阅读说明：本技术 颗粒分析仪的表征和分选 (Characterization and sorting of particle analyzers ) 是由 乔纳森·林 约瑟夫·特罗特 基根·奥斯利 于 2019-08-30 设计创作，主要内容包括：非参数变换(例如t-分布随机近邻嵌入(tSNE))可用于分析多参数数据,例如源自流式细胞术或其他颗粒分析系统及方法的数据。本发明可以包括这些变换,以用于降维和亚群识别(例如设门)。从本质上讲,非参数变换在未基于所述整个数据集(包括新观测值)训练新变换的情况下无法变换所述新观测值。所述功能利用神经网络使非参数变换参数化,从而允许采用非参数技术变换小型训练数据集。所述训练数据集之后可用于生成准确的参数模型,以便采用与所述初始事件一致的方式评估其他事件。(Non-parametric transformations, such as t-distributed random neighbor embedding (tSNE), may be used to analyze multi-parameter data, such as data derived from flow cytometry or other particle analysis systems and methods. The present invention may include these transformations for both descent and subpopulation identification (e.g., gating). Essentially, a non-parametric transformation cannot transform a new observation without training the new transformation based on the entire dataset (including the new observation). The function parameterizes the non-parametric transformation with a neural network, allowing a small training data set to be transformed using non-parametric techniques. The training data set can then be used to generate an accurate parametric model to evaluate other events in a manner consistent with the initial event.)

1. A computer-implemented method, comprising:

is controlled by one or more processing devices,

receiving from a particle analyzer measurements of a first portion of particles associated with an experiment;

converting the measured value into an initial transformed measured value by using nonparametric transformation;

generating a parametric model to receive as input the measured values of the first portion of particles relevant to the experiment and to generate as output the initial transformed measured values; and

configuring the particle analyzer to:

transforming measurements of particles included in a second portion of the particles based at least in part on the parametric model; and

classifying the particle based at least in part on the measurement and a sorting criterion.

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

receiving gate information identifying a range of measurements in a nonparametric space for classifying the particle, wherein the sorting criterion includes the gate information.

3. The computer-implemented method of any of claims 1-2, further comprising:

transmitting the parametric model to the particle analyzer; and

causing the particle analyzer to configure a sorting circuit based at least in part on the parametric model and the sorting criterion.

4. The computer-implemented method of any of claims 1-3, further comprising:

selecting a measurement value of the particle and an initial post-transform measurement value; and

generating a composite measurement value pair based at least in part on:

(i) a measurement of the value of the particle,

(ii) an initial post-transform measurement of the particle; and

(iii) a noise value, wherein the synthetic measurement value pair comprises a synthetic measurement value and a synthetic transformed measurement value; and

wherein the parametric model further receives the synthetic measurements as input and generates the synthetic transformed measurements as output.

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

receiving gate information identifying a range of measurements for classifying the particles, an

Wherein the particle is selected based at least in part on the measurement value corresponding to the measurement value range.

6. The computer-implemented method of any of claims 1-5, further comprising:

receiving a hardware identifier from the particle analyzer, the hardware identifier indicating sorting circuitry implemented in the particle analyzer;

determining an architecture of the parametric model based at least in part on the hardware identifier, wherein the architecture of the parametric model is applicable to the sorting circuit; and

wherein generating the parametric model comprises generating the parametric model consistent with the architecture.

7. The computer-implemented method of claim 6, wherein the parametric model includes a neural network, and wherein the architecture identifies a number of layers of the neural network and a number of nodes per layer.

8. The computer-implemented method of any of claims 1-7, wherein the parametric model comprises a neural network, and wherein generating the parametric model comprises:

assigning an initial weight to a connection between a first node in a first layer of the parametric model and a second node in a second layer of the parametric model of a second layer;

generating an output suitable for the particle measurement;

comparing the output to the transformed measurements of the particle; and

adjusting the initial weights based at least in part on a difference between the output and the post-transform measurement values,

wherein a subsequent difference value of the particle is less than the difference value after adjusting the initial weight.

9. A system, comprising:

one or more processing devices; and

a computer-readable storage medium containing instructions that, when executed by the one or more processing devices, cause the system to:

receiving from a particle analyzer measurements of a first portion of particles associated with an experiment;

converting the measured value into an initial transformed measured value by using nonparametric transformation;

generating a parametric model to receive as input raw measurements of the first fraction of particles relevant to the experiment and to generate as output the initial transformed measurements;

configuring the particle analyzer to:

transforming measurements of particles included in a second portion of the particles based at least in part on the parametric model;

generating a control signal to adjust an operational state of an analysis device included in the particle analyzer; and

transmitting the control signal to the analysis device to achieve the operational state.

10. The system of claim 9, wherein the analysis device comprises sorting electronics communicatively coupled with a deflection plate, and

wherein the particle analyzer is further configured to identify a target container applicable to a particle based at least in part on the post-conversion measurement value corresponding to a sorting criterion associated with the target container, and

wherein adjusting the operating condition comprises applying an electrical charge through the deflection plate to direct the particles into a target vessel.

11. The system of claim 9, wherein the analysis device comprises a fluidic system, and

wherein adjusting the operating condition comprises adjusting a pressure applied during the experiment.

12. A method, comprising:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

13. A system comprising a processor comprising a memory operably coupled to the processor, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

14. A system, comprising:

a light source configured to irradiate a sample (containing particles) in a fluid medium;

a light detection system comprising a photodetector configured to generate a data signal using light detected from particles within the fluid medium;

a processor comprising a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to:

applying a non-parametric transform to a first set of data signals from the light detection system to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second set of data signals from the light detection system; and

a cell sorter.

15. An integrated circuit programmed to:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

Technical Field

The present invention relates generally to the field of automated particle evaluation technology, and more particularly to sample analysis and particle characterization methods.

Background

Particle analyzers (e.g., flow cytometers and scanning cytometers) are analytical tools that can characterize particles based on electro-optical measurements such as light scattering and fluorescence. In flow cytometry, for example, particles (e.g., molecules, microbeads bound to an analyte, or individual cells) in a fluid suspension are passed through a detection zone where they are exposed to excitation light, typically from one or more lasers, and the light scattering and fluorescence characteristics of the particles are determined. The particles or components thereof are typically labeled with a fluorescent dye for detection. The various particles or components can be detected simultaneously by labeling with fluorescent dyes having different spectral characteristics. In some embodiments, the analyzer comprises a plurality of photodetectors, one for each scattering parameter to be determined, and one or more for each different dye to be detected. For example, some embodiments include a spectral structure in which more than one sensor or detector is used for each dye. The obtained data includes the signals measured for each of the light scatter detector and the fluorescence emission.

The particle analyzer may further comprise means for recording the measured data and analyzing said data. For example, a computer coupled to the detection electronics may be used for data storage and analysis. For example, the data may be stored in the form of a list, where each row corresponds to a particle of data and the columns correspond to each feature being measured. Storing data from the particle analyzer using a standard file format (e.g., an "FCS" file format) facilitates analysis of the data using a separate program and/or machine. With current analysis methods, the data is typically presented in the form of a one-dimensional histogram or a two-dimensional (2D) graph for visualization, but other methods may be used to visualize the multidimensional data.

For example, parameters determined using flow cytometry typically include light scattered primarily in the forward direction at the excitation wavelength by the particle at a narrow angle (known as Forward Scatter (FSC)); the excitation light scattered by the particle in a direction orthogonal to the excitation laser (referred to as Side Scatter (SSC)); and light emitted by fluorescent molecules in one or more detectors (for measuring signals over a range of spectral wavelengths), or light emitted by fluorescent dyes detected primarily in the particular detector or detector array. Different cell types can be identified by light scattering properties and fluorescence emissions obtained/generated by labeling various cellular proteins or other components with fluorochrome-labeled antibodies or other fluorescent probes.

Flow cytometers and scanning cytometers are both commercially available from, for example, BD Biosciences (san Jose, Calif.). For a description of flow cytometry, see, e.g., Landy et al (eds.), clinical flow cytometry, New York academy of sciences Ann (1993), vol 677; bauer et al (eds), clinical flow cytometry: principles and applications, Williams & Wilkins (1993); ormerod (eds), flow cytometry: practical method, oxford university press (1994); jarosszeski et al (eds.), flow cytometry protocols, Methods in Molecular Biology, 91 st, Humana Press (1997); and Shapiro, practical flow cytometry, 4 th edition, Wiley-loss (2003); the entire contents of said publication are incorporated herein by reference. For a description of fluorescence imaging microscopy, see, e.g., the Pawley (eds.), the manual for confocal microscopy of biology, 2 nd edition, Plenum Press (1989), which is incorporated herein by reference.

The data obtained from the analysis of cells (or other particles) using multi-color flow cytometry is multi-dimensional, in which each cell corresponds to a point in a multi-dimensional space defined by the measured parameters. A population of cells or particles is identified as a cluster of points in the data space. Clusters and populations can be identified manually by drawing gates around a certain population displayed in one or more two-dimensional maps (called "scatter maps" or "dot-matrix maps") of the data. Alternatively, clusters may be automatically identified and gates defining the population boundaries may be automatically determined. Exemplary methods for automatic door setting are found in the following publications, for example, U.S. patents numbered below: 4,845,653, respectively; 5,627,040, respectively; 5,739,000, respectively; 5,795,727, respectively; 5,962,238, respectively; 6,014,904, respectively; 6,944,338, respectively; and U.S. patent publication No. 2012/0245889, each of which is incorporated herein by reference.

Flow cytometry is an effective method for analyzing and isolating biological particles (e.g., cells and constituent molecules), and thus it is widely used in diagnosis and therapy. The method uses a fluid medium to linearly separate particles so that the particles can be aligned for passage through a detection device. Individual cells may be distinguished by their location in the fluid medium and the presence or absence of a detectable label. Thus, the flow cytometer may be used to characterize and generate diagnostic profiles for populations of biological particles.

Separation of biological particles has been achieved by providing flow cytometers with sorting or collection functions. Particles in the separated stream that are detected to have one or more desired characteristics are separated individually from the sample stream by mechanical or electrical separation. This flow sorting method has been used to sort different types of cells, to separate X and Y chromosome-containing sperm for animal breeding, to sort chromosomes for genetic analysis, and to separate specific organisms from a complex biological population.

Gates are provided to help understand and classify the large amount of data that may be generated from a sample. Given the large amount of data presented by a given sample, there is a need to effectively control the graphical display of that data.

Fluorescence activated particle sorting or cell sorting is a specialized flow cytometry. It provides a method of sorting a heterogeneous mixture of particles into one or more containers, said method being carried out on the basis of the specific light scattering and fluorescence properties of the individual cells, one cell at a time. It records the fluorescent signal from individual cells and physically isolates the particular cells of interest. The acronym FACS, which is a trademark of Becton Dickinson and is owned by Becton Dickinson, may be used to refer to an apparatus that performs fluorescence-activated particle sorting or cell sorting.

The particle suspension is placed near the center of a narrow, fast flowing stream. The flow is arranged such that on average, when particles arrive randomly (poisson process) at the detection zone, there is a large separation between the particles relative to their diameter. The vibration mechanism stabilizes the outgoing fluid medium into a single droplet containing particles previously characterized in the detection zone. The system is typically adjusted to have a low probability of more than one particle being present in the droplet. If a particle is classified as to be collected, a charge is applied to the flow cell and effluent stream over a period of time to form one or more droplets and detach from the stream. These charged droplets are then passed through an electrostatic deflection system that transfers the droplets to a target vessel in accordance with the charge applied to the droplets.

A sample may include (if not millions of) thousands of cells. The cells may be sorted to purify the sample into cells of interest. The sorting process can generally identify three types of cells: cells of interest, non-cells of interest, and cells that are not recognized. In order to sort out cells having a high purity (e.g., a high concentration of cells of interest), if a desired cell is too close to another undesired cell, the cell sorter that generates the droplets typically electronically halts the sorting, thereby reducing contamination of the sorted population by inadvertent inclusion of undesired particles within the droplets containing the particles of interest.

Disclosure of Invention

In an innovative aspect, a computer-implemented method performed under control of one or more processing devices is provided. The method includes receiving from a particle analyzer measurements of a first portion of particles associated with an experiment. The method includes converting the measurements to initial transformed measurements using a non-parametric transform. The method includes generating a parametric model to receive as input the measured values of the first portion of particles associated with the experiment and to generate as output the initial transformed measured values. The method also includes configuring the particle analyzer. The particle analyzer is configured to convert measurements of particles included in the second portion of the particles based at least in part on the parametric model and classify the particles based at least in part on the measurements and sorting criteria.

Some embodiments of the method include receiving gate information identifying a range of measurements in nonparametric space for classifying the particle, wherein the sorting criterion includes the gate information.

Some embodiments of the method include transmitting the parametric model to the particle analyzer, and causing the particle analyzer to configure a sorting circuit based at least in part on the parametric model and the sorting criterion. For example, the sorting circuit may be implemented in the form of a field programmable gate array.

The measurements received from the particle analyzer may include measurements of light emitted by the first portion of particles in fluorescent form. The light emitted in fluorescent form by the first portion of particles may comprise light emitted in fluorescent form by antibodies bound to the first portion of particles.

Some embodiments of the method may include selecting a measurement of the grain and an initial post-transform measurement, and generating a composite measurement value pair based at least in part on: (i) a measurement of the particle, (ii) an initial post-transform measurement of the particle; and (iii) a noise value, wherein the synthetic measurement value pair comprises a synthetic measurement value and a synthetic transformed measurement value. The parametric model may further receive the synthetic measurements as inputs and generate the synthetic transformed measurements as outputs. In some embodiments, the method may further include receiving gate information identifying a range of measurements used to classify the particle. For example, the particle may be selected based at least in part on the measurement corresponding to the measurement range.

Some embodiments of the method may include receiving a hardware identifier from the particle analyzer, the hardware identifier indicating sorting circuitry implemented in the particle analyzer, and including an architecture to determine the parametric model based at least in part on the hardware identifier, wherein the architecture of the parametric model is applicable to the sorting circuitry. In such embodiments, generating the parametric model may include generating the parametric model consistent with the architecture. For example, the parametric model may include a neural network, and wherein the architecture identifies a number of layers of the neural network and a number of nodes per layer.

In embodiments in which the parametric model comprises a neural network, generating the parametric model may comprise assigning an initial weight to a connection between a first node in a first layer of the parametric model and a second node in a second layer of the parametric model of a second layer; generating an output suitable for the particle measurement; comparing the output to the transformed measurements of the particle; and adjusting the initial weight based at least in part on a difference between the output and the post-transform measurement, wherein a subsequent difference of the grain is less than the difference after adjusting the initial weight.

In another innovative aspect, a system is provided that includes one or more processing devices and a computer-readable storage medium containing instructions. The instructions, when executed by the one or more processing devices, cause the system to receive measurements of a first portion of particles associated with an experiment from a particle analyzer; converting the measured value into an initial transformed measured value by using nonparametric transformation; generating a parametric model to receive as input raw measurements of the first fraction of particles relevant to the experiment and to generate as output the initial transformed measurements; and configuring the particle analyzer. The particle analyzer is configured to convert measurements of particles included in a second portion of the particles based at least in part on the parametric model; generating a control signal to adjust an operational state of an analysis device included in the particle analyzer; and transmitting the control signal to the analysis device to achieve the operational state.

In some embodiments, the analysis device may include sorting electronics communicatively coupled with the deflection plate. The particle analyzer may be further configured to identify a target container applicable to a particle based at least in part on the post-conversion measurement value corresponding to the sorting criterion associated with the target container. Adjusting the operating condition may include applying an electrical charge through the deflection plate to direct the particles into a target vessel. In some embodiments, the analytical device may comprise a fluidic system, and adjusting the operating state comprises adjusting a pressure applied during the experiment.

Drawings

FIG. 1 illustrates a functional block diagram of an example of a transformation control system for analyzing and displaying biological events.

FIG. 2A is a schematic view of a particle sorter system according to an embodiment shown herein.

FIG. 2B is a schematic illustration of particle sorter system 200 according to an embodiment shown herein

Fig. 3 illustrates a functional block diagram of a particle analysis system for calculation-based sample analysis and particle characterization.

FIG. 4 is a diagram illustrating an example system for dynamically transforming event data.

FIG. 5 is a diagram illustrating an example system that utilizes a parametric machine learning transformation to dynamically transform event data.

FIG. 6 is a process flow diagram depicting an example of a parametric machine learning transformation method suitable for multi-dimensional event data.

Fig. 7 is a process flow diagram depicting an example of a method for sample sorting using a parametric machine learning transformation.

Detailed Description

Before the present invention is described in greater detail, it is to be understood that this invention is not limited to particular embodiments described, as variations in actual practice will certainly occur. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to limit the inventive concept, the scope of which will be limited only by the appended claims.

Where a range of values is provided, it is understood that each intervening value, to the extent that there is no such difference, to the upper and lower limits of that range, and any other stated or intervening value in that range, is encompassed within the invention. Unless the context clearly dictates otherwise, each intermediate value should be as low as one tenth of the unit of the lower limit. The upper and lower limits of these smaller ranges may independently be included in the smaller ranges, and are also encompassed within the invention, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the invention.

Certain ranges set forth herein precede the value by the term "about". The term "about" is used herein for the purpose of providing literal support for the precise number following the term, as well as numbers that are near or similar to the number following the term. In determining whether a number is near or approximate to a specifically recited number, a near or approximate non-recited number may be a number substantially equal to the specifically recited number in the context of its occurrence.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Representative exemplary methods and materials are described below, although methods and materials similar or equivalent to those described herein can also be used in the practice or testing of the present invention.

All publications and patents cited in this specification are herein incorporated by reference as if each individual publication or patent were specifically and individually indicated to be incorporated by reference and were set forth for the purpose of disclosing and describing the methods and/or materials associated with the cited publications. The citation of any publication is for its disclosure prior to the filing date and should not be construed as an admission that the present invention is not entitled to antedate such publication by virtue of prior invention. Further, the dates of publication provided may be different from the actual publication dates which may need to be independently confirmed.

It is noted that, as used herein and in the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. It should also be noted that claims may be drafted to exclude any optional element. Accordingly, this statement is intended to serve as antecedent basis for use of such exclusive terminology as "solely," "only," and the like in connection with the recitation of claim elements, or use of a "negative" limitation.

As will be apparent to those skilled in the art upon reading this disclosure, each of the individual embodiments described and illustrated in this patent has layered components and features that can be readily separated or combined with the features of any of the other several embodiments without departing from the scope and spirit of the present invention. Any recited method may be implemented in the order of events recited or in any other order that is logically possible.

Although the apparatus and method have or will be described for the sake of grammatical fluidity with functional explanations, it is to be expressly understood that the claims are not to be construed as necessarily limited in any way by the "means" or "steps" unless expressly stated in chapter 35 of the united states code, but rather are to be accorded the full scope of the meaning and equivalents of the definition set forth in the claims under the judicial doctrine of equivalents, and that the claims, when expressly written according to the provisions of chapter 112 of chapter 35 of the united states code, are to be accorded full statutory equivalents within chapter 112 of the united states code.

Non-parametric transformations, such as t-distributed random neighbor embedding (tSNE), may be used to analyze multi-parameter data, such as data derived from flow cytometry or other particle analysis systems and methods. The present invention may include these transformations for both descent and subpopulation identification (e.g., gating). Essentially, a non-parametric transformation cannot transform a new observation without training the new transformation based on the entire dataset (including the new observation). This presents significant challenges for computational-based particle analysis, as well as sample processing (e.g., sorting, instrument control, etc.) based on such analysis. The function parameterizes the non-parametric transformation with a neural network (e.g., an artificial neural network or other computational model trained to approximate the non-parametric transformation), thereby allowing a small training dataset to be transformed using non-parametric techniques, while allowing an accurate parametric model to be generated using the small training dataset to evaluate other events in a manner consistent with the initial events.

In some embodiments, the small training data set is transformed using a non-parametric transform such as tSNE (e.g., tens of thousands of examples). The small training data sets may be used to form a large training data set. For example, transformed/untransformed pairs are randomly extracted from the small training data set, and random noise is added to the transformed and untransformed data. The random draw may be weighted arbitrarily, for example by density, population of interest, or uniform weight. The added noise may be based on a probability distribution, including a uniform distribution, a gaussian distribution, or a poisson distribution.

The resulting large training data set is used to train a neural network to perform the transformation produced by the non-parametric transformation (which is performed on the small training data set). The neural network may have a variety of architectures with any number of layers, any number of neurons in a layer, and any activation function between connected neurons. However, to facilitate translation of the neural network into a configuration of the sorting electronics included in an analytical instrument, the network training may be limited to an architecture that can be translated into the target analytical instrument. For example, the neural network may consist of an ingress layer (with a number of nodes equal to the dimensionality of the untransformed data), four fully connected modified linear unit (re-lu) activation layers (30 nodes per layer), and an output layer (with two nodes, linear activation). Various error functions (e.g., mean square error, Kullback-Leibler divergence) and optimization programs (e.g., adapelta, RMSProp) may be used for training.

After training, the neural network approximates the transformation that was originally used to transform the small training data set. Thus, the neural network represents the transformation between the space occupied by the original, untransformed data and the space occupied by the transformed data. Thus, the neural network may transform new observations by feeding new data to the entry layer and collecting the resulting transformed data in the output layer. The output data can then be used in downstream applications, such as cell sorting decisions.

The functionality is a significant improvement over existing particle analysis systems, such as systems that include non-parametric transformations. Non-parametric transformations are commonly used to analyze multi-parameter data, such as flow cytometry data. For example, tSNE is used to perform dimension reduction processing on flow cytometry data to visualize cell subpopulations. The fundamental limitation of non-parametric transformations is that once trained, they cannot be used to transform new observations. Thus, the introduction of new data requires the re-execution of the entire transformation, which typically results in the appearance of a distinct transformation. Therefore, non-parametric transformations are of limited utility in areas where new observations need to be analyzed (e.g., FACS). The function overcomes these limitations by non-parametric transformations.

Other ways of parameterizing the non-parametric transformation rely on large training sets (e.g., about 100,000 or more data points). The techniques described herein utilize a small training set (useful for synthesizing representative data sets of sufficient size) to train machine learning models that produce accurate results using parametric variations. This saves time and material for the user and is more compatible with the sample size that is common when evaluating biological samples. For example, a typical flow cytometry setup experiment typically contains tens of thousands of data points. The number of these data points may be too small to train an accurate neural network. Requiring users to collect hundreds of thousands to millions of data points is time consuming and requires a large number of samples, which are often expensive and difficult to obtain.

Still other ways of parameterizing non-parametric transforms also rely on large complex neural networks with many layers and many nodes. These large networks are computationally intensive to evaluate and therefore cannot be implemented in sufficiently short computational times for low latency applications such as cell sorting on sorting electronics with limited processing resources. The techniques described herein utilize a relatively small number of layers and nodes so that the requirements of low latency implementations can be met on sorting electronics that are resource efficient, such as Field Programmable Gate Arrays (FPGAs).

In the case of flow sorting, the functionality allows for the collection of small data sets similar to those currently used to establish the sorting gating strategy. Then, a non-parametric transformation may be performed on the small dataset. For example, tSNE may be performed to reduce the dimensionality of the data. The system may then receive a gate specified using the tSNE transformed data. The pre-tSNE data and post-tSNE data are then used to train a neural network that learns the transformation between the raw data space and the reduced-dimension space. The trained neural network can then be programmed into the FPGA and used to sort cells (or other particles) for the actual sorting experiment. In previous implementations, sorting cannot be based on data transformed using a non-parametric transform.

In some embodiments, the transformation model comprises a dynamic algorithm, such as a machine learning algorithm. In a conventional sense, the term "machine learning" as used herein refers to adjustments to programming by computational methods that directly utilize data to determine and implement information without relying on predetermined equations as models. In certain embodiments, machine learning includes learning algorithms that look for patterns in data signals (e.g., from multiple particles in a flow cytometry sample). In these embodiments, the learning algorithm is configured to generate better, more accurate decisions and predictions from the number of data signals (i.e., the learning algorithm becomes more robust as the number of characterized particles from the sample increases). The destination machine learning scheme may include, but is not limited to, artificial neural networks, decision tree learning, decision tree predictive modeling, support vector machines, bayesian networks, dynamic bayesian networks, genetic algorithms, and other machine learning schemes.

A neural network or neural network model may be conceptualized as a network of nodes. The nodes may be organized into layers, with the first layer being the input layer for data inflow. The neural network may also include an output layer, wherein transformed data flows therefrom. Each individual node may have multiple inputs and a single output (e.g., an input layer node has only a single input). The output of the node represents a linear combination of the inputs. In other words, the input may be multiplied by a correlation constant. The products may be accumulated along a path of nodes having a constant offset. The constant offset or "bias" may represent another degree of freedom that may be adjusted during the training process. For example, for a re-lu based neural network model, the constant offset may be a threshold value because it has the ability to reduce the node value below zero, thereby causing the activation function to output zero.

The resulting value is evaluated using an activation function and used as the output of the node. A node in a given layer in the neural network is connected to each node in an adjacent layer. The neural network may be trained by: the expected output of the network is compared to the actual output of the network using a gradient descent algorithm and an error function to minimize the error of the network. The weights of one or more nodes may be adjusted to model the desired outcome produced by the network.

Typically, neural networks require a large amount of data to train. In this context, the small dataset may be trained by bootstrapping (generating synthetic data based on the small dataset). The term "bootstrapping" as used herein refers, in a conventional sense, to generating one or more additional data sets having a greater amount of data than an original data set, e.g., generating data sets having 10 or more data points (e.g., 100 or more data points, 1000 or more data points, 10000 or more data points, 100,000 or more data points, and including 1,000,000 or more data points) than the original data set. In some cases, the generated data set includes 2-fold or more, e.g., 3-fold or more, 5-fold or more, 10-fold or more, 25-fold or more, 50-fold or more, 100-fold or more, and including 1000-fold or more, data points as compared to the original data set. In some embodiments, bootstrapping includes random replaceable sampling. In some embodiments, bootstrapping includes estimating a characteristic of the estimation function (e.g., by variance) by measuring the characteristic when sampled from an approximate distribution (e.g., an empirical distribution function). Bootstrapping according to some embodiments may include using data from the flow cytometer in conjunction with noise components generated by the flow cytometer (e.g., vibration from the laser assembly or detection system and noise from the photodetector).

The neural network architecture is carefully chosen to minimize the number of nodes and layers per layer, allowing the neural network to be implemented in computing hardware (e.g., an FPGA). In addition, selecting an activation function that is compatible with the FPGA can further facilitate hardware implementation of the network. Hardware implementation is crucial for the data transformation (with sufficiently low latency for cell sorting applications).

The terms specifically set forth below as used herein have the following definitions. Unless otherwise defined in this section, all terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

As used herein, "system," "apparatus," "device" and "means" typically encompass hardware (e.g., mechanical and electronic hardware) and, in some embodiments, also the relevant software (e.g., a special-purpose computer program for graphics control) components.

As used herein, "event" or "event data" generally refers to a data packet measured from a single particle (e.g., a cell or synthetic particle). Typically, data measured from a single particle includes a number of parameters, including one or more light scattering parameters, and at least one parameter or characteristic, e.g., the intensity of the fluorescence, derived from the fluorescence detected from the particle. Thus, each event is represented as a measurement and a feature vector, where each measured parameter or feature corresponds to one dimension of the data space. In some embodiments, the data measured from a single particle includes image, electrical, temporal, or acoustic data. Events can be associated with an experiment, assay, or sample source (which can be identified by measurement data).

As used herein, a "population" or "subpopulation" of particles (e.g., cells or other particles) generally refers to a group of particles having a characteristic (e.g., optical, impedance, or temporal characteristics) relative to one or more measured parameters, which causes measured parameter data to form clusters in data space. Thus, the population is identified as a cluster in the data. Conversely, each data cluster is generally interpreted as corresponding to a particular type of cell or particle population, although clusters corresponding to noise or background are also typically observed. Clusters may be defined in a subset of the dimensions (e.g., relative to the subset of measured parameters) that correspond to populations that differ only in the subset of measured parameters or features (extracted from the cell or particle measurements).

As used herein, "gate" generally refers to a classifier boundary that identifies a subset of data of interest. In cytometry, a gate may be associated with a set of events of particular interest. "gating" as used herein generally refers to the process of classifying a given data set using defining gates, which may be one or more destination areas combined with Boolean logic, for that data.

As used herein, an "event" generally refers to an assembled package of data measured from a single particle (e.g., a cell or synthetic particle). Typically, the data measured from a single particle includes a number of parameters or characteristics, including one or more light scattering parameters or characteristics, and at least one other parameter or characteristic derived from the measured fluorescence. Thus, each event is represented as a vector of parameter and feature measurements, where each measured parameter or feature corresponds to a dimension of the data space.

Aspects of the invention also include systems having an optical detection system to characterize particles in a sample (e.g., cells in a biological sample) within a fluid medium. A system according to some embodiments includes: a light source configured to irradiate a sample (having particles) in a fluid medium; a light detection system comprising a photodetector configured to generate a data signal using light detected from particles within the fluid medium; and a processor having a memory operably coupled to itself, wherein the memory has instructions stored thereon that, when executed by the processor, cause the processor to: applying a non-parametric transform to a first set of data signals from the light detection system to generate a first transformed data set; generating a second transformed data set using the first transformed data set; generating a transformation model using the second transformed data set; and applying the transformation model to a second set of data signals from the light detection system.

The system of interest includes a light source configured to irradiate a sample in a fluid medium. In embodiments, the light source may be any suitable broadband or narrowband light source. Depending on the components (e.g., cells, microbeads, non-cellular particles, etc.) in the sample, the light source may be configured to emit light at different wavelengths ranging from 200nm to 1500nm, e.g., 250nm to 1250, 300nm to 1000nm, 350nm to 900nm, including 400nm to 800 nm. For example, the light source may include a broadband light source that emits light in the wavelength range of 200nm to 900 nm. In other cases, the light source comprises a narrow band light source that emits light in the wavelength range of 200nm to 900 nm. For example, the light source may be a narrow band LED (1 nm-25 nm) emitting light in the wavelength range of 200nm to 900 nm.

In some embodiments, the light source is a laser. The laser of interest may comprise a pulsed laser or a continuous wave laser. For example, the laser may be a gas laser, e.g., a helium-neon laser, an argon laser, a krypton laser, a xenon laser, a nitrogen laser, a CO2 laser, a CO laser, an argon fluoride (ArF) excimer laser, a krypton fluoride (KrF) excimer laser, a xenon chloride (XeCl) excimer laser, or a xenon fluoride (XeF) excimer laser, or a combination thereof; or a metal vapor laser, such as a helium cadmium (HeCd) laser, a helium mercury (HeHg) laser, a helium selenium (HeSe) laser, a helium silver (HeAg) laser, a strontium laser, a neon copper (NeCu) laser, a copper laser, or a gold laser, and combinations thereof; or a solid-state laser, for example a ruby laser, Nd: YAG laser, NdCrYAG laser, Er: YAG laser, Nd: YLF laser, Nd: YVO4 laser, Nd: YCa4O (BO3)3 laser, Nd: YCOB laser, titanium sapphire laser, thulim YAG laser, YAG ytterbium laser, ytterbium trioxide laser or cerium-doped laser and combinations thereof; or a semiconductor diode laser, an Optically Pumped Semiconductor Laser (OPSL), or a frequency doubled or tripled embodiment of any of the above.

In other embodiments, the light source is a non-laser light source, such as a lamp (including but not limited to halogen, deuterium, xenon arc), a light emitting diode (e.g., broadband LED with continuous spectrum, super-radiant LED, semiconductor light emitting diode, broad spectrum LED white light source, multi-LED integrated light source). In some cases, the non-laser light source is a stable fiber-coupled broadband light source, a white light source, and other light sources, or any combination thereof.

In some embodiments, the light source is a light beam generator configured to generate two or more beams of frequency-shifted light. In some cases, the optical beam generator includes a laser and a radio frequency generator configured to apply a radio frequency drive signal to the acousto-optic device to generate two or more angularly deflected laser beams. In these embodiments, the laser may be a pulsed laser or a continuous wave laser. For example, the laser in the target beam generator may be a gas laser, e.g., a helium-neon laser, an argon laser, a krypton laser, a xenon laser, a nitrogen laser, a CO2 laser, a CO laser, an argon fluoride (ArF) excimer laser, a krypton fluoride (KrF) excimer laser, a xenon chloride (XeCl) excimer laser, or a xenon fluoride (XeF) excimer laser, or a combination thereof; or a metal vapor laser, such as a helium cadmium (HeCd) laser, a helium mercury (HeHg) laser, a helium selenium (HeSe) laser, a helium silver (HeAg) laser, a strontium laser, a neon copper (NeCu) laser, a copper laser, or a gold laser, and combinations thereof; or solid-state lasers, such as ruby lasers, Nd: YAG lasers, NdCrYAG lasers, Er: YAG lasers, Nd: YLF lasers, Nd: YVO4 lasers, Nd: YCa4O (BO3)3 lasers, Nd: YCOB lasers, titanium sapphire lasers, thulim YAG lasers, YAG ytterbium lasers, ytterbium trioxide lasers or cerium-doped lasers and combinations thereof.

The acousto-optic device may be any suitable acousto-optic scheme configured to frequency shift the laser light using an applied acoustic wave. In some embodiments, the acousto-optic device is an acousto-optic deflector. The acousto-optic device in the destination system is configured to generate the angularly deflected laser beam using light from the laser and the applied radio frequency drive signal. The radio frequency drive signal may be applied to the acousto-optic device using any suitable radio frequency drive signal source, such as a direct digital frequency synthesizer (DDS), an Arbitrary Waveform Generator (AWG) or an electrical pulse generator.

In an embodiment, the controller is configured to apply an rf drive signal to the acousto-optic device to produce a desired number of angularly deflected laser beams in the output laser beam, for example, 3 or more rf drive signals, 4 or more rf drive signals, 5 or more rf drive signals, 6 or more rf drive signals, 7 or more rf drive signals, 8 or more rf drive signals, 9 or more rf drive signals, 10 or more rf drive signals, 15 or more rf drive signals, 25 or more rf drive signals, 50 or more rf drive signals, including being configured to apply 100 or more rf drive signals.

In some cases, to obtain the intensity distribution of the angularly deflected laser beam in the output laser beam, the controller is configured to apply the rf drive signals with different amplitudes, for example, in a range of about 0.001V to about 500V, about 0.005V to about 400V, about 0.01V to about 300V, about 0.05V to about 200V, about 0.1V to about 100V, about 0.5V to about 75V, about 1V to 50V, about 2V to 40V, 3V to about 30V, including about 5V to about 25V. In some embodiments, each applied rf drive signal has a frequency in the range of about 0.001MHz to about 500MHz, for example, about 0.005MHz to about 400MHz, about 0.01MHz to about 300MHz, about 0.05MHz to about 200MHz, about 0.1MHz to about 100MHz, about 0.5MHz to about 90MHz, about 1MHz to about 75MHz, about 2MHz to about 70MHz, about 3MHz to about 65MHz, about 4MHz to about 60MHz, including about 5MHz to about 50 MHz.

In certain embodiments, the controller has a processor with a memory operably coupled to itself such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam having an angularly deflected laser beam (with a desired intensity profile). For example, the memory may include instructions to generate two or more angularly deflected laser beams having the same intensity, e.g., 3 or more, 4 or more, 5 or more, 10 or more, 25 or more, 50 or more, and include the following memories: that is, a memory may be included that generates instructions to angularly deflect laser beams of 100 or more beams with the same intensity. In other embodiments, the memory may include instructions to generate two or more angularly deflected laser beams having different intensities, for example, 3 or more, 4 or more, 5 or more, 10 or more, 25 or more, 50 or more, and include the following memories: that is, a memory may be included that generates instructions to angularly deflect the laser beam 100 or more beams with different intensities.

In certain embodiments, the controller has a processor with a memory operably coupled to itself such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam having an intensity along a transverse axis that increases from edge to center thereof. In these cases, the intensity of the angularly deflected laser beam measured at the center of the output beam can be 0.1% to about 99% of the intensity of the angularly deflected laser beam measured at the edges of the output laser beam along the transverse axis, e.g., 0.5% to about 95%, 1% to about 90%, about 2% to about 85%, about 3% to about 80%, about 4% to about 75%, about 5% to about 70%, about 6% to about 65%, about 7% to about 60%, about 8% to about 55%, including about 10% to about 50% of the intensity of the angularly deflected laser beam measured at the edges of the output laser beam along the transverse axis. In other embodiments, the controller has a processor with a memory operably coupled to itself such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam having an intensity that increases along a transverse axis from an edge to a center thereof. In these cases, the intensity of the angularly deflected laser beam measured at the edge of the output beam can be 0.1% to about 99% of the intensity of the angularly deflected laser beam measured at the center of the output laser beam along the transverse axis, e.g., 0.5% to about 95%, 1% to about 90%, about 2% to about 85%, about 3% to about 80%, about 4% to about 75%, about 5% to about 70%, about 6% to about 65%, about 7% to about 60%, about 8% to about 55%, including about 10% to about 50% of the intensity of the angularly deflected laser beam measured at the center of the output laser beam along the transverse axis. In other embodiments, the controller has a processor with a memory operably coupled to itself such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam whose intensity distribution along the transverse axis conforms to a gaussian distribution. In certain embodiments, the controller has a processor with a memory operably coupled to itself such that the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate an output laser beam having a top hat shaped intensity profile along a transverse axis.

In an embodiment, the destination beam generator may be configured to generate spatially separated angularly deflected laser beams in the output laser beam. The spacing between the angularly deflected laser beams may be 0.001m or more, for example, 0.005 μm or more, 0.01 μm or more, 0.05 μm or more, 0.1 μm or more, 0.5 μm or more, 1 μm or more, 5 μm or more, 10 μm or more, 100 μm or more, 500 μm or more, 1000 μm or more, including 5000 μm or more, depending on the applied radio frequency drive signal and the desired irradiation profile of the output laser beam. In some embodiments, the system is configured to produce overlapping angularly deflected laser beams in the output laser beam, e.g., overlapping adjacent angularly deflected laser beams distributed along a horizontal axis of the output laser beam. The overlap (e.g., spot overlap) between adjacent angularly deflected laser beams may be 0.001 μm or more, e.g., 0.005 μm or more, 0.01 μm or more, 0.05 μm or more, 0.1 μm or more, 0.5 μm or more, 1 μm or more, 5 μm or more, 10 μm or more, including 100 μm or more.

In some cases, a light beam generator configured to generate two or more beams of frequency-shifted light includes a laser excitation module described in the following publications: us patent nos. 9,423,353, 9,784,661, and 10,006,852; and U.S. patent publication nos. 2017/0133857 and 2017/0350803, the contents of which are incorporated herein by reference.

In an embodiment, the system comprises a light detection system having one or more of a bright field photodetector, a light scatter (forward light scatter, side light scatter) detector, and a fluorescence detector for detecting and measuring light from said sample. The bright field, light scattering and fluorescence detectors of interest may include, but are not limited to, optical sensors such as Active Pixel Sensors (APS), avalanche photodiodes, image sensors, Charge Coupled Devices (CCD), enhanced charge coupled devices (ICCD), light emitting diodes, photon counters, bolometers, pyroelectric detectors, photoresistors, photovoltaic cells, photodiodes, photomultipliers, phototransistors, quantum dot photoconductors or photodiodes, and combinations thereof, among others. In certain embodiments, light from the sample is measured using a Charge Coupled Device (CCD), a semiconductor Charge Coupled Device (CCD), an Active Pixel Sensor (APS), a Complementary Metal Oxide Semiconductor (CMOS), an image sensor, or an N-type metal oxide semiconductor (NMOS) image sensor. In certain embodiments, the bright field photodetector comprises an Avalanche Photodiode (APD). In some cases, the light scattering detector is an avalanche photodiode. In some cases, one or more of the fluorescence detectors are avalanche photodiodes.

In some embodiments, the light detection system of interest comprises a plurality of fluorescence detectors. In some cases, the light detection system includes a plurality of solid state detectors, such as photodiodes. In some cases, the light detection system includes an array of fluorescence photodetectors, such as an array of photodiodes. In these embodiments, the photodetector array may include 4 or more photodetectors, for example, 10 or more photodetectors, 25 or more photodetectors, 50 or more photodetectors, 100 or more photodetectors, 250 or more photodetectors, 500 or more photodetectors, 750 or more photodetectors, including 1000 or more photodetectors. For example, the detector may be a photodiode array having 4 or more photodiodes, e.g., 10 or more photodiodes, 25 or more photodiodes, 50 or more photodiodes, 100 or more photodiodes, 250 or more photodiodes, 500 or more photodiodes, 750 or more photodiodes, including 1000 or more photodiodes.

The photodetectors may be arranged in any geometric configuration as desired, with the intended arrangement configurations including, but not limited to, square configurations, rectangular configurations, trapezoidal configurations, triangular configurations, hexagonal configurations, heptagonal configurations, octagonal configurations, nonagonal configurations, decagonal configurations, dodecagonal configurations, circular configurations, elliptical configurations, and irregular pattern configurations. The photodetectors in the photodetector array may be oriented at an angle to one another (as described by the X-Z plane) in the range of 10 ° to 180 °, for example, 15 ° to 170 °,20 ° to 160 °, 25 ° to 150 °,30 ° to 120 °, including 45 ° to 90 °. The photodetector array may be any suitable shape and may be rectilinear (e.g., square, rectangular, trapezoidal, triangular, hexagonal, etc.), curvilinear (e.g., circular, elliptical), and irregular (e.g., parabolic bottom connected to a planar top). In certain embodiments, the photodetector array has a rectangular active surface.

The active surface width of each photodetector (e.g., photodiode) in the array ranges from 5 μm to 250 μm, e.g., 10 μm to 225 μm, 15 μm to 200 μm, 20 μm to 175 μm, 25 μm to 150 μm, 30 μm to 125 μm, including 50 μm to 100 μm; a length in the range of 5 μm to 250 μm, for example, 10 μm to 225 μm, 15 μm to 200 μm, 20 μm to 175 μm, 25 μm to 150 μm, 30 μm to 125 μm, including 50 μm to 100 μm; wherein each photodetector (e.g., photodiode) in the array has a surface area in the range of 25 μm2 to 10000 μm2, e.g., 50 μm2 to 9000 μm2, 75 μm2 to 8000 μm2, 100 μm2 to 7000 μm2, 150 μm2 to 6000 μm2, including 200 μm2 to 5000 μm 2.

The size of the photodetector array may vary depending on the amount and intensity of light, the number of photodetectors, and the desired sensitivity, and may range in length from 0.01mm to 100mm, e.g., from 0.05mm to 90mm, from 0.1mm to 80mm, from 0.5mm to 70mm, from 1mm to 60mm, from 2mm to 50mm, from 3mm to 40mm, from 4mm to 30mm, including from 5mm to 25 mm. The width of the photodetector array may also vary, ranging from 0.01mm to 100mm, for example, 0.05mm to 90mm, 0.1mm to 80mm, 0.5mm to 70mm, 1mm to 60mm, 2mm to 50mm, 3mm to 40mm, 4mm to 30mm, including 5mm to 25 mm. Thus, the photodetector array active surface has a surface area in the range of 0.1mm2 to 10000mm2, for example 0.5mm2 to 5000mm2, 1mm2 to 1000mm2, 5mm2 to 500mm2, including 10mm2 to 100mm 2.

The photodetector of interest is configured to measure light collected at one or more wavelengths, e.g., 2 or more wavelengths, 5 or more different wavelengths, 10 or more different wavelengths, 25 or more different wavelengths, 50 or more different wavelengths, 100 or more different wavelengths, 200 or more different wavelengths, 300 or more different wavelengths, including light emitted by a sample in a fluid medium at 400 or more different wavelengths.

In some embodiments, the light detection system comprises a bright field photodetector configured to generate a bright field data signal. The bright field photodetector may be configured to detect light from the sample at one or more wavelengths, such as 5 or more different wavelengths, 10 or more different wavelengths, 25 or more different wavelengths, 50 or more different wavelengths, 100 or more different wavelengths, 200 or more different wavelengths, 300 or more different wavelengths, including detecting light at 400 or more different wavelengths. The bright field photodetector may be configured to detect light at one or more wavelengths in the wavelength range 200 nm-1200 nm. In some cases, the method includes detecting light from the sample with a bright field photodetector over a range of wavelengths, e.g., 200nm to 1200nm, 300nm to 1100, 400nm to 1000nm, 500nm to 900nm, including 600nm to 800 nm.

In certain embodiments, the bright field photodetector in the light detection system of interest is configured to generate one or more bright field data signals responsive to the detected light, e.g., 2 or more, 3 or more, 4 or more, 5 or more, and including 10 or more bright field data signals responsive to the detected light. Where the bright field photodetector is configured to detect light at multiple wavelengths of light (e.g., 400nm to 800nm), the method may, in some cases, include generating one or more bright field data signals responsive to each wavelength of the detected light. In other cases, a single bright field data signal is generated in response to light detected by the bright field photodetector over the entire wavelength range.

In some embodiments, the light detection system includes a light scatter photodetector configured to generate a light scatter data signal. The light scattering photodetector may be configured to detect light from the sample at one or more wavelengths, such as 5 or more different wavelengths, 10 or more different wavelengths, 25 or more different wavelengths, 50 or more different wavelengths, 100 or more different wavelengths, 200 or more different wavelengths, 300 or more different wavelengths, including detecting light at 400 or more different wavelengths. The light scattering photodetector may be configured to detect light at one or more wavelengths in the wavelength range 200 nm-1200 nm. In some cases, the method includes detecting light from the sample with a light scattering photodetector over a range of wavelengths, e.g., 200nm to 1200nm, 300nm to 1100, 400nm to 1000nm, 500nm to 900nm, including 600nm to 800 nm.

In certain embodiments, the light scattering photodetectors in the destination light detection system are configured to generate one or more light scattering data signals responsive to the detected light, e.g., 2 or more, 3 or more, 4 or more, 5 or more, and including 10 or more light scattering data signals responsive to the detected light. Where the light scattering photodetector is configured to detect light at multiple wavelengths of light (e.g., 400nm to 800nm), methods may, in some cases, include generating one or more light scattering data signals responsive to each wavelength of the detected light. In other cases, a single light scatter data signal is generated in response to light detected by the light scatter photodetector over the entire wavelength range.

The light detection system comprises one or more bright field, light scattering or fluorescence detectors, for example 2 or more, 3 or more, 4 or more, 5 or more, 6 or more, 7 or more, 8 or more, 9 or more, 10 or more, 15 or more, including 25 or more detectors. In an embodiment, each of the detectors is configured to generate a data signal. Light from the sample can be detected with each detector independently at one or more wavelengths in the wavelength range 200 nm-1200 nm. In some cases, one or more detectors are configured to detect light from the sample over a range of wavelengths, e.g., 200nm to 1200nm, 300nm to 1100, 400nm to 1000nm, 500nm to 900nm, including 600nm to 800 nm. In other cases, one or more detectors are configured to detect light at one or more specific wavelengths. For example, the fluorescence may be detected at one or more wavelengths of 450nm, 518nm, 519nm, 561nm, 578nm, 605nm, 607nm, 625nm, 650nm, 660nm, 667nm, 670nm, 668nm, 695nm, 710nm, 723nm, 780nm, 785nm, 647nm, 617nm, and any combination thereof, depending on the number of different detectors in a target light detection system. In certain embodiments, one or more detectors are configured to detect wavelengths of light corresponding to fluorescence peak wavelengths of certain fluorophores in the sample.

The light detection system is configured to measure light continuously or at discrete time intervals. In some cases, a detector in the light detection system is configured to continuously measure the collected light. In other cases, the light detection system is configured to measure at discrete time intervals, e.g., the time intervals of the measurement light are 0.001 milliseconds, 0.01 milliseconds, 0.1 milliseconds, 1 millisecond, 10 milliseconds, 100 milliseconds, including 1000 milliseconds, or some other time interval.

In some embodiments, the system is configured to generate frequency-encoded fluorescence data by irradiating a sample (having particles) in a fluid medium. In some embodiments, the light source includes a light generator assembly that generates a plurality of angularly deflected laser beams, each laser beam having an intensity that is based on the magnitude of an applied radio frequency drive signal (e.g., from a direct digital combiner coupled to an acousto-optic device). For example, the subject system may include a light generator assembly that generates 2 or more angularly deflected laser beams, e.g., 3 or more, 4 or more, 5 or more, 6 or more, 7 or more, 8 or more, 9 or more, 10 or more, and including 25 or more angularly deflected laser beams. In an embodiment, each of the angularly deflected laser beams has a different frequency that is offset by a predetermined radio frequency compared to the frequency of the input laser beam.

The subject systems are configured, according to certain embodiments, to generate angularly deflected laser beams that are additionally spatially offset from one another. Depending on the applied radio frequency drive signal and the desired irradiation profile of the output laser beam, the subject system may be configured to generate angularly deflected laser beams, the spacing between which may be 0.001 μm or more, e.g., 0.005 μm or more, 0.01 μm or more, 0.05 μm or more, 0.1 μm or more, 0.5 μm or more, 1 μm or more, 5 μm or more, 10 μm or more, 100 μm or more, 500 μm or more, 1000 μm or more, including 5000 μm or more. In some embodiments, the angularly deflected laser beams overlap, e.g., with adjacent angularly deflected laser beams distributed along a horizontal axis of the output laser beam. The overlap (e.g., spot overlap) between adjacent angularly deflected laser beams may be 0.001 μm or more, e.g., 0.005 μm or more, 0.01 μm or more, 0.05 μm or more, 0.1 μm or more, 0.5 μm or more, 1 μm or more, 5 μm or more, 10 μm or more, including 100 μm or more.

In some embodiments, a system includes a processor having a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to: frequency encoded fluorescence data is generated by calculating the difference between the optical frequencies of incident overlapping beamlets of light on the fluid medium. In an example, a system includes a processor having a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to calculate a beat frequency at each position on a lateral axis of the fluid medium. In these embodiments, the frequency encoded fluorescence emitted by the particles is a beat frequency corresponding to a difference between the frequency of the local oscillator beam (fLO) and the frequency of the radio frequency offset sub-beam. For example, the frequency encoded fluorescence data includes the beat frequency of the fLO-fRF shifted sub-beams. Where the irradiation of the fluid medium includes a local oscillator spanning a width (e.g., the entire transverse axis) of the fluid medium, the frequency encoded fluorescence data includes beat frequencies corresponding to differences between the frequency of the local oscillator beam (fLO) and the frequencies of the radio frequency offset sub-beams (f1, f2, f3, f4, f5, f6, etc.). In these embodiments, the frequency encoded fluorescence data may include a plurality of beat frequencies, each beat frequency corresponding to a position on a lateral axis of the fluid medium.

In some embodiments, the subject systems include a particle sorting component for sorting particles (e.g., cells) in the sample. In some cases, the particle sorting component is a particle sorting module, for example, a particle sorting module described in the following publications: U.S. patent publication No. 2017/0299493, filed on 28.3.2017; and U.S. provisional patent application No. 62/752,793, filed on 30.10.2018, the contents of which are incorporated herein by reference. In certain embodiments, the particle sorting assembly includes one or more droplet deflectors, such as those described in U.S. patent publication No. 2018/0095022, filed on 2017, 6, and 14, the contents of which are incorporated herein by reference.

In some embodiments, the subject system is a flow cytometry system. Suitable flow cytometry systems may include, but are not limited to, the systems described in the following publications: ormerod (eds), flow cytometry: practical methods, oxford university press (1997); jarosszeski et al (eds.), flow cytometry protocols, Methods in Molecular Biology, 91 st, Humana Press (1997); practical flow cytometry, third edition, Wiley-Liss (1995); virgo et al (2012), annual clinical biochemistry, 1 month; 49(pt 1) 17 to 28; linden et al, seminal symposium on thrombosis and hemostasis, 10 months 2004; 30(5): 502-11; alison et al, Pathology journal, 12 months 2010; 222(4): 335-344; and Herbig et al (2007) Crit Rev Therg Drug Carrier Syst.24 (3): 203-255; the contents of these publications are incorporated herein by reference. In some cases, the flow cytometry systems of interest include BD Biosciences FACSCPitot II flow cytometer, BD AccuriTM flow cytometer, BD Biosciences FACSCesta flow cytometer, BD Biosciences FACSCLyricTM flow cytometer, BD Biosciences FACSCverse flow cytometer, BD Biosciences FACSCEPH flow cytometer, BD Biosciences LSROrient TM X-20 flow cytometer, BD Biosciences FACSCalibur TM cell sorter, BD Biosciences FACSCALIG TM cell sorter, BD Biosciences FACSCOlRIcTM cell sorter, BD Biosciences ViaTM cell sorter, BD Biosciences BD bioscience cell sorter, BD Biosciences ViaTM cell sorter, BD Biosciences FACSCUx flow cytometer, BD Biosciences FACSCOr flow cytometer, BD Biosciences FACSCOrthoRTM flow cytometer, BD Biosciences FACSCORE flow cytometer, and BD Biosciences^TMCell sorter, BD Biosciences Jazz^TMCell sorter, BD Biosciences Aria^TMCell sorter and BD Biosciences FACCSmolody^TMCell sorters, and the like.

In some embodiments, the subject particle sorting system is a flow cytometry system, for example, the system described in the following numbered U.S. patents: 10,006,852, respectively; 9,952,076, respectively; 9,933,341, respectively; 9,784,661, respectively; 9,726,527, respectively; 9,453,789, respectively; 9,200,334, respectively; 9,097,640, respectively; 9,095,494, respectively; 9,092,034, respectively; 8,975,595, respectively; 8,753,573, respectively; 8,233,146, respectively; 8,140,300, respectively; 7,544,326, respectively; 7,201,875, respectively; 7,129,505, respectively; 6,821,740, respectively; 6,813,017, respectively; 6,809,804, respectively; 6,372,506, respectively; 5,700,692, respectively; 5,643,796, respectively; 5,627,040, respectively; 5,620,842, respectively; 5,602,039; the contents of these documents are incorporated herein by reference in their entirety.

In some cases, the subject system is a flow cytometry system configured to characterize and image particles in fluid media by fluorescence imaging using radio frequency marker emission (FIRE), such as the systems described in the following publications: diebold et al, natural photonics, volume 7 (10); 806, 810 (2013); us patent nos. 9,423,353, 9,784,661, and 10,006,852; and U.S. patent publication nos. 2017/0133857 and 2017/0350803, the contents of which are incorporated herein by reference.

Specific examples of various embodiments and systems are described further below (which can be implemented).

FIG. 1 illustrates a functional block diagram of an example of a transformation control system for analyzing and displaying biological events. The analysis controller 190 may be configured to implement various processes for controlling the graphical display of biological events.

The particle analyzer 102 may be configured to collect biological event data. For example, a flow cytometer may generate flow cytometry event data. The particle analyzer 102 may be configured to provide the biological event data to the analysis controller 190. A data communication channel may be included between particle analyzer 102 and analysis controller 190. The biological event data may be provided to the analysis controller 190 via the data communication channel.

The analysis controller 190 may be configured to receive the biological event data from the particle analyzer 102. The biological event data received from the particle analyzer 102 may include flow cytometry event data. The analysis controller 190 may be configured to provide a graphical display including the first plot of biological event data to the display device 106. The analysis controller 190 may be further configured to render a destination area as a door, overlaid on the first map, around the plurality of biological event data shown by the display device 106. In some embodiments, the gate may be a logical combination of one or more destination graphic regions plotted on a single parameter histogram or bivariate plot.

The analysis controller 190 may be further configured to display the bio-event data on the display device 106 inside the door differently than other events in the bio-event data outside the door. For example, the analysis controller 190 may be configured to cause the color of the biological event data contained inside the door to be different from the color of the biological event data outside the door. Display device 106 may be implemented in the form of a display, tablet computer, smart phone, or other electronic device configured to present a graphical interface.

The analysis controller 190 may be configured to receive a gate selection signal from the first input device identifying the gate. For example, the first input device may be implemented in the form of a mouse 110. The mouse 110 may issue a door selection signal to the analysis controller 190 to determine a door to be displayed on the display device 106 or manipulated via the display device 106 (e.g., click on or within a desired door when a cursor is located there). In some embodiments, the first device may be implemented in the form of a keyboard 108 or other means for providing input signals to the analysis controller 190 (e.g., a touch screen, a stylus, an optical detector, or a voice recognition system). Some input devices may include multiple input functions. In such embodiments, the input functions may each be considered an input device. For example, as shown in FIG. 1, mouse 110 may include a right mouse button and a left mouse button, both of which may generate a triggering event.

The triggering event may cause the analysis controller 190 to change the manner in which data is displayed (the partial data is actually displayed on the display device 106), or provide input for further processing, such as selecting a population of interest for particle sorting.

In some embodiments, analysis controller 190 may be configured to detect when mouse 110 activates a door selection. The analysis controller 190 may be further configured to automatically modify the graph visualization to facilitate the gating process. The modification may be based on a particular distribution of the biological event data received by the analysis controller 190.

The analysis controller 190 may be connected to the storage device 104. The storage device 104 may be configured to receive and store the biological event data from the analysis controller 190. The storage device 104 may also be configured to receive and store flow cytometry event data from the analysis controller 190. The storage device 104 may also be configured to allow the analysis controller 190 to retrieve biological event data, e.g., flow cytometry event data.

The display device 106 may be configured to receive display data from the analysis controller 190. The display data may comprise a graph of the biological event data and a gate summarizing a portion of the graph. Display device 106 may be further configured to change the information presented based on input received from analysis controller 190 and input received from particle analyzer 102, storage device 104, keyboard 108, and/or mouse 110.

A common flow sorting technique (known as "electrostatic cell sorting") employs droplet sorting in which a stream or moving liquid column containing linearly separated particles is broken up into droplets, and the droplets containing the particles of interest are charged and deflected by an electric field into a collection tube. Current droplet sorting systems are capable of forming droplets at a rate of 100,000 drops/second in a fluid medium passing through a nozzle having a diameter of less than 100 microns. Droplet sorting generally requires that the droplets break off from the stream at a distance from the nozzle head. The distance is typically about a few millimetres from the nozzle head and for undisturbed fluid media the distance is stable and can be maintained by oscillating the nozzle head at a predetermined frequency and an amplitude at which the detachment remains constant. For example, in some embodiments, the amplitude of the sinusoidal waveform voltage pulses is adjusted at a given frequency to keep the detachment stable and constant.

Typically, the linearly separated particles in the stream are characterized in that they pass through an observation point located within the flow cell or cuvette or directly below the nozzle head. Once it is determined that the particles meet one or more desired criteria, the time at which they reach the point of droplet break-up and break off from the stream in the form of droplets can be predicted. Ideally, the fluid medium is briefly charged before the droplets containing the selected particles are separated from the stream, and then immediately grounded after droplet break-up. The droplets to be sorted remain charged when they are detached from the fluid medium, while all other droplets are uncharged. The charged liquid drops are deviated from the descending tracks of other liquid drops under the action of the electric field and are collected in the sample tube. The uncharged droplets fall directly into a discharge pipe.

FIG. 2A is a schematic view of a particle sorter system according to an embodiment shown herein. Particle sorter system 250 shown in fig. 2A includes deflection plates 252 and 254. The charge is applied through the flow charging wire in barb 256. This produces a stream of particles 260 for analysis. The particles may be illuminated with one or more light sources (e.g., lasers) to produce light scattering and generate fluorescence information. The particle information is analyzed, such as by sorting electronics or other detection systems (not shown in fig. 2A). The deflection plates 252 and 254 may be independently controlled to attract or repel the charged droplets to direct the droplets into a destination collection container (e.g., one of 272, 274, 276, or 278). As shown in fig. 2A, the deflector plate may be controlled to direct particles along the first path 262 toward the vessel 274 or along the second path 268 toward the vessel 278. If the particle is not a particle of interest (e.g., does not display scattering or illumination information within a specified sorting range), the deflector plate may continue the flow of the particle along the flow path 264. Such uncharged droplets may enter a waste container via, for example, aspirator 270.

The sorting electronics may be included to begin collecting measurements, receive fluorescence signals of particles, and determine how to adjust the deflection plates to sort the particles. An exemplary embodiment of the embodiment shown in FIG. 2A includes BD FACSAria marketed by Becton, Dickinson and Company, san Jose, Calif^TMA series of flow cytometers.

FIG. 2B is a schematic view of a particle sorter system 200 according to an embodiment shown herein. In some embodiments, particle sorter system 200 is a cell sorter system. As shown in fig. 2B, a drop formation sensor 202 (e.g., a piezoelectric oscillator) is coupled to a fluid conduit 201 (e.g., a nozzle). Within the fluid conduit 201, the sheath fluid 204 hydrodynamically focuses the sample fluid 206 into a moving liquid column 208 (e.g., stream). Within the moving fluid column 208, particles (e.g., cells) are aligned to pass through a monitored region 210 (e.g., laser flow intersection) irradiated by an irradiation source 212 (e.g., laser). The drop formation sensor 202 vibrates so that the moving liquid column 208 breaks up into droplets 209.

In operation, the detection station 214 (e.g., event detector) determines when a particle of interest (or cell of interest) passes through the monitored region 210. The detection station 214 feeds a timing circuit 228, which in turn feeds an instantaneous charging circuit 230. At the droplet break-up point, after notification of a timed droplet delay (Δ t), an instantaneous charge is applied to the moving liquid column 208 to charge the target droplet. The droplet of interest may comprise one or more particles or cells to be sorted. The charged droplets may then be sorted by activating a deflection plate (not shown), deflecting the droplets into a container such as a collection tube or a multi-well sample plate, where the wells may be associated with specific purpose droplets. As shown in fig. 2B, the droplets are collected in a drain container 238.

A detection system 216 (e.g., a drop boundary detector) is used to automatically determine the phase of the drop drive signal as the particle of interest passes through the monitored region 210. An exemplary drop boundary detector is described in U.S. patent No. 7,679,039, which is incorporated herein by reference in its entirety. Detection system 216 allows the instrument to accurately calculate the position of each detected particle in the droplet. The detection system 216 may be fed with the amplitude signal 220 and/or the phase 218 signal, which in turn is fed (via amplifier 222) to the amplitude control circuit 226 and/or the frequency control circuit 224. Amplitude control circuit 226 and/or frequency control circuit 224 in turn control drop formation sensor 202. The amplitude control circuit 226 and/or the frequency control circuit 224 may be included in the control system.

In some embodiments, the sorting electronics (e.g., detection system 216, detection station 214, processor 240) may be coupled with a memory configured to store detected events and sorting decisions based thereon. The sorting decision may be included in event data of the particle. In some embodiments, detection system 216 and detection station 214 may be implemented in the form of a single detection unit or communicatively coupled such that event measurements may be collected by one of detection system 216 or detection station 214 and provided to the non-collecting element.

In some embodiments, one or more of the components described as suitable for particle sorter system 200 may be used for particle analysis and characterization, whether or not the particles are physically sorted into collection vessels. Likewise, one or more of the following components suitable for use in particle analysis system 300 (fig. 3) may also be used for particle analysis and characterization, whether or not the particles are physically sorted into collection containers. For example, particles may be grouped or displayed in a tree comprising at least three groupings as described herein using one or more of the components described in particle sorter system 200 or particle analysis system 300.

Fig. 3 illustrates a functional block diagram of a particle analysis system for calculation-based sample analysis and particle characterization. In some embodiments, particle analysis system 300 is a flow system. The particle analysis system 300 shown in fig. 3 may be configured to perform, in whole or in part, the methods described herein, e.g., the method shown in fig. 5 or the method shown in fig. 8. Particle analysis system 300 includes fluidic system 302. The fluidic system 302 may include or be coupled with a sample tube 310 in which particles 330 (e.g., cells) in a sample move along a common sample path 320 and a moving liquid column within the sample tube.

The particle analysis system 300 includes a detection system 304 configured to collect a signal from each particle as it passes through one or more detection stations along the common sample path. The detection station 308 generally refers to the monitored region 340 of the common sample path. In some embodiments, detecting may include detecting light or one or more other characteristics of the particle 330 as it passes through the monitored region 340. In fig. 3, an inspection station 308 having an area 340 to be monitored is shown. Some embodiments of the particle analysis system 300 may include multiple detection stations. In addition, some inspection stations may monitor more than one area.

A signal value is assigned to each signal to form a data point for each particle. As described above, the data may be referred to as event data. The data points may be multi-dimensional data points comprising values for each measured characteristic of the particle. The detection system 304 is configured to collect a series of the data points over a first time interval.

Particle analysis system 300 also includes a control system 306. The control system 306 may include one or more processors, amplitude control circuitry 226, and/or frequency control circuitry 224 (shown in fig. 2B). Control system 306 is shown operatively associated with fluidic system 302. Control system 306 is configured to generate a calculated signal frequency for at least a portion of the first time interval based on the poisson distribution and the number of data points collected by detection system 804 during the first time interval. Control system 306 is further configured to generate an experimental signal frequency based on the number of data points within the portion of the first time interval. Further, the control system 306 may compare the experimental signal frequency to a calculated signal frequency or a predetermined signal frequency.

FIG. 4 is a diagram illustrating an example system for dynamically transforming event data. The system 400 includes a transformation device 420. The transformation device 420 comprises an event data receiver 422. Event data receiver 422 may receive event data 402 from a particle analyzer (e.g., particle analyzer 102 shown in fig. 1). In some embodiments, event data 402 may be generated by a particle analyzer by receiving from, for example, an analysis workstation. For example, a user may provide event data 402 obtained from a particle analyzer to event data receiver 422. The event data receiver 422 may include a transceiver for wireless communication or a port for connecting to a wired network such as an Ethernet local area network or a network interface such as via a universal serial bus orThe connector is connected to a port of the device.

The event data receiver 422 may provide at least a portion of the event data 402 to transform a selection circuit 424 included in the transformation device 420. The transform selection circuit 424 may identify transforms that are applicable to the event data. The identifying may comprise detecting a value in the event data, such as an identifier for the assay or experiment. The available transforms may be stored in a data store 440 accessible to transform selection circuitry 424. As discussed above, the transformation may be a parametric transformation or a non-parametric transformation. In some embodiments, the transformation may be specified by the device providing the event data 402. For example, the analysis workstation may submit a message requesting processing of the event data 402. The message may include the required transformation (e.g., tSNE).

The transformation device 420 may include event data transformation circuitry 426 to process the event data 402 according to the identified transformation. Event data transformation circuitry 426 may generate transformed event data 490. The transformed event data 490 may be transmitted from the transformation device 420 to a target device, such as a device that verifies the event data 402, a storage location, or an addressable network service (e.g., a laboratory information system).

As described above, when performing non-parametric transformations, consistent handling of event data received at different times is unpredictable due to the statistical nature of the transformations. Thus, the transformation device 420 in FIG. 4 may provide appropriate results when the event data 402 represents measurements for a completed experiment. However, for certain analysis methods (e.g., establishing sorting criteria), the user may make a best guess as to the expected location of the population of interest, or take measurements of a portion of the sample, which are then used to specify a gate. In the first case, the guessing may not be effective in finding the true population of interest, and often may require taking a larger sample size, which may be impractical for some experiments. In the case of an initial portion of the analysis, any resulting gate requires a transformation of a future event in the same manner as the initial portion of the event if the event data is transformed using a non-parametric transformation. As described above, non-parametric transformations are typically uncertain, and thus event data processed at two different times may provide two different results.

FIG. 5 is a diagram illustrating an example system that utilizes a parametric machine learning transformation to dynamically transform event data. The transformation device 520 in fig. 5 includes functionality to parameterize the non-parametric transformation using a neural network, thereby allowing the small training dataset to be transformed using non-parametric techniques, while allowing an accurate parametric model to be generated using the small training dataset to evaluate other events in a manner consistent with the initial events.

The system 500 includes a transformation device 520. The transformation device 520 comprises an event data receiver 522. Event data receiver 522 may receive event data 502 from a particle analyzer (e.g., particle analyzer 102 shown in fig. 1). In some embodiments, event data 502 may be generated by a particle analyzer by receiving from, for example, an analysis workstation. For example, a user may provide event data 502 obtained from a particle analyzer to event data receiver 522. The event data receiver 522 may include a transceiver for wireless communication or a network port for connection to a wired network such as an ethernet local area network.

Event data receiver 522 may activate switch 524 or other routing element for further processing. Switch 524 may split the connection between transform selection circuit 540 and training circuit 530. Switch 524 may be activated based on: event data 502; information included in the event data 502 or included with the event data 502, such as a message requesting processing of the event data 502; characteristics of the device providing the event data 502; or other factors that may be detected by the event data receiver 522 or the switch 524.

When the switch 524 activates the path to the transform selection circuit 540, at least a portion of the event data 502 is provided to the transform selection circuit 540. As with the transform selection circuit 424 of fig. 4, the transform selection circuit 540 may also identify transforms that are applicable to the event data. The identifying may comprise detecting a value in the event data, such as an identifier for the assay or experiment. The available transformations may be stored in a data store 540 accessible to transformation selection circuitry 540. As discussed above, the transformation may be a parametric transformation or a non-parametric transformation. In some embodiments, the transformation may be specified by the device providing the event data 502. For example, the analysis workstation may submit a message requesting processing of the event data 502. The message may include the required transformation (e.g., tSNE).

The transformation device 520 may include an event data transformation circuit 542 to process the event data 502 according to the identified transformation. Event data transform circuitry 542 may generate transformed event data 570. The transformed event data 570 may be transmitted from the transformation device 520 to a target device, such as a device that verifies the event data 502, a storage location, or an addressable network service (e.g., a laboratory information system).

In the second mode, the switch 524 may be activated to route at least a portion of the event data 502 to the training circuit 530. The training circuit 530 may coordinate parametric model generation for transforming the event data 502. The transformation device 520 may include a training event data generator 532 to generate a suitable training data set for the machine learning model. The neural network may have a variety of architectures with any number of layers, any number of neurons in a layer, and any activation function between connected neurons. However, to facilitate translation of the neural network into a configuration of the sorting electronics included in an analytical instrument, the network training may be limited to an architecture that can be translated into the target analytical instrument. For example, the neural network may consist of an ingress layer (with a number of nodes equal to the dimensionality of the untransformed data), four fully connected modified linear unit (re-lu) activation layers (30 nodes per layer), and an output layer (with two nodes, linear activation). Various error functions (e.g., mean square error, Kullback-Leibler divergence) and optimization programs (e.g., adapelta, RMSProp) may be used for training.

After training, the neural network approximates the transformation that was originally used to transform the small training data set. Thus, the neural network represents the transformation between the space occupied by the original, untransformed data and the space occupied by the transformed data. Thus, the neural network may transform new observations by feeding new data to the entry layer and collecting the resulting transformed data in the output layer.

The generated event data may then be transmitted to transformation device 520 for processing. The transformation of the training data may comprise a non-parametric transformation. The non-parametric transformation may be identified in a request message that is accompanied or associated with the generated event data. The generated event data may be processed as described above via the event data receiver 522, the transform selection circuit 540, and the event data transform circuit 542 to generate transformed generated event data.

The transformed generated event data and the generated event data may then be received from the training circuit 530 to generate a transformed model. Generating the transformation model may include generating a neural network that approximates a transformation from the generated event data to the transformed generated event data. As discussed above, the configuration of the model may be constrained based on factors such as the target particle analyzer or the sorting electronics. This may facilitate the generation of a transformation model that may be applied to the target particle analyzer or sorting electronics. The particle analyzer or sorting electronics may be indicated using a hardware identifier, which may be associated with a model architecture or a constraint.

The transformation model 560 may be stored in the data store 514 and used for subsequent transformations of the event data for the experiment or other experiments.

FIG. 6 is a process flow diagram depicting an example of a parametric machine learning transformation method suitable for multi-dimensional event data. Method 600 may be implemented in whole or in part by a transformation device, such as transformation device 520 shown in fig. 5.

Multidimensional event data 602, such as particle analyzer data, may be provided. The multi-dimensional event data 602 may be processed using a non-parametric transform 604 (e.g., tSNE) to generate transformed event data 606 representing initial transformed measurements of the event. The transformed event data 606 may be used to train the model through machine learning 608. The result of the learning may include a parametric transformation neural network model 610. The neural network model 610 may generate transformed data 614 for the multi-dimensional event data 602 or new multi-dimensional event data 612. The transformed data 614 may then be provided for other processing or control, such as gate selection 680, event sort determination 682, or instrument adjustment 684 (e.g., pressure adjustment of a flow cytometer). Gate selection 680 can include transmitting the transformed event data to a device to present one or more graphical user interfaces. The graphical user interface may include control elements for selecting a destination event and assigning a sort destination (e.g., container, plate, well, chamber, tube, etc.) to the event. The selecting may include receiving gate information identifying a range of measurement values for classifying the particle. The gates may be specified using measurements in transformed (e.g., nonparametric) space. The gate information may form or be used to generate sorting criteria to configure the particle analyzer to sort particles suitable for the experiment.

The event sort determination 682 may include transmitting the transformation along with the sorting criteria to the particle analyzer. The particle analyzer (including sorting electronics or other analysis means for processing particles during the experiment) may be configured to transform raw event data and apply the specified sorting criteria. As discussed above, the transformation may be a parameterized neural network model representing a non-parametric transformation that is used to expose events suitable for gate selection. Instrument adjustment 684 may include transmitting control signals to the particle analyzer or an analytical device included therein to adjust its operational state. For example, the flow rate may be adjusted based on the distribution of the transformed event data as compared to an ideal distribution. If the distribution deviates from the ideal distribution, the flow rate may decrease. The amount of reduction can be estimated using the magnitude of the deviation from the ideal distribution. Other operating characteristics of the particle analyzer may be adjusted using the transformed event data (e.g., detector voltage or data scaling factor).

Fig. 7 is a process flow diagram depicting an example of a method for sample sorting using a parametric machine learning transformation. Method 700 may be implemented in whole or in part by the apparatus shown in fig. 1. Aspects of method 700 may be coordinated using a management device (e.g., transformation device 520 shown in fig. 5). The method 700 illustrates how event data is processed to generate a parametric transformation based on a non-parametric transformation of an initial set of events.

The method 700 begins at block 702. At block 710, initial event data is collected for a portion of a sample. The collecting may include activating a particle analyzer to process the portion of the sample. Processing the sample may include measuring a property of the particles, such as a graphical, electrical, temporal or acoustic property.

At block 720, the initial event data may be transmitted to a transformation device, such as transformation device 520 shown in FIG. 5. The transmitting may include transmitting a message requesting transformation of the initial event data. The message may include an identifier suitable for the experiment, sample, source of the sample, or other unique information to identify the event data. The identifier may be used to select a transformation applicable to the event data. The identifier may be used to associate a parameter transformation generated using the event data with a subsequent transformation request.

At block 730, the initial event data is transformed using a non-parametric transform. The transformation may be selected from a set of transformations based on, for example, event data. For example, if the event data includes acoustic measurements, a tSNE transform tailored to the acoustic values may be used. In some embodiments, the transformation may be specified in a message requesting processing of the event data. For example, if a researcher requests to provide a chart of the event data for gating, the request may include an indication of the type of transformation to be applied to graphically represent the dimensions of the event data for gating.

At block 740, the initial event data may be augmented to generate model training event data. The amplifying may include generating training data based on the measurements of the events. For example, the measurement may be used to generate a second synthetic event within a threshold distance of the measurement. The composite measurement may comprise a measurement value pair, where one value is a composite untransformed measurement value and the second value is a composite transformed measurement value. Table 1 lists event data applicable to measurements using a threshold of 0.001 (e.g., the composite measurement based on actual measurements should be within 0.001 of the actual measurements).

TABLE 1

Event Id	Is it synthesized?	Measuring
			1	Whether or not	2.30495
2	Is that	2.30492
			3	Is that	2.30499
4	Is that	2.30404

The threshold value may be determined based on the type of measurement value. For example, optical measurements (e.g., fluorescence-based values) may be more tolerable than image-based measurements. The threshold may be specified in a configuration accessible to the device performing the training. Aspects of the generation of the transformation data are included in fig. 5 and 6 described above.

At block 750, the parametric model is generated based at least in part on the model training event data. Generating the parametric model may include training a neural network model that approximates a non-parametric transformation from actual and synthetic event data to transformed event data. Generating the parametric model may include generating a neural network that approximates a transformation from the generated event data to the transformed generated event data. As discussed above, the configuration of the model may be constrained based on factors such as the target particle analyzer or the sorting electronics. This may facilitate the generation of a transformation model that may be applied to the target particle analyzer or sorting electronics.

At block 760, a selection of a destination event may be received. For example, the researcher may draw polygons on a chart to define a range of transformed data values to be sorted. The polygon may define a gate that may be or be associated with a sorting criterion. The selection may be made using initial event data transformed by a non-parametric transform or by a parametric model.

At block 770, the sorting electronics (e.g., sorting circuitry) may be configured using the parametric model and gating criteria specified at block 760. The configuration of the sorting electronics may include storing the model generated at block 750 in a storage location accessible to the sorting electronics. The model may then be used to transform the received event data for evaluation against sorting criteria, such as the population identified at block 760. As discussed above, the model may be used to configure a field programmable gate array included in the sorting electronics.

At block 780, the analyzer may evaluate and sort the remainder of the sample using the configured sorting electronics. As new event measurements are collected, the measurements can be transformed in real time using configured sorting electronics and sorted into designated containers. For example, a deflector plate of the particle analyzer may be activated to direct particles of interest into a designated collection tube.

The method 700 ends at block 790. However, it should be understood that method 700 may be repeated for other events, samples, or experiments. In some embodiments, it may be desirable to retrain the model to accommodate any variations in the sample or to account for variations in the source of the sample. For example, in a therapeutic setting, a biological sample may be collected during administration of a drug or other compound. The model may need to be adjusted to take into account the presence of the drug or compound after administration. In such cases, the model may be retrained and data collected, given that the original model has already been trained. This may provide for higher resource utilization than training a new model using a corpus of event data collected from the samples.

The function may sort the particles based on their membership to a target cluster. Mahalanobis distance can be used to measure the distance between a particle and a known cluster. A sorting decision for a particle may be determined based on the closest cluster of the particle. For some samples, particles near the contamination cluster may be excluded or transferred to a spare collection container. The criterion for determining closeness may be determined based on a minimum error probability. The probabilities can be defined by configuration, or specified as part of a sample experimental setup. Known clusters can be identified by a training set. The training set may comprise a subset of the measured values of the particles belonging to a certain cluster. The mean and covariance of the training set may be generated as a metric representative of a certain cluster. The metric can then be used as a factor in the mahalanobis distance equation, which can be evaluated with hardware that can analyze the measurements and, in time, generate sorting decisions that direct the particles to a particular collection vessel.

The term "determining" as used herein encompasses various activities. For example, "determining" can include calculating, computing, processing, deriving, investigating, looking up (e.g., looking up in a table, a database or another data structure), ascertaining and the like. Further, "determining" can include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), and the like. Further, "determining" may include resolving, selecting, establishing, and the like.

The term "providing" as used herein encompasses various activities. For example, "providing" may include storing the value in a location of the storage device that facilitates subsequent retrieval, transmitting the value directly to the recipient over at least one wired or wireless communication medium, transmitting or storing a reference to the value, and so forth. "providing" may also include encoding, decoding, encrypting, decrypting, validating, authenticating, etc., via hardware elements.

The terms "selectively" or "selectively" as used herein may encompass various activities. For example, a "selective" process may include determining one of a plurality of options. The "selective" process may include one or more of the following: dynamically determined input, pre-configured input, or input initiated by a user for determination. In some embodiments, n-input switching may be included to provide selective functionality, where n is the number of inputs used to make the selection.

The term "message" as used herein encompasses various formats for conveying (e.g., sending or receiving) information. The messages may include machine-readable aggregations of information, such as XML documents, fixed field messages, comma separated messages, and the like. In some embodiments, a message may include a signal for transmitting one or more representations of information. Although recited in the singular, it should be understood that a message can be composed, sent, stored, received, etc., in multiple parts.

As used herein, a "user interface" (also referred to as an interactive user interface, graphical user interface, or UI) may refer to a network-based interactive control that includes data fields, buttons, or other interactive controls for receiving input signals or providing electronic information or providing information to a user in response to any received input signals. Can be implemented in a form such as hypertext markup language (HTML), Javascript^TM、FLASH^TM、JAVA^TM,.NET^TM、WINDOWS OS^TM、macOS^TMWeb services, and Rich Site Summary (RSS) to implement the UI in whole or in part. In some embodiments, the UI may be included in a standalone client (e.g., a thick client) configured to communicate (e.g., send or receive data) according to one or more aspects described.

As used herein, a "data store" may be embodied in a hard drive, solid state memory, and/or any other type of non-transitory computer readable storage medium that may be accessed by or by a device, including an access device, server, or other computing device as described, and the like. As is known in the art, a data store may also or alternatively be distributed or partitioned among multiple local and/or remote storage devices without departing from the scope of the present invention. In other embodiments, the data store may be included or embodied in a data storage network service.

As used herein, a phrase referring to "at least one item" in a list of items refers to any combination of the items, including a single member. For example, "at least one of the following items: a. b or c "is intended to encompass: a. b, c, a-b, a-c, b-c and a-b-c.

Those of skill in the art would understand that information, messages, and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The techniques described herein may be implemented in hardware, software, firmware, or any combination thereof. The techniques may be implemented in any of a variety of devices, such as a specially programmed event processing computer, a wireless communication device, or an integrated circuit device. Any functions described as modules or components may be performed together in an integrated logic device or separately in discrete but interoperable logic devices. If implemented in software, the techniques may be realized at least in part by a computer-readable data storage medium comprising instructions that, when executed, perform one or more of the methods described above. The computer-readable data storage medium may form part of a computer program product, which may include packaging materials. The computer-readable medium may comprise memory or a data storage medium, such as Random Access Memory (RAM) (e.g., Synchronous Dynamic Random Access Memory (SDRAM)), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), flash memory, a magnetic or optical data storage medium, and so forth. The computer readable medium may be a non-transitory storage medium. Additionally or alternatively, the techniques may be realized at least in part by a computer readable communication medium that carries or communicates program code in the form of instructions or data structures and that can be accessed, read, and/or executed by a computing device, such as a propagated signal or wave.

The program code may be executed by a specially programmed transform processor, which may include one or more processors, such as one or more Digital Signal Processors (DSPs), configurable microprocessors, Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. The graphics processor may be specially configured to perform any of the techniques described in this disclosure. A combination of computing devices (e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration, at least in part, of a data connection) may perform one or more of the functions described. In some aspects, the functionality described herein may be provided within a dedicated software module or hardware module configured for encoding and decoding, or incorporated into a dedicated transform control card.

The methods disclosed herein comprise one or more steps or actions for achieving the method. The method steps and/or actions may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and/or use of specific steps and/or actions may be modified without departing from the scope of the claims.

Various embodiments of the invention are described herein. These and other embodiments are within the scope of the following claims.

Aspects of the above-described subject matter, including embodiments, can be used alone or in combination with one or more aspects or embodiments. Without limiting what is described herein, the following provides some non-limiting aspects of the invention. After reading this disclosure, it will be apparent to those skilled in the art that each of the individually numbered items can be used individually, or in combination with any of the previously or subsequently individually numbered items. This is intended to provide support for all such combinations of items of content and is not limited to the following explicitly provided combinations of content:

1. a method, comprising:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

2. The method of 1, wherein the non-parametric transformation comprises t-distributed random neighbor embedding (tSNE).

3. The method of any of claims 1-2, wherein the first raw data set comprises data from 10,000 or fewer detected cells of the flow cytometer.

4. The method of any of claims 1-3, wherein generating the second transformed data set comprises adding a noise component.

5. The method of 4, wherein generating the second transformed data set comprises adding the noise component to one or more of the first original data set, the first transformed data set, and combinations thereof.

6. The method of 5, wherein generating the second transformed data set comprises adding the noise component to a data pair comprising an original data point and a transformed data point.

7. The method of 6, wherein the original data point and the transformed data point are selected by random selection.

8. The method of 7, wherein the original data point and the transformed data point are selected by weighted random selection.

9. The method of 8, wherein the random selection is weighted by one or more of density and cell population.

10. The method of any of claims 4-9, wherein the noise component is determined based on a probability distribution.

11. The method of 10, wherein the probability distribution is selected from the group consisting of a uniform distribution, a gaussian distribution, or a poisson distribution.

12. The method of any of claims 1-11, wherein the transformation model is a dynamic algorithm.

13. The method of 12, wherein the dynamic algorithm is a machine learning algorithm.

14. The method of any of claims 1-13, wherein the transformation model is applied to real-time data from the flow cytometer.

15. The method of any one of claims 1-14, further comprising classifying particles in the sample within the fluidic medium of the flow cytometer.

16. The method of 15, further comprising generating a sorting decision based on the particle classification.

17. The method of any one of claims 15-16, further comprising sorting one or more particles in the sample.

18. The method of any of claims 1-17, wherein the method is performed by an integrated circuit device.

19. The method of 18, wherein the integrated circuit device is a Field Programmable Gate Array (FPGA).

20. The method of 18, wherein the integrated circuit device is an Application Specific Integrated Circuit (ASIC).

21. The method of 18, wherein the integrated circuit device is a Complex Programmable Logic Device (CPLD).

22. A computer-implemented method, comprising:

is controlled by one or more processing devices,

receiving from a particle analyzer measurements of a first portion of particles associated with an experiment;

converting the measured value into an initial transformed measured value by using nonparametric transformation;

generating a parametric model to receive as input the measured values of the first portion of particles relevant to the experiment and to generate as output the initial transformed measured values; and

configuring the particle analyzer to:

transforming measurements of particles included in a second portion of the particles based at least in part on the parametric model; and

classifying the particle based at least in part on the measurement and a sorting criterion.

23. The computer-implemented method of 22, further comprising:

receiving gate information identifying a range of measurements in a nonparametric space for classifying the particle, wherein the sorting criterion includes the gate information.

24. The computer-implemented method of any of claims 22-23, further comprising:

transmitting the parametric model to the particle analyzer; and

causing the particle analyzer to configure a sorting circuit based at least in part on the parametric model and the sorting criterion.

25. The computer-implemented method of claim 24, wherein the sorting circuit includes a field programmable gate array.

26. The computer-implemented method of any of claims 22-25, wherein the measurements received from the particle analyzer comprise measurements of light emitted by the first portion of particles in fluorescent form.

27. The computer-implemented method of 26, wherein the light emitted in fluorescent form by the first portion of particles comprises light emitted in fluorescent form by antibodies bound to the first portion of particles.

28. The computer-implemented method of any of claims 22-27, further comprising:

selecting a measurement value of the particle and an initial post-transform measurement value; and

generating a composite measurement value pair based at least in part on:

(i) a measurement of the value of the particle,

(ii) an initial post-transform measurement of the particle; and

(iii) noise values, wherein the integrated measure-value pairs comprise an integration

A measured value and a synthetically transformed measured value; and

wherein the parametric model further receives the synthetic measurements as input and generates the synthetic transformed measurements as output.

29. The computer-implemented method of 28, further comprising:

receiving gate information identifying a range of measurements for classifying the particles, an

Wherein the particle is selected based at least in part on the measurement value corresponding to the measurement value range.

30. The computer-implemented method of any of claims 22-29, further comprising:

receiving a hardware identifier from the particle analyzer, the hardware identifier indicating sorting circuitry implemented in the particle analyzer;

determining an architecture of the parametric model based at least in part on the hardware identifier, wherein the architecture of the parametric model is applicable to the sorting circuit; and

wherein generating the parametric model comprises generating the parametric model consistent with the architecture.

31. The computer-implemented method of claim 30, wherein the parametric model includes a neural network, and wherein the architecture identifies a number of layers of the neural network and a number of nodes per layer.

32. The computer-implemented method of any of claims 22-31, wherein the parametric model comprises a neural network, and wherein generating the parametric model comprises:

assigning an initial weight to a connection between a first node in a first layer of the parametric model and a second node in a second layer of the parametric model of a second layer;

generating an output suitable for the particle measurement;

comparing the output to the transformed measurements of the particle; and

adjusting the initial weights based at least in part on a difference between the output and the post-transform measurement values,

wherein a subsequent difference value of the particle is less than the difference value after adjusting the initial weight.

33. A system comprising a processor comprising a memory operably coupled to the processor, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

34. The system of 33, wherein the non-parametric transformation comprises t-distributed random neighbor embedding (tSNE).

35. The system of any of claims 33-34, wherein the first raw data set comprises data from 10,000 or fewer detected cells of the flow cytometer.

36. The system of any of claims 33-35, wherein the processor comprises a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component.

37. The system of 36, wherein the processor comprises a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component to one or more of the first original data set, the first transformed data set, and combinations thereof.

38. The system of 37, wherein the processor comprises a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component to a data pair comprising an original data point and a transformed data point.

39. The system of 38, wherein the original data point and the transformed data point are selected by random selection.

40. The system of 39, wherein the original data point and the transformed data point are selected by weighted random selection.

41. The system of 40, wherein the random selection is weighted by one or more of density and cell population.

42. The system of any of claims 36-41, wherein the noise component is determined based on a probability distribution.

43. The system of 42, wherein the probability distribution is selected from the group consisting of a uniform distribution, a Gaussian distribution, or a Poisson distribution.

44. The system of any of claims 33-43, wherein the transformation model is a dynamic algorithm.

45. The system of 44, wherein the dynamic algorithm is a machine learning algorithm.

46. The system of any of claims 33-45, wherein the transformation model is applied to real-time data from the flow cytometer.

47. The system of any one of claims 33-46, wherein the processor comprises a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to classify a particle in a sample within a fluidic medium of the flow cytometer.

48. The system of 47, wherein the processor comprises a memory operably coupled with itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to generate a sorting decision based on the particle classification.

49. The system of any of claims 33-48, further comprising an integrated circuit device.

50. The system of 49, wherein the integrated circuit device is a Field Programmable Gate Array (FPGA).

51. The system of 49, wherein the integrated circuit device is an Application Specific Integrated Circuit (ASIC).

52. The system of 49, wherein said integrated circuit device is a Complex Programmable Logic Device (CPLD).

53. A system, comprising:

one or more processing devices; and

a computer-readable storage medium containing instructions that, when executed by the one or more processing devices, cause the system to:

receiving from a particle analyzer measurements of a first portion of particles associated with an experiment;

converting the measured value into an initial transformed measured value by using nonparametric transformation;

generating a parametric model to receive as input raw measurements of the first fraction of particles relevant to the experiment and to generate as output the initial transformed measurements;

configuring the particle analyzer to:

transforming measurements of particles included in a second portion of the particles based at least in part on the parametric model;

generating a control signal to adjust an operational state of an analysis device included in the particle analyzer; and

transmitting the control signal to the analysis device to achieve the operational state.

54. The system of 53, wherein the analysis device comprises sorting electronics communicatively coupled with a deflection plate, an

Wherein the particle analyzer is further configured to identify a target container applicable to a particle based at least in part on the post-conversion measurement value corresponding to a sorting criterion associated with the target container, and

wherein adjusting the operating condition comprises applying an electrical charge through the deflection plate to direct the particles into a target vessel.

55. The system of 53, wherein the analysis device comprises a fluidic system, an

Wherein adjusting the operating condition comprises adjusting a pressure applied during the experiment.

56. A system, comprising:

a light source configured to irradiate a sample (containing particles) in a fluid medium;

a light detection system comprising a photodetector configured to generate a data signal using light detected from particles within the fluid medium; and

a processor comprising a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to:

applying a non-parametric transform to a first set of data signals from the light detection system to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second set of data signals from the light detection system.

57. The system of 56, wherein the non-parametric transformation comprises t-distributed random neighbor embedding (tSNE).

58. The system of any one of 56-57, wherein the first set of data signals from the light detection system comprises data from 10,000 or fewer detected cells of the flow cytometer.

59. The system of any of 56-58, wherein the processor comprises a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component.

60. The system of 59, wherein the processor comprises a memory operably coupled to itself, wherein the memory includes instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component to one or more of the first set of data signals from the light detection system, the first transformed data set, and combinations thereof.

61. The system of 60, wherein the processor comprises a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to generate the second transformed data set by adding a noise component to a data pair comprising an original data point and a transformed data point.

62. The system of 61, wherein the original data point and the transformed data point are selected by random selection.

63. The system of 62, wherein the original data point and the transformed data point are selected by weighted random selection.

64. The system of 63, wherein the random selection is weighted by one or more of density and cell population.

65. The system of any of 59-64, wherein the noise component is determined based on a probability distribution.

66. The system of 65, wherein the probability distribution is selected from the group consisting of a uniform distribution, a Gaussian distribution, or a Poisson distribution.

67. The system of any of 56-66, wherein the transformation model is a dynamic algorithm.

68. The system of 67, wherein the dynamic algorithm is a machine learning algorithm.

69. The system according to any one of claims 56-68, wherein the transformation model is applied to real-time data signals from the light detection system.

70. The system of any one of claims 56-69, wherein the processor comprises a memory operably coupled to itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to classify a particle in a sample within the fluid medium.

71. The system of 70, wherein the processor comprises a memory operably coupled with itself, wherein the memory comprises instructions stored thereon that, when executed by the processor, cause the processor to generate a sorting decision based on the particle classification.

72. The system of any one of 56-71, further comprising a cell sorter.

73. The system of 72, wherein the cell sorter comprises a droplet deflector.

74. The system of any of claims 56-73, further comprising an integrated circuit device.

75. The system of 74, wherein the integrated circuit device is a Field Programmable Gate Array (FPGA).

76. The system of 74, wherein the integrated circuit device is an Application Specific Integrated Circuit (ASIC).

77. The system of 74, wherein the integrated circuit device is a Complex Programmable Logic Device (CPLD).

78. An integrated circuit programmed to:

applying a non-parametric transform to a first raw data set from a flow cytometer to generate a first transformed data set;

generating a second transformed data set using the first transformed data set; and

generating a transformation model using the second transformed data set; and

applying the transformation model to a second raw data set from the flow cytometer.

79. The integrated circuit of 78, wherein the nonparametric transformation includes t-distributed random neighbor embedding (tSNE).

80. The integrated circuit of any of 78-79, wherein the first raw data set comprises data from 10,000 or fewer detected cells of the flow cytometer.

81. The integrated circuit of any of 78-80, wherein the integrated circuit is programmed to generate the second transformed data set by adding a noise component.

82. The integrated circuit of 78, wherein the integrated circuit is programmed to generate the second transformed data set by adding a noise component to one or more of the first original data set, the first transformed data set, and a combination thereof.

83. The integrated circuit of 82, wherein the integrated circuit is programmed to generate the second transformed data set by adding a noise component to a data pair (comprising an original data point and a transformed data point).

84. The integrated circuit of 83, wherein the original data point and the transformed data point are selected by random selection.

85. The integrated circuit of 84, wherein the original data point and the transformed data point are selected by weighted random selection.

86. The integrated circuit of 85, wherein the random selection is weighted by one or more of density and cell population.

87. The integrated circuit of any of 81-86, wherein the noise component is determined based on a probability distribution.

88. The integrated circuit of 87, wherein the probability distribution is selected from the group consisting of a uniform distribution, a gaussian distribution, or a poisson distribution.

89. The integrated circuit of any of 78-88, wherein the transformation model is a dynamic algorithm.

90. The integrated circuit of 89, wherein the dynamic algorithm is a machine learning algorithm.

91. The integrated circuit of any of 78-90, wherein the transformation model is applied to real-time data from the flow cytometer.

92. The integrated circuit of any of 78-91, wherein the integrated circuit is programmed to classify a particle in a sample within a fluidic medium of the flow cytometer.

93. The integrated circuit of 92, wherein the integrated circuit is programmed to generate a sorting decision based on the particle classification.

94. The integrated circuit of any of claims 78-93, wherein the integrated circuit device is a Field Programmable Gate Array (FPGA).

96. The integrated circuit of any of claims 78-93, wherein the integrated circuit device is an Application Specific Integrated Circuit (ASIC).

97. The integrated circuit of any of claims 78-93, wherein the integrated circuit device is a Complex Programmable Logic Device (CPLD).

Although the foregoing invention has been described in some detail by way of illustration and example for purposes of clarity of understanding, it will be readily apparent to those of ordinary skill in the art in light of the teachings of this invention that certain changes and modifications may be made thereto without departing from the spirit or scope of the appended claims.

Accordingly, the foregoing merely illustrates the principles of the invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described or shown herein, embody the principles of the invention and are thus within its spirit and scope. Furthermore, all examples and conditional language recited herein are principally intended expressly to aid the reader in understanding the principles of the invention and the concepts contributed by the inventors to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure. Furthermore, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.

Accordingly, the scope of the present invention is not limited to the exemplary embodiments shown and described herein. Rather, the scope and spirit of the invention is embodied by the appended claims. In the claims, section 112 (f) of chapter 35 of the united states codex or section 112 (6) of chapter 35 of the united states codex are explicitly defined as being cited only if the phrase "means for … …" or "step for … …" is explicitly used at the beginning of the limitations of the claims; if the phrase is not used in the limitations of the claims, chapter 35, section 112 (f) of the United states code or chapter 35, section 112 (6) of the United states code is not cited.