas Velocity (SGV): 1 for low SGV, with S-shaped drop as SGV increases, and 0 for larger SGV to reflect increased risk of blistering as SGV increases

Power per volume: a normal distribution around a certain maximum;

-stirring speed: 0 for 0 … … 5% and 95% … … 100% of the bioreactor agitation speed, and 1 otherwise (preferably run without the system at its limits); or linearly increasing from 0 at 0% to 1 at 5%, then flattening to 95%, then linearly decreasing to 0 at 100%;

-vortex size: 1 for vortex sizes greater than 5x organism size, then linearly becomes 0 for 2x organism size to reflect increased risk to the organism as vortex size decreases;

product concentration at harvest: 0 for 0, where saturation increases as the titer increases (tends to 1 as the product concentration tends to infinity);

-DO: 0 for 0 … … 10%, then sigmoidal up to 1 between 10% and 2, 0%, and then 1 for > 20% to ensure proper oxygen for the organism in the sensitive area;

-SGV + protein concentration: 1 for low SGV or low protein concentration, decreasing to 0 as either becomes larger; similarly, the stirring speed + SGV may also affect the risk of foaming:

power per volume + cell density: a normal distribution around an optimum, but as cell density increases, the optimum goes from low power per volume to high power per volume (reflecting the protective effect of the cell on other cells);

product concentration + product mass: if either is 0, it is 0 and then increases in proportion to the product of the concentration and the mass.

The relative acceptability function maps to an evaluation from a combination of the process parameter(s) at the source scale and the corresponding process parameter(s) at the target scale. For example, a way to combine two corresponding process parameters at different scales is to calculate their difference or relative difference, i.e. the absolute value of (value in source-value in target)/(maximum (value in source, value in target)). An example of a relative acceptability function defines the following condition:

-0 if the mixing time at the source scale is less than the target scale, and 1 otherwise (ensuring no loss of mixing on scale up);

normal distribution of PPV around increments of 0 (ensuring PPV, typical parameters for matching are maintained between scales);

for k at target scale greater than source scale_La is 1, otherwise with k_La gradually varies and decreases in sigmoid form to 0;

normal distribution of n around 0, standard deviation 0.2 for cell density, (ensure that the growth curve is maintained before source and target).

Generally, the acceptability function may be one-dimensional, two-dimensional or have a higher dimension. Some examples of canonical forms of one-dimensional and two-dimensional acceptability functions are reported below.

The one-dimensional acceptability function may take one of the following canonical forms:

always zero except at a given exact value, at which is a one (which expresses the need to spatially constrain the solution to an exact value, e.g. if a specific fill volume is required at small scale);

normal distribution (this expresses e.g. k)_LThe idea that the parameter value of a has the best value, but there is still room for convolution on this problem); one within a certain range and zero outside this range (this expresses the idea that the parameter value should remain within this range, e.g. the mixing time should be less than a given maximum value)

The two-dimensional acceptability function may take one of the following canonical forms:

-f (x, y) is 0 unless x is y, in which case f (x, y) is 1 (this expresses the need to find an exact match between scales, e.g. for power input per volume);

-f (x, y) ═ 0 unless x ≧ y, in which case f (x, y) ═ 1 (which expresses the need to spatially constrain the solution to the case where the parameter value exceeds that at the source scale at the target scale, e.g., in terms of oxygen transfer)

-f (x, y) ═ 0 unless x ≦ y, in which case f (x, y) ≦ 1 (which expresses the need to constrain the solution space to the case where the parameter values are smaller at the target scale than at the source scale, e.g. in terms of mixing time)

-f (x, y) ═ N (x-y; m, s) (this expresses the benefit of being as close as possible, but not necessarily a perfect match between scales, for example for power per volume)

-f (x, y) ═ N ((x-y)/max (x, y); m, s) (supra, but in a ratio meter).

In the case of recipe scaling, the acceptability function may be appended to a specified portion of the process template. For example, the process template may specify "start of batch" and "end of batch," and the acceptability function will then be appended to the interval between these two waypoints and considered applicable only to those parts of the process. For example, in the batch phase only those acceptability functions that do not set the acceptability value to be dependent on the cell density may be applied, since in the batch phase there may be some variation as the cells grow to run out of their nutrients. In contrast, in the feed phase only those acceptability functions whose acceptability value depends on the cell density can be applied, since at the beginning of the feed phase the changes due to the initial inoculum should "level" since they all have the same amount of nutrients in the batch phase. In another example, only those acceptability functions whose acceptability value depends on the product titer (concentration) can be applied during the harvest stage, since before the harvest point the titer is irrelevant.

At step S105, source agency specifications 220 and target agency specifications 230 are also received. In an exemplary implementation, the user selects the source and target institutions from a list. The software retrieves information about a given source or target facility, including permitted configurations, minimum stirring speeds, maximum stirring speeds, mixing properties, etc., from structured data in an XML file stored on a file system in a location accessible to the software.

The institutional specification is a description of the scale, i.e., the plant at the source scale and target scale, specifying, for example, the volume and number/type of plant components. The source mechanism specification 220 may be, for example, with a mammalian impeller250 and the target agency specification 230 may be, for example, any of 2L UniVessel, 50LSTR with 3+6 impellers and a combined distributor, 2000L STR with 3+6 impellers and a combined distributor.

Further, at step S105, a recipe 240 for the source scale is also received. In an exemplary implementation, the software provides a user interface whereby a user can design a recipe or recipe template by adding and removing steps in succession and by parameterizing the steps. Net serializers are used by the software to hold recipe templates in XML-based repositories to hold object models of recipes, represented internally as objects, within the software implementation.

An example of a source recipe 240 may correspond to the following process: "fill the bioreactor with 0.2L of a given medium; heating to 35 ℃; seeding with clones to a density of 1e6 cells mL-1; incubating and stirring at 600rpm for 36 hours; control pH to 7.4 with bottom and top control, i.e. add acid or base as needed to push pH back to 7.4; maintaining the temperature; gassing with air at a rate of 0.1 total volume per minute; feeding for 36 hours by using composite feed; continuously monitoring and controlling the pH and the temperature; DO is controlled by stirring and gassing; an inducer is added to trigger production. And harvesting after 36 hours. "

The combination of the source authority specification 220, source recipe 240, and parameter evolution information 200 is then used to simulate the execution of the production process at the source scale at step S107. The source authority specification 220 provides a framework for simulation, while the source recipe 240 and the parameter evolution information 200 define how the process evolves.

Physical, chemical and/or biological aspects of the production process are simulated. In particular, process simulations include purely physical modeling (e.g., of the fill volume), bioreactor modeling derived from physical characterization of the bioreactor (e.g., mapping to power per volume from fill volume and agitation speed), and biological modeling of organisms (in terms of growth, oxygen consumption, and pH inhibition).

From the simulation at source scale, the source trajectory of the process parameters is determined at S109. This means that the values of the process parameters are recorded at different times during the simulation, so that the time dependence of the process parameters can be determined. The data points may be fitted to obtain a fitted function of the time evolution of the parameters. Fig. 6 shows an example of a trajectory, which will be discussed below.

Thereafter, at steps S111 and S113, a tentative target recipe is selected. A recipe template is selected among the plurality of recipe templates 210 and some input values are provided for variable parameters in the selected recipe template. The combination of the selected recipe template and the provided input values provides a tentative target recipe that can be used at step S115 to simulate a production process at a target scale, similar to step S107.

FIG. 4 illustrates portions of exemplary inputs for recipe scaling, where the parameter evolution information 200, the source recipe 240, the input values of the selected recipe template 210, and the acceptability function 250 are visible, without showing the source and target agency specifications.

Further, at step S117, a target trajectory 270 corresponding to the time evolution of the process parameters at the target scale is determined, just as for the source trajectory.

The initial guess of the variable parameters provided by the input values is then modified at step S119 to "best" satisfy the conditions given by the acceptability function 250. In particular, the simulation may be run multiple times to explore the space available for the variable parameters until a preferred point or surface in the space is found, i.e., the point or surface that best fits the target trajectory 270 to the acceptability function 250. Compliance of the target trajectory 270 with the acceptability function 250 is indicated by an acceptability score 280. Different degrees of compliance may be of interest, such as considering only the value of the variable parameter that maximizes the acceptability score 280, or also considering multiple values that produce an acceptability score 280 above some threshold. The multiple values may form a single (possibly multi-dimensional) range or a non-adjacent range.

At step S121, it is checked whether there are other applicable recipe templates that can be used as a basis for the target recipe, and if so, steps S111 to S115 are repeated.

Finally, at step S123, one or more target recipes 260, i.e., recipe templates having corresponding values of the variable parameters, are selected among the tested tentative target recipes. The selection is based on the acceptability score 280 and may be such that only the tentative target formula with the highest acceptability score 280 is considered, or more tentative target formulas with acceptability scores 280 above a certain threshold are considered.

The method ends at S125.

In some cases, the target recipe 260 may be output, for example, as a file. Examples of outputs indicative of the target recipe 260 may be: "on ambr 250, a formulation template named" increase agitation speed "was used, where the initial inoculum was set to 0.2% of the total volume and the DO was controlled to 35% all the time". An acceptability score 280 may also be output, for example, as the text "this would give an 80% score of the best transition from your 50L recipe" or as a graph as a function of time, as shown in fig. 5. Further, the simulated target trajectory 270 may be output, for example, as a graph.

Fig. 5 shows acceptability scores over time, and fig. 6 shows an exemplary target trajectory.

As can be seen, the acceptability score 280 in fig. 5 has two valleys around hours 4 and 6. An advantage of outputting the target trajectory 270 is that the cause of poor performance in these stages can be easily identified. For example, considering the target trajectories in the work on cell density part of fig. 6, it is clear that cell density evolves similarly at the source scale and at the target scale. Since one of the goals is to maximize the similarity between processes at different scales, the cell density will score higher. In contrast, FIG. 6 for k_LTarget trajectories in the lower half of a show, k_LThe evolution of a at the target scale is dissimilar to the evolution at the source scale. Thus, k_La may be at least partially the cause of a trough in the overall acceptability score 280.

In general, the method involves mapping a recipe at one scale to a recipe at a second scale subject to some constraints by evaluating the trajectories caused by simulating the recipe at each scale in question. In particular, matching the best trajectory according to the acceptability function is used to infer the value of properly filling the recipe template.

The method can be generalized to queue scaling involving any number of scales, i.e., a transition from a source scale and to a final target scale by traversing multiple intermediate target scales and transitioning for each intermediate target scale.

When the recipe scaling method of fig. 2 is applied, the agency specifications for the intermediate target scale are received at S105, and the acceptability function 250 covers all scales, i.e., conditions that define all scales. In addition, steps S113, S115 and S117 must also be performed for each intermediate target scale. Further, step S119 involves a simultaneous search for the "best" values (optimal values or ranges) for all target scales (i.e., one or more intermediate target scales and the final target scale).

Exemplary scenarios for queue scaling may be selected from

15 to2L to50L to1000L to

2000L. In particular, each of these scales may belong to a scale group, as follows:

-configuration 1:250, attached to small scale groups

-configuration 2:

2L, attached to small-scale and intermediate gauge modules

-configuration 3:

50, attached to the intermediate and large scale groups

-configuration 4:

1000, attached to Large Scale groups

-configuration 5:

2000, attached to Large-Scale groups

The following acceptability functions 250 are defined for each group:

1) small scale group:

-relative acceptability function: k is a radical of_La normal function of the increment of a, with a mean of 0 and a standard deviation of 1hr^-1；

-relative acceptability function: normal function of the increment of PPV, with mean 0 and standard deviation of 0.2 Wm^-3；

-absolute acceptability function: when between 0 and 5% and 95% and 100%, the value is 0, otherwise 1 (% of the stirring speed of the bioreactor is taken as maximum);

-absolute acceptability function: a sigmoid function, where the value for 0 is 0 for dissolved oxygen, spikes to 1 around 20.

2) The intermediate gauge module:

relative acceptability function: normal function of the increment of PPV, with mean 0 and standard deviation of 0.1 Wm^-3；

3) Large-scale group:

-relative acceptability function: normal function of the increment of PPV, with mean 0 and standard deviation of 0.1 Wm^-3

Absolute acceptability function: 1 kW/(1 kW + power input)

It can be seen that in queue scaling, in particular, the acceptability function 250 relates different scales such that the process is scaled for each scale, taking into account also the requirements that will arise for the next scale.

Queue scaling can be computationally expensive, however, when exploring the space of variable parameters, an acceptability function can be used.

For example, it is conceivable to have 3Simple cases of target sizes A, B and C. A and B may be represented by_La differs by more than 1hr^-1An acceptability function that falls below 0.5 is linked, while an acceptability function that links B and C is such that if k between B and C is_La is completely different, then is reduced to 0, otherwise is 1. The variable parameter is the stirring speed, so the stirring speed for scale a, the stirring speed for scale B, and the stirring speed for scale C require input values. A threshold of 0.5 is set for the final acceptability score and the acceptability score functions are combined by taking into account the product of the acceptability scores.

After the input value of the stirring speed at the scale a is selected and the input value of the stirring speed at the scale B is selected, k may be calculated based on the stirring speed values_La, and in particular k between A and B_La is poor. Will only exist k_La has a difference of not more than 1hr^-1And thus certain regions with an acceptability score greater than 0.5. Since, by definition, the acceptability scores are all lower than or equal to 1, and since the acceptability score functions are combined as a product, it can be easily concluded that any value of B that is not in the above region will result in a final acceptability score that is lower than the threshold. Thus, an input value based on a gives a range of possible input values for B.

Conversely, after proposing a candidate stirring speed for scale B, this may propagate immediately to scale C, since only what is acceptable given the same k at scale C_La stirring speed (using the same logic as before).

The conclusion is that optimization should be performed primarily on the value of a, and the range of B values used for optimization is reduced according to the candidate stirring speed of a, and no optimization on C is needed, as it is automatically derived from B. This gives the optimizer a set of appropriate clues to enable it to get a solution quickly, rather than exploring all possible values of stirring speed at A, B and C.

Time point scaling

FIG. 7 illustrates an exemplary method for time point scaling of the state of a production facility. The method will be described in connection with fig. 8, which shows a block diagram of inputs and outputs indicating time point scaling.

In the following, the method will be described for a production plant comprising a bioreactor which can be used to perform any of the production processes discussed with reference to recipe scaling.

Examples of bioreactor configurations include, but are not limited to:

15 fermenting,15 cell culture,

250 of mammals,

250 microorganisms, Univessel

2L, with annular distributor and 2x3 bladed impeller50. With mini-distributors and 3+ 6-bladed impellers200, and with annular distributor and 2x3 bladed impeller1000。

The state of the production equipment is defined by state parameters that may include, but are not limited to: stirring speed (rpm), fill volume (L), Total outgassing Rate (Lhr)^-1) Gas percentage (%) O2 and gas percentage (%) CO 2.

The method starts at S701 and the first step at S703 is to retrieve mapping information 800. In an exemplary implementation, the mapping information is stored with the bioreactor configuration in an XML file accessible to software.

The mapping information 800 characterizes how the state parameters relate to derived parameters characterizing a given point in time during the production process. Derived parameters include, but are not limited to: tip speed (mps), k_La(hr^-1) Mixing time(s), power input (W), power input per volume (W/m)³) Reynolds number, Froude number, minimum vortex size (. mu.m) and superficial gas velocity.

In particular, the mapping information 800 may include experimental bioreactor data fits derived from previous executions of the production process and/or equations derived through theoretical models.

The mapping information 800 may include different relationships between the status parameters and the derived parameters that are applicable to different production equipment. Therefore, in the search step S703, only the relationship suitable for the case at hand can be searched.

For example, can be at250 bioreactor and UnivesselTime point scaling was applied between 2L bioreactors. The retrieved mapping information 800 may include k from stirring speed, fill volume, and outgassing rate_La mapping of PPV and an equation relating the agitation speed and tip speed of a bioreactor of a given geometry.

At step S705, source agency specifications 810, target agency specifications 820 and an acceptability function 840 are received. In an exemplary implementation, the source and target institutions are specified by a user selecting from the combo box through a user interface. The details of the source and target organizations, in terms of associated parameters and mappings, are stored in XML files on a software accessible file system.

The mechanical specification is a description of the scale, that is, the description of the plant at the source scale and the target scale, specifying the body such as the plant partProduct and quantity/type. Continuing with the example above, the source authority specification 810 may be250, and the target agency specification 840 may be2L。

The acceptability function 840 may have any of the canonical forms previously described. For the example above, three acceptability functions 840 may be received: relative acceptability function, which is k for source scale and target scale_LThe difference between a is 0hr on average^-1And standard deviation of 1hr^-1Normal distribution of (2); a relative acceptance function, which is a normal distribution with a mean of 0s and a standard deviation of 5s for the difference between mixing times at the source scale and the target scale; and an absolute acceptability function that requires a tip speed at the target scale of 5% of the maximum in the bioreactor (i.e., 0 for tip speeds below 5% and 1 for tip speeds above 5%).

At step S705, a first set of state parameters at the source scale and a second set of state parameters at the target scale 830 are also received. The first set of state parameters of the above example may be: stirring speed: 400 rpm; gas outlet rate: 0.02L min^-1(ii) a Filling volume: 0.2L; gas: 100% air. The second set of state parameters may be: filling volume: 2L; gas outlet rate: 0.2Lmin^-1(ii) a Gas: 100% air, where stirring speed is a variable parameter, which will be filled afterwards.

Fig. 9 shows a portion of an exemplary input for time point scaling, where the source and target agency specifications 810 and 820, the first and second sets of status parameters 830, and the acceptability function 840 are visible, and the mapping information 800 is not shown.

Then, at step S707, a first set of derived parameters for the source size is calculated. In the given example, considering that the stirring speed was 400rpm, the gas outlet rate was 0.02L min^-1Fill volume 0.2L and gas 100% air and use the retrieved mapping information 800 to calculate k at ambr 250 scale_La and mixing time.

Then, at step S709, an input value of the variable parameter is selected. For example, the input value for the stirring speed at the target scale may be 40rpm, which is the midpoint of the minimum and maximum stirring speeds of the UniVessel 2L. Using the input values, at step S711, a second set of derived parameters 860 for the target scale is calculated, similar to step S707. Thus, in the given example, considering a stirring speed of 40rpm, an outgassing rate of 0.2L min^-1A fill volume of 2L and a gas of 100% air and calculated using the retrieved mapping information 800K at 2L Scale_La and mixing time. In addition, according to the mapping information 800

Tip speed was calculated from the 2L scale geometry and candidate agitation speed 40 rpm.

The initial guess of the variable parameters provided by the input values is then modified at step S713 to "best" satisfy the conditions given by the acceptability function 840. In particular, the space available for variable parameters is explored until a preferred point or surface in the space is found, i.e., the point or surface that best fits the second set of state parameters to the acceptability function 840.

Compliance of the state parameters at the target scale with the acceptability function 840 is indicated by an acceptability score 870. Different degrees of compliance may be of interest, such as considering only the value of the variable parameter that maximizes the acceptability score 870, or also considering multiple values that produce an acceptability score 870 above some threshold. The multiple values may form a single (possibly multi-dimensional) range or a non-adjacent range. If each acceptability function 840 gives a partial acceptability score, the overall acceptability score may be given by the product or average or other combination of all partial acceptability scores.

The result is an optimized second set of state parameters 850 for the target size.

In the given example, k is derived again each time a new input value from the space of stirring speeds is selected_La. The time and tip speed are mixed and a corresponding acceptability score is calculated. The end result is therefore an optimized stirring speed or an acceptable range of stirring speeds on the UniVessel 2L scale.

The method ends at S715.

The method can be generalized to queue scaling involving any number of scales, i.e., a transition from a source scale and to a final target scale by traversing multiple intermediate target scales and transitioning for each intermediate target scale.

When the time point scaling method of fig. 7 is applied, the agency specifications for the intermediate target scale are received at S705, and the acceptability function 840 covers all scales, i.e., conditions defining all scales. In addition, steps S709 and S711 must also be performed for each intermediate target scale. Further, step S713 involves a simultaneous search for the "best" values (optimal values or ranges) for all target scales (i.e., one or more intermediate target scales and the final target scale).