Method and device for predicting protein-ligand binding affinity

Abstract

The invention discloses a method and device for predicting protein-ligand binding affinity based on distance and atomic contact number, comprising: acquiring a data set of protein-ligand binding affinity, the data set comprising: atomic information of a protein receptor and atomic information of a ligand; classifying atoms of a protein-ligand compound in the data set according to a preset atom type, and calculating a distance and an atom contact number of an atom pair of a protein-ligand in the data set according to the atom type; training a preset model by taking the protein-ligand distance and the atom contact number as training data to obtain a protein-ligand binding affinity prediction model; and obtaining to-be-predicted protein-ligand data information, inputting the to-be-predicted protein-ligand data information into the protein-ligand binding affinity prediction model, and outputting the to-be-predicted protein-ligand binding affinity. The method is not limited to the power exponent of the distance in the traditional electrostatic interaction energy and Van der Waals interaction energy, and the prediction capability of the affinity is improved.

Description

Technical Field

[0001] The invention belongs to the field of computer biology, and in particular to a method and device for predicting protein-ligand binding affinity. Background Art

[0002] Among the existing protein-ligand binding affinity prediction methods, the traditional linear model is just a linear superposition of some traditional energy terms and geometric terms. Among them, the problem with the traditional linear model is that "it is difficult to further improve the prediction effect without reconsidering the effective form of the energy term for protein-ligand interactions at the atomic level". The prediction effect of nonlinear machine learning models is often better than that of traditional linear models, but without specific mathematical expressions, the generalization ability is not very good.

[0003] Therefore, how to find an effective mathematical expression to calculate the energy of protein-ligand interactions at the atomic level in order to improve the accuracy of protein-ligand binding affinity prediction has become an urgent problem to be solved. Summary of the invention

[0004] To solve the above technical problems, the embodiments of the present invention provide a method for predicting protein-ligand binding affinity based on distance and atomic contact number, comprising:

[0005] Acquiring a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand;

[0006] Classifying atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating distances and atomic contacts of atomic pairs of the protein-ligand in the data set according to the atomic types;

[0007] Using the protein-ligand distance and the number of atomic contacts as training data to train a preset model to obtain a protein-ligand binding affinity prediction model;

[0008] The protein-ligand data information to be predicted is obtained and input into the protein-ligand binding affinity prediction model, and the protein-ligand binding affinity to be predicted is output.

[0009] Furthermore, the method of obtaining a data set of protein-ligand binding affinity comprises:

[0010] Obtain protein-ligand binding affinity data from a preset database;

[0011] Data satisfying preset conditions are selected from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, no heavy atom collision occurs between the protein and the ligand and the distance between all heavy atom pairs in the protein and the ligand is greater than a third preset value, and no preset atom exists in the protein and the ligand.

[0012] Furthermore, the classifying the atoms of the protein-ligand complex in the data set according to preset atom types, and calculating the distances and atomic contacts of the atom pairs of the protein-ligand in the data set according to the atom types, comprises:

[0013] Comparing the protein receptor atoms and the ligand atoms in the protein-ligand with the preset atomic types respectively, and classifying the protein-ligand into the preset atomic binding type;

[0014] Calculate the atomic contact number variable n according to the preset atomic bonding type k The distance variable d between the atom pairs k .

[0015] Furthermore, the method of using the atomic distance and atomic contact number of the protein-ligand as training data to train a preset model to obtain a protein-ligand binding affinity prediction model includes:

[0016] Multiple preliminary linear models were trained using the protein-ligand atom pair distances and atomic contact numbers in the dataset;

[0017] A linear model is trained by selecting a prediction value of a preliminary linear model whose prediction accuracy is greater than a preset value from the trained multiple preliminary linear models, and the linear model is combined to obtain the protein-ligand binding affinity prediction model.

[0018] Furthermore, the step of obtaining the protein-ligand data information to be predicted includes:

[0019] Compare the protein receptor atoms and ligand atoms in the protein-ligand to be predicted with the preset atomic types, and determine the atomic binding type;

[0020] Calculate the atomic contact number variable n according to the determined atomic bonding type k The distance variable d between the atom pairs k .

[0021] A device for predicting protein-ligand binding affinity based on distance and atomic contact number, comprising:

[0022] An acquisition module is used to acquire a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand;

[0023] A processing module, used for classifying atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating distances and atomic contacts of atomic pairs of the protein-ligand in the data set according to the atomic types;

[0024] The processing module is used to train a preset model using the protein-ligand distance and the atomic contact number as training data to obtain a protein-ligand binding affinity prediction model;

[0025] The execution module is used to obtain the protein-ligand data information to be predicted, input it into the protein-ligand binding affinity prediction model, and output the protein-ligand binding affinity to be predicted.

[0026] Furthermore, the acquisition module includes:

[0027] The first acquisition submodule is used to obtain protein-ligand binding affinity data from a preset database;

[0028] The second acquisition submodule is used to select data that meets preset conditions from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, there is no heavy atom collision between the protein and the ligand and the distance between all heavy atom pairs in the protein-ligand is greater than a third preset value, and there is no preset atom in the protein-ligand.

[0029] Furthermore, the processing module includes:

[0030] A first processing submodule, for comparing the protein receptor atoms and the ligand atoms in the protein-ligand with preset atomic types respectively, and classifying the protein-ligand into preset atomic binding types;

[0031] The second processing submodule is used to calculate the atomic contact number variable n according to the preset atomic bonding type. k The distance variable d between the atom pairs k .

[0032] Furthermore, the processing module includes:

[0033] a third processing submodule for respectively training a plurality of preliminary linear models using the atom pair distances and the atomic contact numbers of the protein-ligand in the data set;

[0034] The fourth processing submodule is used to select a prediction value training linear model of a preliminary linear model with a prediction accuracy greater than a preset value from the trained multiple preliminary linear models, and combine them to obtain the protein-ligand binding affinity prediction model.

[0035] Furthermore, the execution module includes:

[0036] The first execution submodule is used to compare the protein receptor atoms and ligand atoms in the protein-ligand to be predicted with the preset atomic types respectively, and determine the atomic binding type;

[0037] The second execution submodule is used to calculate the atomic contact number variable n according to the determined atomic bonding type k The distance variable d between the atom pairs k .

[0038] A computer device comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor performs the steps of the method for predicting protein-ligand binding affinity based on distance and atomic contact number as described above.

[0039] A storage medium storing computer-readable instructions, when the computer-readable instructions are executed by one or more processors, causes the one or more processors to perform the steps of the method for predicting protein-ligand binding affinity based on distance and atomic contact number as described above.

[0040] The present invention re-characterizes the quantitative distance-energy relationship and quantitative contact number-energy relationship for protein-ligand interaction at the atomic level. It is not limited to the power exponent of distance in traditional electrostatic interaction energy and van der Waals interaction energy. Compared with traditional linear models, it has certain improvements and enhancements in predicting protein-ligand affinity, and has reference significance for the development of future protein-ligand affinity prediction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A schematic flow chart of a method for predicting protein-ligand binding affinity based on distance and atomic contact number provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0045] like Figure 1 As shown, the present invention provides a method for predicting protein-ligand binding affinity based on distance and atomic contact number, comprising:

[0046] S1. Obtaining a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand;

[0047] S2, classifying atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating distances and atomic contacts of atomic pairs of the protein-ligand in the data set according to the atomic types;

[0048] S3, using the protein-ligand distance and atomic contact number as training data to train a preset model to obtain a protein-ligand binding affinity prediction model;

[0049] S4. Obtaining the protein-ligand data information to be predicted, and inputting it into the protein-ligand binding affinity prediction model, and outputting the protein-ligand binding affinity to be predicted.

[0050] The present invention re-characterizes the quantitative distance-energy relationship and quantitative contact number-energy relationship for protein-ligand interaction at the atomic level. It is not limited to the power exponent of distance in traditional electrostatic interaction energy and van der Waals interaction energy. Compared with traditional linear models, it has certain improvements and enhancements in predicting protein-ligand affinity, and has reference significance for the development of future protein-ligand affinity prediction methods.

[0051] In some embodiments, obtaining a protein-ligand binding affinity dataset in S1 includes:

[0052] Step 1: Obtain protein-ligand binding affinity data from a preset database;

[0053] Step 2: Select data that meets preset conditions from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, there is no heavy atom collision between the protein and the ligand and the distance between all heavy atom pairs in the protein-ligand is greater than a third preset value, and there is no preset atom in the protein-ligand.

[0054] In one embodiment of the present invention, the protein-ligand binding affinity data comes from PDBbind v.2020 [1,2] and the protein-ligand binding affinity prediction method evaluation dataset CASF-2016 [3]. Since the data quality in PDBbind v.2020 varies, data screening is required.

[0055] The preconditions for selecting high-quality data are as follows: there is a confirmed experimentally determined binding affinity, i.e., Ki or Kd (Ki or Kd is the equilibrium dissociation constant of the inhibitor or ligand); the number of heavy atoms of the ligand is ≤70, and the total number of atoms of the ligand is ≤150. There are no heavy atom collisions in the experimentally determined complex structure. The distances of all heavy atom pairs in the complex structure are In the ligand binding pocket, there are no unidentified or uncommon atom types (e.g., B, Du, Ni, Cd, As, Ir, Y, Yb, V, Ru, Te, Pr, Cs, Be, Sr, Rh, Re, Rb, Pt, In, Ga, Eu, Au).

[0056] After screening, 5171 protein-ligand complexes that met the above conditions were from the general set (SET-1) of PDBbind v.2020; 4958 protein-ligand complexes were from the refined set (SET-2) of PDBbind v.2020; and 285 protein-ligand complexes were from CASF-2016 (SET-3).

[0057] In some embodiments, S2 classifies atoms of the protein-ligand complex in the data set according to preset atom types, and calculates distances and atomic contacts of atom pairs of the protein-ligand in the data set according to the atom types, including:

[0058] Step 1: comparing the protein receptor atoms and the ligand atoms in the protein-ligand with the preset atomic types respectively, and classifying the protein-ligand into the preset atomic binding type;

[0059] In this example, the preset atomic types are as follows: For each complex, the atom types were specified using Open Babel 2.4.1. In the receptor, there are 33 heavy atom types (O.2, C.3, C.2, N.am, O.3, C.ar, O.co2, N.ar, N.pl3, S.3, C.cat, N.3, N.2, Zn, N.4, Mg, P.3, C.1, Ca, Cl, Mn, Na, SO, Se, Fe, N.1, K, F, S.O2, Co.oh, S.2, I, Br); in the ligand, there are 25 heavy atom types (C.3, C.ar, O.2, C.2, O.3, N.am, N.a r, N.3, N.pl3, O.co2, P.3, N.2, S.3, S.O2, Cl, F, S.2, C.1, C.cat, N.4, N.1, Br, I, Fe, SO); in standard amino acids, there are 18 heavy atom types (O.2, C.3, C.2, N.am, O.3, C.ar, O.co2, N.ar, N.pl3, S.3, C.cat, N.3, N.2, N.4, C.1, N.1, S.2, SO); the oxygen atom type in water molecules is O.3. The definition of atom types is shown in Table 1.

[0060] Table 1 Definition of heavy atom types

[0061]

[0062]

[0063] Step 2: Calculate the atomic contact number variable n according to the preset atomic bonding type k The distance variable d between the atom pairs k .

[0064] In this example, the distance between two atoms from the protein receptor and the ligand is The two heavy atoms are defined as atomically touching. The interaction strength of an atom pair decreases as the distance increases. In order to reduce the amount of calculation, the distance is cut off. Used to define the binding pocket. If an atom pair consists of different atom types (atom type 1 and 2), there are two cases: atom type 1 comes from the receptor and atom type 2 comes from the ligand; atom type 2 comes from the receptor and atom type 1 comes from the ligand. These two cases are considered separately, and the atomic environments of the receptor and ligand are different.

[0065] When dealing with protein-ligand interactions, four scenarios were used according to the different components of the receptor. In scenario 1, all heavy atoms in the receptor binding pocket were considered; in scenario 2, only the heavy atoms of the standard amino acids in the receptor binding pocket were considered; in scenario 3 (S3), the heavy atoms of the standard amino acids in the receptor binding pocket and other heavy atoms (excluding water molecules) were considered; in scenario 4, the heavy atoms in the standard amino acids and water molecules in the receptor binding pocket were considered. In scenarios 1 and 3, there were 738 interacting atom type pairs in the experimental structure; in scenarios 2 and 4, there were 415 interacting atom type pairs in the experimental structure.

[0066] Calculate the variables based on the number of atomic contacts (n) for different pairs of atom types k ) and distance-based variables (d k ), n, d, and k are the contact number, distance, and power index, respectively.

[0067] The calculation process can be referred to in the form of the preliminary linear model of the atom pair distance and atomic contact number of the protein-ligand below, which will not be described in detail here.

[0068] In some embodiments, S3 uses the atomic distance and atomic contact number of the protein-ligand as training data to train a preset model to obtain a protein-ligand binding affinity prediction model, including:

[0069] Step 1: Use the protein-ligand atom pair distances and atomic contact numbers in the dataset to train multiple preliminary linear models respectively;

[0070] In the embodiment of the present invention, a preparatory linear model for predicting protein-ligand binding affinity is trained, wherein the binding energy (E) between the protein and the ligand is proportional to log(K), that is, E=constant log(K).

[0071] The preliminary linear model of the protein-ligand atom pair distances is of the form:

[0072] log(K)=a 1 ∑d 1j -k +…+a i ∑d ij -k +…+a m ∑d mj -k +b;

[0073] Wherein, k is the number of prepared linear models, k = 1, 2, ..., 18;

[0074] The preliminary linear model for the number of atomic contacts between protein and ligand is of the form:

[0075] log(K)=a 1 n 1 k +…+a i n i k +…+a m n m k +b;

[0076] Where k is the number of prepared linear models, k = 1, 2, ..., 18; K is the inhibition or dissociation constant, a 1 ,a 2 ,…,a m is the coefficient, b is the constant term, d ij is the distance between the jth atom pair belonging to the i-th atom type pair, k is the power index, ∑d ij -k is the d of all atom pairs belonging to the ith atom type pair ij -k sum, n i is the number of atom pairs belonging to the ith atom type pair, and m is the number of atom type pairs (m = 738 or 415). In the preliminary linear model, there are 738 or 415 variables based on distance or contact number. The coefficients of the preliminary linear model are obtained using the least squares method.

[0077] Step 2: Select a linear model for training the prediction value of a preliminary linear model with a prediction accuracy greater than a preset value from the trained multiple preliminary linear models, and combine them to obtain the protein-ligand binding affinity prediction model.

[0078] In the embodiment of the present invention, the data in the training set are randomly selected from SET-1 (5171 protein-ligand complexes), SET-2 (4958 protein-ligand complexes), and SET-3 (285 protein-ligand complexes). SUBSET-1 consists of 4000 complexes randomly selected from SET-1; SUBSET-2 consists of 4000 complexes randomly selected from SET-2; SUBSET-3 consists of 200 complexes randomly selected from SET-3; SUBSET-4 consists of SUBSET-1 and SUBSET-2; SUBSET-5 consists of SUBSET-1 and SUBSET-3; SUBSET-6 consists of SUBSET-2 and SUBSET-3; SUBSET-7 consists of SUBSET-1, SUBSET-2, and SUBSET-3. These subsets are used as training sets to determine the coefficients in the preliminary linear model. The process of randomly selecting training sets was repeated 10,000 times, so that 70,000 preliminary linear models were generated for each scenario (S1-S4), each power index (k), and each distance- or contact-number-based variable.

[0079] According to the prediction effect of the model, a preliminary linear model with an accuracy greater than a preset value is selected from a large number of trained models.

[0080] The linear model trained in the embodiment of the present invention for predicting protein-ligand binding affinity includes:

[0081] Based on the predictions of the preliminary linear model, a new linear model is trained. The mathematical expression of the linear model is as follows: The form of the linear model of the distance between the atom pairs of protein-ligand is as follows:

[0082] log(K)=a 1 dlm 1 +…+a i dlm i +…+a n dlm n +b;

[0083] n is the number of prepared linear models, n = 1, 2, ..., 18

[0084] The preliminary linear model for the number of atomic contacts between protein and ligand is of the form:

[0085] log(K)=a 1 nlm 1 +…+a i nlm i +…+a n nlm n +b

[0086] n is the number of prepared linear models, n = 1, 2, ..., 18

[0087] where K is the inhibition or dissociation constant, a 1 ,a 2 ,…,a n is the coefficient, b is the constant term, dlm i and nlm i is the ith representative preliminary linear model based on distance and number of contacts. All combinations of preliminary linear models were considered, and each new linear model contained no more than 18 representative preliminary linear models. The training set was generated in the same way as the training of the preliminary linear models. The process of randomly selecting complexes was repeated 100 times. Based on these training sets, 1,835,00100 ((218–1)×700) linear models were generated. Based on the prediction effect of the model, multiple representative linear models were selected from the large number of trained models, and the model with the highest accuracy was selected as the final trained model.

[0088] These models can be implemented using MATLAB programs.

[0089] In one embodiment, the complex structure file 4ciw.pdb, the receptor is dehydroquinate synthetase, and the ligand is: XH2

[0090] ((1R,4R,5R)-1,4,5-trihydroxy-3-(2-hydroxy)ethylcyclohex-2-ene-1-carboxylic acid)

[0091] $cdscripts_20240713

[0092] #Enter the folder scripts_20240713

[0093] $. / run_script_list_20240703.pl 4ciw.pdb A 1148 XH2

[0094] #Calculate the variables based on distance and contact number. 4ciw.pdb is the structure file, A is the ligand chain number, 1148 is the ligand sequence number, XH2 is the ligand name, and the output files reciprocal_distance_sum_S1-S4 save the variables based on distance; contact_number_S1-S4 save the variables based on contact number.

[0095] $matlab-nodesktop-nosplash-rpredict_affintiy_predictoin_20240713

[0096] #After running, the affinity values ​​predicted by 8 linear models are stored in the file predicted_results_8.

[0097] $mv predicted_results_8 4ciw_predicted_results_8

[0098] #Rename the result file predicted_results_8 to 4ciw_predicted_results_8

[0099] Complex structure file 5c28.pdb, receptor is phosphodiesterase, ligand is 4XV (6-chloro-2-cyclopropyl-5-methylpyrimidin-4-amine)

[0100] $cdscripts_20240713

[0101] #Enter the folder scripts_20240713

[0102] $. / run_script_list_20240703.pl 5c28.pdb A 803 4XV

[0103] #Calculate the variables based on distance and contact number. 5c28.pdb is the structure file, A is the ligand chain number, 803 is the ligand sequence number, 4XV is the ligand name, and the output files reciprocal_distance_sum_S1-S4 save the variables based on distance; contact_number_S1-S4 save the variables based on contact number.

[0104] $matlab-nodesktop-nosplash-rpredict_affintiy_predictoin_20240713

[0105] #After running, the affinity values ​​predicted by 8 linear models are stored in the file predicted_results_8.

[0106] $mv predicted_results_8 5c28_predicted_results_8

[0107] #Rename the result file predicted_results_8 to 5c28_predicted_results_8

[0108] In some embodiments, S4 obtains protein-ligand data information to be predicted, including:

[0109] Step 1: compare the protein receptor atoms and ligand atoms in the protein-ligand to be predicted with the preset atomic types, and determine the atomic binding type;

[0110] Step 2: Calculate the atomic contact number variable n according to the determined atomic bonding type k The distance variable d between the atom pairs k .

[0111] An embodiment of the present invention also provides a device for predicting protein-ligand binding affinity based on distance and atomic contact number, comprising: an acquisition module, used to acquire a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand; a processing module, used to classify the atoms of the protein-ligand complex in the data set according to preset atomic types, and calculate the distance and atomic contact number of the atomic pairs of the protein-ligand in the data set according to the atomic type; the processing module, used to train a preset model using the protein-ligand distance and atomic contact number as training data to obtain a protein-ligand binding affinity prediction model; an execution module, used to acquire the protein-ligand data information to be predicted, and input it into the protein-ligand binding affinity prediction model, and output the protein-ligand binding affinity to be predicted.

[0112] The present invention re-characterizes the quantitative distance-energy relationship and quantitative contact number-energy relationship for protein-ligand interaction at the atomic level. It is not limited to the power exponent of distance in traditional electrostatic interaction energy and van der Waals interaction energy. Compared with traditional linear models, it has certain improvements and enhancements in predicting protein-ligand affinity, and has reference significance for the development of future protein-ligand affinity prediction methods.

[0113] In some embodiments, the acquisition module includes: a first acquisition submodule, used to acquire protein-ligand binding affinity data from a preset database; a second acquisition submodule, used to select data that meets preset conditions from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, there is no heavy atom collision between the protein and the ligand and the distance between all heavy atom pairs in the protein-ligand is greater than a third preset value, and there is no preset atom in the protein-ligand.

[0114] In some embodiments, the processing module includes: a first processing submodule, for comparing the protein receptor atoms and the ligand atoms in the protein-ligand with the preset atomic types, and classifying the protein-ligand into preset atomic binding types; a second processing submodule, for calculating the atomic contact number variable n according to the preset atomic binding type. k The distance variable d between the atom pairs k .

[0115] In some embodiments, the processing module includes: a third processing submodule, used to train multiple preliminary linear models respectively using the atom pair distances and atomic contact numbers of the protein-ligand in the data set; a fourth processing submodule, used to select a linear model for training the prediction value of the preliminary linear model whose prediction accuracy is greater than a preset value from the trained multiple preliminary linear models, and combine them to obtain the protein-ligand binding affinity prediction model.

[0116] In some embodiments, the execution module includes: a first execution submodule, which is used to compare the protein receptor atoms and the ligand atoms in the protein-ligand to be predicted with the preset atomic types respectively, and determine the atomic binding type; a second execution submodule, which is used to calculate the atomic contact number variable n according to the determined atomic binding type. k The distance variable d between the atom pairs k .

[0117] To solve the above technical problems, an embodiment of the present invention also provides a computer device. The computer device includes a processor, a non-volatile storage medium, a memory and a network interface connected by a system bus. Among them, the non-volatile storage medium of the computer device stores an operating system, a database and a computer-readable instruction, and the database may store a control information sequence. When the computer-readable instruction is executed by the processor, the processor can implement a method for predicting protein-ligand binding affinity based on distance and atomic contact number. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor may execute a method for predicting protein-ligand binding affinity based on distance and atomic contact number. The network interface of the computer device is used to connect and communicate with the terminal. In this embodiment, the processor is used to execute the specific contents of the acquisition module and the processing module, and the memory stores the program code and various types of data required to execute the above modules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all submodules in the image processing method, and the server can call the program code and data of the server to execute the functions of all submodules.

[0118] The present invention also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, causes the one or more processors to execute the steps of a method for predicting protein-ligand binding affinity based on distance and atomic contact number as described in any of the above embodiments.

[0119] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0120] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0121] The above descriptions are only some embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0122] The above content is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for predicting protein-ligand binding affinity based on distance and atomic contact number, characterized in that: include: Acquiring a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand; Classifying atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating distances and atomic contacts of atomic pairs of the protein-ligand in the data set according to the atomic types; Using the protein-ligand distance and the number of atomic contacts as training data to train a preset model to obtain a protein-ligand binding affinity prediction model; The protein-ligand data information to be predicted is obtained and input into the protein-ligand binding affinity prediction model, and the binding affinity of the protein-ligand to be predicted is output.

2. The method according to claim 1, characterized in that The method for obtaining a protein-ligand binding affinity data set comprises: Obtain protein-ligand binding affinity data from a preset database; Data satisfying preset conditions are selected from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, no heavy atom collision occurs between the protein and the ligand and the distance between all heavy atom pairs in the protein and the ligand is greater than a third preset value, and no preset atom exists in the protein and the ligand.

3. The method according to claim 1, characterized in that The step of classifying the atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating the distances and atomic contacts of the atom pairs of the protein-ligand in the data set according to the atomic types, comprises: Comparing the protein receptor atoms and the ligand atoms in the protein-ligand with the preset atomic types respectively, and classifying the protein-ligand into the preset atomic binding type; Calculate the atomic contact number variable n according to the preset atomic bonding type k The distance variable d between the atomic pairs k .

4. The method according to claim 1, characterized in that: The method of using the atomic distance and atomic contact number of the protein-ligand as training data to train a preset model to obtain a protein-ligand binding affinity prediction model comprises: Multiple preliminary linear models were trained using the protein-ligand atom pair distances and atomic contact numbers in the dataset; A linear model is trained by selecting a prediction value of a preliminary linear model whose prediction accuracy is greater than a preset value from the trained multiple preliminary linear models, and the linear model is combined to obtain the protein-ligand binding affinity prediction model.

5. The method according to claim 1, characterized in that The step of obtaining the protein-ligand data information to be predicted includes: The protein receptor atoms and ligand atoms in the protein-ligand to be predicted are respectively compared with the preset atomic types, and the atomic binding type is determined; Calculate the atomic contact number variable n according to the determined atomic bonding type k The distance variable d between the atomic pairs k .

6. A device for predicting protein-ligand binding affinity based on distance and atomic contact number, characterized in that: include: An acquisition module is used to acquire a data set of protein-ligand binding affinity, wherein the data set includes: atomic information of the protein receptor and atomic information of the ligand; A processing module, used for classifying atoms of the protein-ligand complex in the data set according to preset atomic types, and calculating distances and atomic contacts of atomic pairs of the protein-ligand in the data set according to the atomic types; The processing module is used to train a preset model using the protein-ligand distance and the atomic contact number as training data to obtain a protein-ligand binding affinity prediction model; The execution module is used to obtain the protein-ligand data information to be predicted, input it into the protein-ligand binding affinity prediction model, and output the protein-ligand binding affinity to be predicted.

7. The device according to claim 6, characterized in that The acquisition module comprises: The first acquisition submodule is used to obtain protein-ligand binding affinity data from a preset database; The second acquisition submodule is used to select data that meets preset conditions from the binding affinity data as the data set, wherein the preset conditions are: including an equilibrium dissociation constant, the number of heavy atoms of the ligand is not less than a first preset value, the total number of atoms of the ligand is not less than a second preset value, there is no heavy atom collision between the protein and the ligand and the distance between all heavy atom pairs in the protein-ligand is greater than a third preset value, and there is no preset atom in the protein-ligand.

8. The device according to claim 6, characterized in that The processing module comprises: A first processing submodule, for comparing the protein receptor atoms and the ligand atoms in the protein-ligand with preset atomic types respectively, and classifying the protein-ligand into preset atomic binding types; The second processing submodule is used to calculate the atomic contact number variable n according to the preset atomic bonding type. k The distance variable d between the atomic pairs k .

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the method for predicting protein-ligand binding affinity based on distance and atomic contact number as described in any one of claims 1 to 5 above.

10. A storage medium storing computer-readable instructions, which, when executed by one or more processors, causes the one or more processors to perform the steps of the method for predicting protein-ligand binding affinity based on distance and atomic contact number as described in any one of claims 1 to 5.

Images