Polypeptide activity screening method and system based on artificial intelligence and entity library

Abstract

The invention discloses a polypeptide activity screening method and system based on artificial intelligence and an entity library. The method comprises the following steps: acquiring physicochemical properties and structural information of polypeptide molecules, physicochemical properties and structural information of target proteins and experimental data of combination of the target proteins and the polypeptide molecules from the entity library; processing experimental data of combination of the target protein and the polypeptide molecules; performing training verification on the initial model based on the physicochemical property and structure information of the polypeptide molecules, the physicochemical property and structure information of the target proteins and the processed experimental data to obtain a prediction model; and inputting the to-be-screened protein structure and the polypeptide sequence into the prediction model to carry out affinity prediction so as to complete screening of polypeptide activity. Through the scheme of the invention, a mode of combining artificial intelligence and entity library real big data is realized, the artificial intelligence model is trained by utilizing a large amount of real experimental data, the magnitude of compound screening is expanded through the intelligent model, the screening cost is rapidly reduced, and the experimental efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of computer-assisted drug discovery, and in particular to a method and system for screening polypeptide activity based on artificial intelligence and entity libraries. Background Art

[0002] High-throughput solid compound screening is one of the most effective innovative drug discovery approaches in the past, present and future. In the statistical analysis of previously marketed drugs, the vast majority of drugs come from solid compound screening. In the field of small molecules, international pharmaceutical giants, compound library companies, the National Center for New Drug Screening and other companies and institutions have built solid small molecule libraries ranging from hundreds of thousands to millions. However, as an important class of drug molecules, peptides have no solid molecule libraries available so far.

[0003] As high-throughput screening continues, the limitations of traditional physical, high-throughput screening have gradually become apparent, including high cost (consuming a large number of researchers, a large amount of scientific research reagents, materials, etc.), limited number of physical molecules (millions of levels), and low efficiency (screening of hundreds of thousands of different peptide compounds takes months, etc.).

[0004] Therefore, this application proposes a technical concept based on the inherent defects existing in such real experimental research and the large amount of real data generated in real experiments combined with the current rapidly developing artificial intelligence technology to improve the scale of screening and reduce screening costs. Summary of the invention

[0005] The purpose of this application is to provide a method and system for screening polypeptide activity based on artificial intelligence and physical library, to realize the combination of artificial intelligence + real big data of physical library, to build, train and optimize artificial intelligence model using a large amount of real experimental data, to rapidly expand the scale of screening compounds through artificial intelligence model, to rapidly reduce screening costs and greatly improve experimental efficiency.

[0006] To achieve the above objectives, this application provides the following solutions: In the first aspect, the present application provides a method for screening polypeptide activity based on artificial intelligence and entity library, comprising: Obtain the physicochemical properties and structural information of peptide molecules, the physicochemical properties and structural information of target proteins, and experimental data on the binding of target proteins to peptide molecules from the entity library; Processing the experimental data of the binding between the target protein and the polypeptide molecule; Based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data, the initial model is trained and verified to obtain a prediction model; The protein structure and polypeptide sequence to be screened are input into the prediction model for affinity prediction to complete the screening of polypeptide activity.

[0007] Optionally, the step of processing the target protein and polypeptide molecule experimental data includes: The structure of the target protein combined with the polypeptide molecule and the three-dimensional coordinates of the internal atoms are converted into feature data that can be recognized by the initial model to obtain a feature matrix; Combine and split the experimental data of target proteins and peptide molecules.

[0008] Optionally, the step of training and verifying the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the experimental data after combination and splitting to obtain the prediction model includes: Obtaining a training data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to train and verify the initial model; A test data set is obtained from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to perform prediction evaluation on the trained model to obtain a prediction model.

[0009] Optionally, the step of obtaining a test data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to perform predictive evaluation on the trained model to obtain a predictive model comprises: Acquire a test data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to predict the trained model; The prediction results are evaluated based on the preset evaluation indicators.

[0010] Optionally, the preset evaluation index includes a consistency index, which is calculated as follows:

[0011] in, Z is the normalization constant; b x For greater affinity d x The predicted value of b y For greater affinity d y The predicted value of CI is the consistency index, CI The higher the value, the smaller the difference between the sample prediction value and the true value; h(x) is a step function; in, h(x) The formula is as follows: .

[0012] Optionally, the preset evaluation index includes a root mean square error, which is calculated as follows:

[0013] in, n is the number of samples; MSE For the predicted value p i ( i =1, 2, …, n ) and the true value y i The average of the sum of squares of the differences between MSE The smaller it is, the closer the predicted value of the sample is to the true value.

[0014] Optionally, before the step of training and validating the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain the prediction model, the step includes: Modifying the algorithm of the initial model based on the physicochemical properties and structural information of the polypeptide molecule; The initial model is trained with a smaller number of iterations based on a grid search method to determine learning parameters of the initial model, wherein the learning parameters include: an initial learning rate and / or a batch size.

[0015] In the second aspect, the present application provides a polypeptide activity screening system based on artificial intelligence and entity library, comprising: A data acquisition module is used to obtain the physicochemical properties and structural information of peptide molecules, the physicochemical properties and structural information of target proteins, and the experimental data of target proteins and peptide molecules from the entity library; A data processing module, used to process the experimental data of the binding between the target protein and the polypeptide molecule; A model training module, used to train and verify the initial model according to the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain a prediction model; The model execution module is used to input the target protein information and peptide information into the prediction model to predict the affinity strength.

[0016] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-mentioned methods for screening polypeptide activity based on artificial intelligence and entity libraries.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-mentioned methods for screening polypeptide activity based on artificial intelligence and a physical library.

[0018] According to the specific embodiments provided in this application, this application discloses the following technical effects: The present application provides a method and system for screening polypeptide activity based on artificial intelligence and entity library, which utilizes the real big data generated by real screening experiments and a large amount of real data of entity library, combined with artificial intelligence algorithm model, to realize the combination of artificial intelligence + real big data of entity library, and utilizes a large amount of real experimental data to build, train and optimize artificial intelligence model, and predicts affinity of target protein and polypeptide in advance through artificial intelligence model to complete polypeptide activity screening, thereby rapidly expanding the magnitude of screening compounds, rapidly reducing screening costs and greatly improving experimental efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A schematic diagram of a method for screening polypeptide activity based on artificial intelligence and a physical library provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a polypeptide activity screening system based on artificial intelligence and a physical library provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0023] refer to Figure 1 As shown, the present application provides a method for screening polypeptide activity based on artificial intelligence and entity library, comprising: Obtaining the physicochemical properties and structural information of the peptide molecules, the physicochemical properties and structural information of the target protein, and the experimental data information of the binding between the target protein and the peptide molecules from the entity library; Processing the experimental data of the binding between the target protein and the polypeptide molecule; Based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data, the initial model is trained and verified to obtain a prediction model; The protein structure and polypeptide sequence to be screened are input into the prediction model for affinity prediction to complete the screening of polypeptide activity.

[0024] In the embodiments of the present application, artificial intelligence is first used to obtain the physicochemical properties of polypeptide molecules such as acidity and alkalinity, optical rotation, solubility, stability, isoelectric point, folding and inactivation, and structural information such as primary structure, secondary structure, tertiary structure and quaternary structure from a pre-constructed entity library; obtain the physicochemical properties of the target protein such as ultraviolet absorption, zwitterionic dissociation, colloidal properties, deformation and hydrolysis, and structural information such as primary structure, secondary structure, tertiary structure and quaternary structure of the target protein; obtain experimental data on the binding of polypeptide molecules to target proteins, such as specifically labeled ELISA affinity binding data, ELISA competition binding data, BLI binding data, SPR (surface plasmon resonance) binding data, cell model data, and possibly animal model data.

[0025] Furthermore, the experimental data of the binding of the target protein and the polypeptide molecule are processed. For example, the structure of the target protein-peptide binding molecule and the three-dimensional coordinates of its internal atoms are converted into feature data such as secondary structure, three-dimensional structure coordinates, physicochemical properties, topological structure information, etc. that can be recognized by the deep learning model to obtain a feature matrix. The experimental data of the target protein and polypeptide molecules (including protein level, cell level, animal level) with specific annotations, ELISA, affinity binding data, ELISA competitive binding data, BLI binding data, SPR binding data, cell model data and possible animal model data are combined and split to supervise the multi-dimensional output of the model. The label value of each dimension is the affinity strength calculated by the experimental data and various real experimental data.

[0026] Furthermore, the acquired physicochemical properties and structural information of the polypeptide molecules, the physicochemical properties and structural information of the target protein, and the experimental data of the binding of the processed target protein to the polypeptide molecules are input into an initial model based on the prediction of the binding strength of small molecule ligands and receptor proteins for training and verification. At the same time, the algorithm hyperparameters are adjusted according to the performance of the model. After the initial model training is completed, the test set data is used to obtain the final model performance.

[0027] Furthermore, after obtaining the final model, the target protein structure and the structure of the polypeptide molecule to be screened are input into the model to predict the affinity strength of the target protein and the polypeptide molecule, thereby completing the screening of polypeptide activity.

[0028] In an exemplary embodiment, the step of processing the target protein and polypeptide molecule experimental data includes: The structure of the target protein binding to the polypeptide molecule and the three-dimensional coordinates of the internal atoms are converted into feature data that can be recognized by the deep learning model to obtain a feature matrix; In the embodiments of the present application, since the numerical ranges of the three-dimensional coordinates of atoms may vary greatly, the coordinate data needs to be normalized first. Usually, the coordinate value can be subtracted from its mean and then divided by its standard deviation to make the coordinate data within a relatively small and standardized numerical range, such as mapping the coordinate values ​​of each dimension (x, y, z) to the interval [-1, 1] or [0, 1]. For example, for a set of x coordinate values ​​of target protein atoms, first calculate its mean as 100 and standard deviation as 20, then each x coordinate value is calculated by the formula ( x -100) / 20 for normalization, and the same applies to y and z coordinates. Normalization helps to improve the stability and convergence speed of model training, and avoid problems in model training caused by large differences in coordinate values. In this embodiment, further, according to specific research needs and model complexity, it may be necessary to remove some redundant atomic information. For example, in the binding structure of the target protein and polypeptide molecule, atoms that are far away from the binding site and have little effect on the interaction, or some atoms that are repeated in the structure and do not contribute much to the overall structural characteristics (such as repeated atoms in some highly symmetric structures), can be removed by setting a distance threshold or a method based on structural analysis. For example, atoms that are more than 10 nanometers away from the binding site are set as redundant atoms for elimination, and only the atomic coordinate information that has an important potential impact on the binding is retained to simplify subsequent data processing and model training.

[0029] After normalizing the coordinate data, feature information is extracted from it. First, the type of atom is encoded to convert it into a numerical form that can be processed by the deep learning model. A common method is to use one-hot encoding. For example, it is known that the types of atoms involved in the binding structure of the target protein and the polypeptide molecule are several common elements such as carbon (C), hydrogen (H), oxygen (O), and nitrogen (N). A one-hot encoding vector can be set for each element. For example, for the carbon element, it is encoded as [1, 0, 0, 0], the hydrogen element is encoded as [0, 1, 0, 0], the oxygen element is encoded as [0, 0, 1, 0], and the nitrogen element is encoded as [0, 0, 0, 1], etc. When encountering a specific atom, it is represented by the corresponding one-hot encoding vector, so that the non-numerical information such as the atom type is converted into a numerical feature that can be recognized by the model. Then, the interatomic distance feature is extracted, and the distance between atoms is calculated as an important structural feature. For each atom, its distance from other atoms within a certain range (such as other atoms within a radius of 5 angstroms with the atom as the center) can be calculated. These distance values ​​can be directly used as part of the feature data. For example, the distances between atom A and the 10 atoms around it are 2.5 angstroms, 3.0 angstroms, etc., and these distance values ​​are recorded in sequence to reflect the spatial relationship between atoms and help the model understand the geometric characteristics of the structure. Extract bond angle and dihedral angle features. The bond angle can be determined by analyzing the three atoms connected by chemical bonds, and the dihedral angle can be determined by four consecutive atoms. For example, in the main chain atoms of a polypeptide molecule, the bond angle formed by three adjacent atoms and the dihedral angle formed by four consecutive atoms are calculated, and these angle values ​​are used as features. These angle features can further enrich the spatial information of the structure, which is critical for understanding the conformation and interaction of molecules.

[0030] After the features are extracted, the feature data is organized, and the various feature data extracted above are organized and arranged in a certain order. Usually, the atom type code, interatomic distance features, bond angle and dihedral angle features, and local environment features (if any) corresponding to each atom can be arranged in a row in sequence, taking atoms as units. For example, for a target protein and polypeptide molecule binding structure with 100 atoms, the first row is where all the feature data corresponding to the first atom are placed, the second row is where the feature data corresponding to the second atom is placed, and so on, to form a 100-row matrix prototype. Among them, if the number of feature data for different atoms is inconsistent (for example, the number of atoms around some atoms is different, resulting in differences in the number of distance features, etc.), it may be necessary to perform padding operations to make the number of features in each row the same, thereby forming a regular matrix. For example, by using zero padding, the length of the feature vectors corresponding to all atoms is made consistent, and finally a feature matrix with a fixed dimension is formed, such as a N× M The matrix ( N represents the number of atoms, M Represents the length of the feature vector corresponding to each atom), this matrix can be used as the input feature data of the deep learning model for subsequent model training, analysis and other operations.

[0031] The experimental data of the target protein and the polypeptide molecule are combined and split, and output in multiple dimensions of the supervised model, wherein the label value based on the supervised model is the affinity strength of the target protein and the polypeptide molecule calculated through the experimental data and the corresponding real experimental data.

[0032] In the embodiments of the present application, when splitting the experimental data, key features are specifically selected from the experimental data, such as the amino acid sequence features of the polypeptide (including amino acid composition, specific amino acid motifs, etc.), the structural features of the target protein (secondary structure ratio, active site amino acid residues, etc.), binding affinity data (such as KD value), binding kinetic parameters (kon, koff), etc. For example, for the binding affinity data, its value can be extracted separately as a feature dimension.

[0033] In this embodiment, when combining experimental data, specifically, different types of features are combined into a new feature vector. For example, the One-Hot code of the amino acid sequence of the polypeptide is spliced ​​with the secondary structure ratio data of the target protein to form a new feature vector, so that the model can learn multiple aspects of information about the polypeptide and the target protein at the same time. New cross features are generated based on the original features, such as multiplying the binding affinity data with the charge properties of the polypeptide to obtain a new feature for exploring the effect of the interaction between different factors on the binding of the target protein to the polypeptide.

[0034] In this embodiment, when outputting based on multiple dimensions of a supervision model, a suitable supervision model is first selected: Classification model: To predict whether a peptide binds to a target protein (yes / no), you can use classification models such as logistic regression, decision tree, support vector machine (SVM), etc. For example, using a logistic regression model, the constructed feature vector is used as input, and the model output is the probability of binding between the peptide and the target protein, with 0.5 as the threshold for classification judgment.

[0035] Regression model: For predicting continuous data such as binding affinity values ​​and binding kinetic parameters, linear regression, polynomial regression, random forest regression and other models can be used. For example, a linear regression model is used to predict the binding constant KD value between the target protein and the peptide. The model takes the combined feature vector as input and outputs the predicted KD value.

[0036] Deep learning model: For complex nonlinear relationships, neural network models can be used, such as multi-layer perceptron (MLP), convolutional neural network (CNN), etc. For example, a multi-layer perceptron is constructed for multi-dimensional output. The input layer receives the combined feature vector, and the complex pattern of the data is learned through multiple hidden layers. The output layer can simultaneously output the results of multiple dimensions such as binding affinity, binding kinetic parameters, and predicted binding sites.

[0037] Furthermore, set the multi-dimensional output: Output layer design: Determine the number of output layer neurons based on the different dimensions of data that need to be predicted. For example, if you want to output binding affinity, binding rate constant, and binding dissociation constant at the same time, the output layer can have 3 neurons, corresponding to the predicted values ​​of these three different dimensions.

[0038] Model training and optimization: Use the training set data to train the model, adjust the model parameters through the back propagation algorithm (such as in the neural network), and minimize the loss function. During the training process, use the validation set data to evaluate the model and adjust the hyperparameters (such as learning rate, regularization coefficient, etc.) to prevent overfitting and improve the generalization ability of the model. For example, use the stochastic gradient descent algorithm to adjust the weights and biases of the neural network, and optimize the hyperparameters through indicators such as validation set accuracy and mean square error.

[0039] Model evaluation: Use the test set data to evaluate the trained model and calculate the evaluation indicators of different dimensional outputs. For classification tasks, commonly used indicators include accuracy, recall, F1 value, etc.; for regression tasks, commonly used indicators include mean square error (MSE), mean absolute error (MAE), etc. For example, in the regression task of predicting binding affinity, the mean square error on the test set is calculated to evaluate the deviation between the model prediction value and the true value.

[0040] In an exemplary embodiment, the step of training and validating the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain a prediction model includes: Obtaining a training data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to train and verify the initial model; A test data set is obtained from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to perform prediction evaluation on the trained model to obtain a prediction model.

[0041] In an embodiment of the present application, the acquired physicochemical properties and structural information of the polypeptide molecules, the physicochemical properties and structural information of the target protein, and the processed experimental data are divided into a training set, a validation set, and a test set according to a certain ratio. In a scenario of one embodiment, the training set accounts for 70%-80%, the validation set accounts for 10%-15%, and the test set accounts for 10%-15%. Among them, it is necessary to ensure that the division process is random and can represent the characteristic distribution of the entire data set. For example, for 1000 groups of polypeptide-target protein interaction data, 700 groups are randomly selected as training sets, 150 groups as validation sets, and 150 groups as test sets. The initial model based on the prediction of the binding strength of small molecule ligands and receptor proteins is trained by the training set, and the trained model is verified by the validation set. The trained model is predicted and evaluated by the test set. In a scenario of one embodiment, the consistency index of the prediction results is calculated to evaluate the evaluation performance of the model, and the calculation formula is as follows:

[0042]

[0043] in, Z is the normalization constant; b x For greater affinity d x The predicted value of b y For greater affinity d y The predicted value of h (x) is a step function; CI is the consistency index, when the consistency index CI The higher the value, the smaller the difference between the sample prediction value and the true value.

[0044] In an exemplary scenario, the root mean square error of the prediction results is calculated to evaluate the evaluation performance of the model, and the calculation formula is as follows:

[0045] in, n is the number of samples; MSE For the predicted value p i ( i =1, 2, …, n ) and the true value y i The average of the sum of squares of the differences between MSE The smaller it is, the closer the predicted value of the sample is to the true value.

[0046] In an exemplary embodiment, before the step of training and validating the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain the prediction model, the step includes: Modifying the algorithm of the initial model based on the physicochemical properties and structural information of the polypeptide molecule; In an embodiment of the present application, the algorithm is modified based on the pear blossom properties of the polypeptide molecule. In an exemplary embodiment, the amino acid composition information of the polypeptide molecule is added to the input feature. In a scenario of an embodiment, One-Hot encoding can be used to represent 20 common amino acids as a 20-dimensional vector. For each polypeptide, the frequency of occurrence of its amino acids is counted and a corresponding vector is formed. For example, for a polypeptide "ACDEG" containing 5 amino acids, its amino acid composition vector can be expressed as [1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], indicating that the polypeptide contains 1 A, 1 C, 1 D, 1 E and 1 G. In an exemplary embodiment, the length of the polypeptide is input as a feature. In this embodiment, longer polypeptides may have more potential binding sites, which may affect their binding strength with receptor proteins. The length of the polypeptide can be used directly as a feature, or it can be normalized (for example, mapped to the interval [0, 1]). In an exemplary embodiment, a secondary structure prediction tool (such as PSIPRED) is used to predict the secondary structure of the polypeptide and convert it into a feature. For example, One-Hot encoding can be used to represent α-helix, β-fold and random coil as [1, 0, 0], [0, 1, 0] and [0, 0,1], and then the secondary structure information of the polypeptide is sequentially spliced ​​into a vector as input. In an exemplary embodiment, the hydrophilicity map of the polypeptide is calculated and used as a feature vector. For example, the hydrophilicity of each amino acid is calculated using the Kyte-Doolittle scale, and then the hydrophilicity map of the entire polypeptide is added as a vector to the input feature. In an exemplary embodiment, the net charge of the polypeptide at different pH is calculated and used as an input feature. The net charge at physiological pH can be calculated based on the pKa value of the amino acid, taking into account the number and distribution of acidic and basic amino acids in the polypeptide. Based on the above-mentioned newly added features, the dimension of the input layer is expanded. For example, if the original input layer is designed for the physicochemical properties of small molecule ligands and receptor proteins, the number of neurons in the input layer needs to be increased accordingly based on the newly added peptide molecular features. Consider inputting peptide features and small molecule ligand features into different input layers, and then merging them together using the concatenate layer to better retain their respective information. Through the above modifications, the model based on the prediction of the binding strength between small molecule ligands and receptor proteins can be better adapted to scenarios containing the physicochemical properties and structural information of peptide molecules, thereby improving the model's ability to predict binding strength. Please adjust and optimize according to the actual situation to achieve the best performance.

[0047] The initial model is trained with a smaller number of iterations based on a grid search method to determine the learning parameters of the initial model, wherein the learning parameters include: an initial learning rate and a batch size.

[0048] refer to Figure 2 As shown, the present application provides a polypeptide activity screening system based on artificial intelligence and entity library, including: a data acquisition module, a data processing module, a model training module and a model execution module. In this embodiment, the physicochemical properties and structural information of polypeptide molecules, the physicochemical properties and structural information of target proteins, and the experimental data of target proteins and polypeptide molecules are obtained from the entity library through the data acquisition module, and the experimental data of the binding of target proteins and polypeptide molecules obtained are processed through the data processing module, and the initial model is trained and verified according to the physicochemical properties and structural information of polypeptide molecules, the physicochemical properties and structural information of target proteins, and the processed experimental data through the model training module to obtain a prediction model; finally, the target protein information and polypeptide information are input into the prediction model through the model execution module to predict the affinity between the target protein and the polypeptide molecule, thereby completing the screening of polypeptide activity.

[0049] In summary, the use of artificial intelligence algorithm models can rapidly expand the scale of screening compounds (random permutations and combinations of 20 natural amino acids and corresponding non-natural amino acids in peptide sequences) from the current million level to tens of millions or even hundreds of billions (random permutations and combinations of 20 natural amino acids and corresponding non-natural amino acids in peptide sequences); rapidly reduce screening costs (only a small number of artificial intelligence algorithms and chemical researchers are required to complete, and only a small number of verification experiments are required, from millions of real experiments to hundreds of real experiments), greatly improving experimental efficiency, from real screening experiments of millions of compounds that take months to a cycle of about 2 weeks of artificial intelligence virtual screening + a small number of real experiments.

[0050] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-mentioned methods for screening polypeptide activity based on artificial intelligence and entity libraries.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-mentioned methods for screening polypeptide activity based on artificial intelligence and a physical library.

[0052] Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0053] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0054] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0055] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0056] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0057] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0058] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0059] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for screening polypeptide activity based on artificial intelligence and entity library, characterized in that: The screening method comprises: Obtain the physicochemical properties and structural information of peptide molecules, the physicochemical properties and structural information of target proteins, and experimental data on the binding of target proteins to peptide molecules from the entity library; Processing the experimental data of the binding between the target protein and the polypeptide molecule; Based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data, the initial model is trained and verified to obtain a prediction model; The protein structure and polypeptide sequence to be screened are input into the prediction model for affinity prediction to complete the screening of polypeptide activity.

2. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 1, characterized in that: The step of processing the experimental data of the binding of the target protein to the polypeptide molecule comprises: The structure of the target protein combined with the polypeptide molecule and the three-dimensional coordinates of the internal atoms are converted into feature data that can be recognized by the initial model to obtain a feature matrix; Combine and split the experimental data of target proteins and peptide molecules.

3. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 2, characterized in that: The step of training and verifying the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the experimental data after combination and splitting to obtain a prediction model includes: Obtaining a training data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to train and verify the initial model; A test data set is obtained from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to perform prediction evaluation on the trained model to obtain a prediction model.

4. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 3, characterized in that: The step of obtaining a test data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to perform prediction evaluation on the trained model to obtain a prediction model comprises: Acquire a test data set from the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein, and the processed experimental data to predict the trained model; The prediction results are evaluated based on the preset evaluation indicators.

5. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 4, characterized in that: The preset evaluation index includes a consistency index, which is calculated as follows: ； in, Z is the normalization constant; b x For greater affinity d x The predicted value of b y For greater affinity d y The predicted value of CI is the consistency index, CI The higher the value, the smaller the difference between the sample prediction value and the true value; h(x) is a step function; in, h(x) The formula is as follows: 。 6. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 4, characterized in that: The preset evaluation index includes the root mean square error, which is calculated as follows: ； in, n is the number of samples; MSE For the predicted value p i ( i =1, 2, …, n ) and the true value y i The average of the sum of squares of the differences between MSE The smaller it is, the closer the predicted value of the sample is to the true value.

7. The method for screening polypeptide activity based on artificial intelligence and entity library according to claim 1, characterized in that: Before the step of training and validating the initial model based on the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain the prediction model, the step includes: Modifying the algorithm of the initial model based on the physicochemical properties and structural information of the polypeptide molecule; The initial model is trained with a smaller number of iterations based on a grid search method to determine learning parameters of the initial model, wherein the learning parameters include: an initial learning rate and / or a batch size.

8. A polypeptide activity screening system based on artificial intelligence and entity library, characterized in that: The system comprises: A data acquisition module is used to obtain the physicochemical properties and structural information of peptide molecules, the physicochemical properties and structural information of target proteins, and the experimental data of target proteins and peptide molecules from the entity library; A data processing module, used to process the experimental data information of the binding between the target protein and the polypeptide molecule; A model training module, used to train and verify the initial model according to the physicochemical properties and structural information of the polypeptide molecule, the physicochemical properties and structural information of the target protein and the processed experimental data to obtain a prediction model; The model execution module is used to input the target protein information and peptide information into the prediction model to predict the affinity strength.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the polypeptide activity screening method based on artificial intelligence and a physical library as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the polypeptide activity screening method based on artificial intelligence and entity library described in any one of claims 1 to 7 are implemented.

Images