阅读说明：本技术 基于人工智能和动物实验数据集的药物代谢动力预测模型 (Drug metabolism dynamic prediction model based on artificial intelligence and animal experiment data set ) 是由 王文乐 陈金拳 朱玉洁 崔亮 于 2020-04-02 设计创作，主要内容包括：基于人工智能和动物实验数据集的药物代谢动力预测模型,它涉及一种药物代谢预测模型的建立,具体涉及一种基于人工智能和动物实验数集的药物代谢动力预测模型。它采用以下技术步骤为：1、利用计算机浅层模拟算法建立人用药物在动物体内的相关活性数据库；2、利用深度学习的方法基于大数据和人工智能建立药物代谢动力学模型；在构建药物代谢动力学模型建立中,比较紧迫的一个问题是如何尽可能模拟每一个特定的个体；药物具有所谓的狭窄治疗指数,即使药物浓度的微小波动也可能产生治疗对象的不希望的意外的结果；3、通过深度学校建立药物活性预测模型；4、将预测模型优化后用于药物设计活性预测和个体化治疗。它全面而系统的建立人体用药在动物体内实验的相关活性数据库；用深度学习的方法构建合适的药代动力学模型,用于个体化治疗以及新药设计的辅助预测。(A drug metabolism dynamic prediction model based on artificial intelligence and an animal experiment data set relates to establishment of a drug metabolism prediction model, in particular to a drug metabolism dynamic prediction model based on artificial intelligence and an animal experiment data set. The method comprises the following technical steps: 1. establishing a related activity database of human medicine in the animal body by utilizing a computer shallow layer simulation algorithm; 2. establishing a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method; in building pharmacokinetic models, one of the pressing questions is how to model each particular individual as much as possible; drugs have a so-called narrow therapeutic index, even small fluctuations in drug concentration can produce undesirable unexpected results in a subject; 3. establishing a drug activity prediction model through deep school; 4. the prediction model is optimized and then used for drug design activity prediction and individualized treatment. It comprehensively and systematically establishes a relevant activity database of the human medicine in the animal body experiment; and (3) constructing a proper pharmacokinetic model by using a deep learning method for individualized treatment and auxiliary prediction of new drug design.)

1. Drug metabolism dynamic prediction model based on artificial intelligence and animal experiment data set, its characterized in that: the method comprises the following technical steps: 1. establishing a related activity database of human medicine in the animal body by utilizing a computer shallow layer simulation algorithm; 2. establishing a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method; in building pharmacokinetic models, one of the pressing questions is how to model each particular individual as much as possible; drugs have a so-called narrow therapeutic index, even small fluctuations in drug concentration can produce undesirable unexpected results in a subject; 3. establishing a drug activity prediction model through deep school; 4. the prediction model is optimized and then used for drug design activity prediction and individualized treatment.

2. The model of claim 1, wherein the model is based on artificial intelligence and animal experimental data set, and is characterized by: the step 1 utilizes a computer shallow layer simulation algorithm to establish a related activity database of human drugs in the animal body: the activity database of the experiment of human drugs in the animal body is established, and not only the intravenous injection drugs are included, but all data including various administration modes such as oral administration, intravenous injection, subcutaneous injection, intragastric administration and the like are collected.

3. The model of claim 1, wherein the model is based on artificial intelligence and animal experimental data set, and is characterized by: and 2, establishing a pharmacokinetic model by utilizing a deep learning algorithm based on big data and artificial intelligence: the deep learning fully refers to the neural network layered structure of the human brain nervous system, and the essence of the deep learning is that the nonlinear combination of feature extraction is carried out on information in each layer of data through a multilayer structure, and each layer of features is converted into higher-layer abstract representation layer by layer from original data, so that the complex structural features of high-dimensional data are found.

4. The model of claim 3, wherein the model is based on artificial intelligence and animal experimental data set, and is characterized by: the deep learning algorithm is mostly based on a deep neural network at present, namely a multilayer neural network, a series of data characteristics are trained through each neural network, and when an actual output result is different from a target function by a certain amount, feedback is carried out and retraining is carried out until the result is converged.

Technical Field

The invention relates to establishment of a drug metabolism prediction model, in particular to a drug metabolism dynamic prediction model based on artificial intelligence and an animal experiment number set.

Background

With the rapid development of information technology, Artificial Intelligence (AI), one of the three most advanced technologies in the world, has been widely spread and developed since 1956, and has penetrated into various industries. In 4 months in 2018, the education department of China more actively deploys action plans to promote multi-subject cross fusion of AI, and plans to 2020, so that the overall AI technology and application are kept at the same level with the world advanced level. The AI is based on the relevant theory of human intelligence, and simulates and extends the edge discipline of the theory, method, technology and application system of human intelligence by applying methods such as big data, machine learning and the like; the branch fields of the method also comprise databases, data mining, statistics, knowledge discovery, pattern recognition, neural networks and the like, and the method is widely applied to the fields of language recognition, intelligent terminals, mobile commerce, medical health and the like at present. Among them, the medical health field currently has practical problems of insufficient high-quality resources, high medical cost, long doctor culture and drug development cycle, and the like, so the AI technology especially shows huge potential and considerable prospect in the field, and large scientific and technological heads such as google, IBM, and the like all strive for a layout AI medical market in the world.

The pharmaceutical field, which is the medical health field in which the AI technology is applied at the earliest time, has achieved wide application and development of the AI technology in many aspects such as health management, auxiliary diagnosis and treatment, drug discovery, drug blending, and even clinical rational administration. The AI technology can be used for perfecting health risk identification (such as a health optimization platform) of a patient, intelligent medication monitoring and adverse reaction risk assessment (such as a computer-assisted Bayesian adverse reaction diagnosis system), assisting clinical treatment drug detection, clinical medication consultation, rationalizing drug design (such as computer-assisted drug design), improving new drug research and development conversion efficiency, providing a new drug targeting means (such as Philips bee colony robot), and comprehensively analyzing various clinical information and drug economic data of the patient to form scientific and reasonable individualized prescriptions.

Today in the 21 st century, the pharmaceutical industry, which should be at the front of science, did not produce much glaring sparks. Statistically, drug research is the least effective segment of the global business chain, and essentially ninety-six percent of drug production will be declared a failure. Generally, whether a drug is ultimately affected by social acceptance is three major factors: firstly, whether the efficacy of the medicine is remarkable or not; secondly, whether the medicine has larger side effect or generates certain toxicity or not; thirdly, whether the drug production mode is consistent with the market economy. In order to reduce the elimination rate of clinical development of drugs as much as possible, valuable information such as the activity of drugs and their intrinsic composition structures must be explored as early as possible. Therefore, in the process of advancing drug design and development, it is of great significance to explore the activity of the drug through a relevant prediction mode.

The activity of the drug is mainly detected by detecting the biological effect of the drug on organisms, the early activity determination mainly depends on animal experiments, and the animal model is the mode which can truly and effectively reflect the clinical pharmacology effect of the drug. In most conventional compound activity studies, the pharmaceutical activity of a compound is measured by animal in vivo tests and detection methods, and a lot of time and cost are undoubtedly consumed in the environment of massive compound data. The activity of an unknown compound is predicted by modern compound activity research through establishing a quantitative structure-activity relationship model by using a mathematical method.

With the continuous development of computer data mining technology, machine learning becomes an active research method in the field of computer science, and scientists apply the machine learning method to improve the prediction efficiency of drug activity. The deep learning algorithm is a machine learning algorithm which is very suitable for big data analysis and has the processing capability of 'abstract concept'. By using the deep learning algorithm, various machine learning models established in the traditional medicine design and medicine information are expected to be improved, and the development of chemical informatics is promoted.

The artificial intelligence needs to have big data as raw materials, and the field of new drug research and development is really a treasure house with very rich big data, so that the artificial intelligence is provided with a great place for use. For example, at least 45 million compounds are published as research objects of medicines in journal of pharmaceutical chemistry in 1959 till now, which is a huge database, and for such big data, artificial intelligence can play its unique role. In the past, the scientific americans and the world economic forum released ten-year-old emerging technologies in 2018, and one of the technologies is artificial intelligence assisted chemical molecule design-machine learning algorithm accelerated new drug development.

At present, at least 100 enterprises in the world are exploring artificial intelligence methods for new drug research and development, and the Puerarin Shike, Merck, Qiangsheng and Sonofiflu companies have distributed artificial intelligence new drug research and development abroad. In China, artificial intelligent new medicine research and development enterprises such as deep intelligence, zero krypton technology, Jingtai technology and the like emerge, and medicine Mingkande also invests in an artificial intelligent new medicine research and development company in the United states strategically.

At present, the machine can be used for learning the combination characteristics of the drug and the drug target, so that the machine can carry out drug design, and the probability of successful design can be greatly improved. The emergence of the artificial intelligence technology provides certain possibility for China to realize curve overtaking in international competition of new drug research and development.

Disclosure of Invention

Aiming at the defects and shortcomings of the prior art, the invention provides a drug metabolism dynamic prediction model based on artificial intelligence and an animal experiment data set, which comprehensively and systematically establishes a relevant activity database of the experiment of human medicine in an animal body; and (3) constructing a proper pharmacokinetic model by using a deep learning method for individualized treatment and auxiliary prediction of new drug design.

In order to realize the purpose, the invention adopts the following technical steps: 1. establishing a related activity database of human medicine in the animal body by utilizing a computer shallow layer simulation algorithm; 2. establishing a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method; one of the pressing questions in building pharmacokinetic modeling is how to model each particular individual as closely as possible. Drugs have a so-called narrow therapeutic index, even small fluctuations in drug concentration can produce undesirable unexpected results in a subject; 3. establishing a drug activity prediction model through deep school; 4. the prediction model is optimized and then used for drug design activity prediction and individualized treatment.

The step 1 utilizes a computer shallow layer simulation algorithm to establish a related activity database of human drugs in the animal body: the activity database of the experiment of human drugs in the animal body is established, and not only the intravenous injection drugs are included, but all data including various administration modes such as oral administration, intravenous injection, subcutaneous injection, intragastric administration and the like are collected.

The step 2 is to establish a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method: the deep learning fully refers to the neural network layered structure of the human brain nervous system, and the essence of the deep learning is that the nonlinear combination of feature extraction is carried out on information in each layer of data through a multilayer structure, and each layer of features is converted into higher-layer abstract representation layer by layer from original data, so that the complex structural features of high-dimensional data are found. The deep learning algorithm is mostly based on a deep neural network at present, the deep neural network can be understood as a multilayer neural network in a popular way, a series of data characteristics are trained through each neural network, and when an actual output result is different from a target function to a certain extent, feedback retraining is carried out until the result is converged. In addition, the big data endows the deep learning with great opportunity, the basic mode of the deep learning at present is also a deep learning mode driven by the big data, and the deep learning is the autonomous learning of the machine, namely the machine is a learning technology, wherein the comparison, optimization, accumulation, refinement, re-comparison, re-optimization, re-accumulation and re-refinement adopt a top-down supervision training method, and the bottom-up unsupervised learning is the artificial intelligent machine learning deep learning path.

With the explosive growth of biological and chemical knowledge, this form cannot meet the needs of people for the knowledge of biotechnological drugs. Data from different sources is increasing (images, models, structures and sequences), but most of the information is still stored in textbooks or print periodicals. The inability to access available data, such as drugs and drug receptors, via the internet, inadequate data and information gathering in different public databases reflects the "two solitary (two)" state of bioinformatics and chemical informatics. As a result, the wealth of existing electronic sequence and structure data has not been well linked to the vast amount of pharmaceutical or chemical knowledge that has accumulated over the past half century. Therefore, a database, which is a completely searchable internet drug resource, is created, and the structure and mechanism data of drug molecules are combined with various activity data after experiments. Not only has obvious educational value, but also can help researchers easily browse and explore the relationship between various experimental methods such as medicine structure, administration mode and the like and activity evaluation indexes such as half-life period, absorption, distribution, metabolism, excretion, bioavailability and the like, and is convenient for further analysis and research.

Currently, drug-related databases are established as exemplified by drug bank, which is the only "bioinformatics and chemical informatics" resource that combines detailed drug data (i.e., chemistry, pharmacology, and pharmacy) with comprehensive drug target information (i.e., sequence, structure, and pathway of action). Drug bank is supported by the health research institute of canada, the alberta innovation-health solutions and metabolomics innovation center (TMIC), which is the core of state-funded research and support for extensive top-end technology metabolomics research. The drug bank database query contains the following information: drug type, drug profile, chemical structure, drug composition, clinical trial, drug target, enzyme, transporter, carrier, drug picture, approved case, approved prescription drug, foreign market name, drug interaction, manufacturer, packager, etc. Drug bank can provide detailed, up-to-date, quantitative analysis or molecular weight information about drugs, drug targets and biological or physiological consequences of drug action. As a chemically-oriented drug database, drug bank can provide a number of built-in tools for viewing, sorting, searching, and extracting text, images, sequences, or structural data. Since the first time information is released from databases, drug banks have been widely used in computer drug retrieval, drug "reconstitution", computer drug structure data retrieval, drug docking or screening, drug metabolism prediction, drug target prediction, and general pharmaceutical education.

The database of pharmaceutical activity to be created by the present invention is similar to drug bank in structure and function, and only the data content and type are changed due to different research purposes, therefore, the tools and methods for creating the database can be the same (for example, PHP programming + MySQ L database software, etc.), and the database creation is feasible, but it takes a certain time to collect and arrange the data.

The second main research content of the invention is to establish a drug activity prediction model based on a deep learning method by means of the established large database, and finally to be used for computer-aided drug research and development or design. In recent years, with the progress of artificial intelligence, machine learning and deep learning, many domestic scholars have made certain progress in the research. In 2016 to 2017, there are 7 relevant documents which can be searched by taking deep learning and drug activity prediction as key words, and in the documents, research and analysis are basically carried out on the basis of data in a database relevant to some pathogenic gene research abroad. The quantity is small, on one hand, the attention degree of research projects carried out by domestic researchers on the aspect needs to be improved, and it is believed that related researches are more and more along with the development of artificial intelligence and computer information processing; on the other hand, the invention also fully shows that the invention is feasible to utilize a deep learning method to carry out prediction research on the activity of the medicine based on the database established by the invention.

The present invention is based on various public, searchable and trustworthy literature databases, such as the PUBMED (literature in the medical and life science) database of NCBI (national center for bioinformatics) and established drug databases, such as DRAGBANK, dragcenra L, etc., to collect experimental data of human drugs in animals, including relevant data on pharmacology, pharmacodynamics, pharmacokinetics, toxicology, etc.

The method utilizes a deep learning method to construct a proper pharmacokinetic model for individualized treatment and auxiliary prediction of new drug design. The method is characterized in that a big data driven deep learning mode is adopted, data in an established database are compared, optimized, accumulated, refined, re-compared, re-optimized, re-accumulated and re-refined, an artificial intelligent deep learning path is established by a top-down supervision training method or a bottom-up unsupervised learning method, and the pharmaceutical activity data are experimentally researched according to a mathematical analysis mode. And finally, the method is used for the auxiliary prediction of individualized treatment and new drug research and development design through an optimization model.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a block diagram illustrating the flow structure of the present invention.

Detailed Description

Referring to fig. 1, the technical solution adopted by the present embodiment is: 1. establishing a related activity database of human medicine in the animal body by utilizing a computer shallow layer simulation algorithm; 2. establishing a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method; one of the pressing questions in building pharmacokinetic modeling is how to model each particular individual as closely as possible. Drugs have a so-called narrow therapeutic index, even small fluctuations in drug concentration can produce undesirable unexpected results in a subject; 3. establishing a drug activity prediction model through deep school; 4. the prediction model is optimized and then used for drug design activity prediction and individualized treatment.

The step 1 utilizes a computer shallow layer simulation algorithm to establish a related activity database of human drugs in the animal body: the activity database of the experiment of human drugs in the animal body is established, and not only the intravenous injection drugs are included, but all data including various administration modes such as oral administration, intravenous injection, subcutaneous injection, intragastric administration and the like are collected.

The step 2 is to establish a pharmacokinetic model based on big data and artificial intelligence by utilizing a deep learning method: the deep learning fully refers to the neural network layered structure of the human brain nervous system, and the essence of the deep learning is that the nonlinear combination of feature extraction is carried out on information in each layer of data through a multilayer structure, and each layer of features is converted into higher-layer abstract representation layer by layer from original data, so that the complex structural features of high-dimensional data are found. The deep learning algorithm is mostly based on a deep neural network at present, the deep neural network can be understood as a multilayer neural network in a popular way, a series of data characteristics are trained through each neural network, and when an actual output result is different from a target function to a certain extent, feedback retraining is carried out until the result is converged. In addition, the big data endows the deep learning with great opportunity, the basic mode of the deep learning at present is also a deep learning mode driven by the big data, and the deep learning is the autonomous learning of the machine, namely the machine is a learning technology, wherein the comparison, optimization, accumulation, refinement, re-comparison, re-optimization, re-accumulation and re-refinement adopt a top-down supervision training method, and the bottom-up unsupervised learning is the artificial intelligent machine learning deep learning path.

With the explosive growth of biological and chemical knowledge, this form cannot meet the needs of people for the knowledge of biotechnological drugs. Data from different sources is increasing (images, models, structures and sequences), but most of the information is still stored in textbooks or print periodicals. The inability to access available data, such as drugs and drug receptors, via the internet, inadequate data and information gathering in different public databases reflects the "two solitary (two)" state of bioinformatics and chemical informatics. As a result, the wealth of existing electronic sequence and structure data has not been well linked to the vast amount of pharmaceutical or chemical knowledge that has accumulated over the past half century. Therefore, a database, which is a completely searchable internet drug resource, is created, and the structure and mechanism data of drug molecules are combined with various activity data after experiments. Not only has obvious educational value, but also can help researchers easily browse and explore the relationship between various experimental methods such as medicine structure, administration mode and the like and activity evaluation indexes such as half-life period, absorption, distribution, metabolism, excretion, bioavailability and the like, and is convenient for further analysis and research.

Currently, drug-related databases are established as exemplified by drug bank, which is the only "bioinformatics and chemical informatics" resource that combines detailed drug data (i.e., chemistry, pharmacology, and pharmacy) with comprehensive drug target information (i.e., sequence, structure, and pathway of action). Drug bank is supported by the health research institute of canada, the alberta innovation-health solutions and metabolomics innovation center (TMIC), which is the core of state-funded research and support for extensive top-end technology metabolomics research. The drug bank database query contains the following information: drug type, drug profile, chemical structure, drug composition, clinical trial, drug target, enzyme, transporter, carrier, drug picture, approved case, approved prescription drug, foreign market name, drug interaction, manufacturer, packager, etc. Drug bank can provide detailed, up-to-date, quantitative analysis or molecular weight information about drugs, drug targets and biological or physiological consequences of drug action. As a chemically-oriented drug database, drug bank can provide a number of built-in tools for viewing, sorting, searching, and extracting text, images, sequences, or structural data. Since the first time information is released from databases, drug banks have been widely used in computer drug retrieval, drug "reconstitution", computer drug structure data retrieval, drug docking or screening, drug metabolism prediction, drug target prediction, and general pharmaceutical education.

The method utilizes a deep learning method to construct a proper pharmacokinetic model for individualized treatment and auxiliary prediction of new drug design. The method is characterized in that a big data driven deep learning mode is adopted, data in an established database are compared, optimized, accumulated, refined, re-compared, re-optimized, re-accumulated and re-refined, an artificial intelligent deep learning path is established by a top-down supervision training method or a bottom-up unsupervised learning method, and the pharmaceutical activity data are experimentally researched according to a mathematical analysis mode. And finally, the method is used for the auxiliary prediction of individualized treatment and new drug research and development design through an optimization model.

The above description is only for the purpose of illustrating the technical solutions of the present invention and not for the purpose of limiting the same, and other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention should be covered within the scope of the claims of the present invention without departing from the spirit and scope of the technical solutions of the present invention.