Protein binding conformation prediction method and system
By constructing a dynamic probability matrix and combining residue contact frequency with motion degrees of freedom, the problem of capturing dynamic features in protein binding conformation prediction in existing technologies has been solved, achieving more accurate conformation prediction and improved system performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-05
- Publication Date
- 2026-03-17
AI Technical Summary
Existing technologies cannot accurately reflect the dynamic conformational fluctuations and interaction changes of residues during the binding process in protein binding conformation prediction, leading to false negative results and failing to meet the needs of drug design and functional studies for dynamic characterization.
A residue index table is constructed by extracting atomic information from static structure files, the distance matrix and motion degrees of freedom between residue pairs are calculated, and a dynamic distance threshold is obtained by adaptive weighting and correlation correction. The result is converted into a binary contact map, and a contact frequency matrix and a symmetric matrix are constructed. These are then fused and probabilistically propagated to generate a dynamic probability matrix to predict protein binding conformation.
This improves the accuracy of conformational prediction, avoids high computational costs, and enhances the overall operating efficiency and performance of the system.
Smart Images
Figure CN121687282A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of protein structure dynamic analysis, in particular to a protein binding conformation prediction method and system. BACKGROUND
[0002] With the rapid development of drug research and development, protein function mechanism analysis, disease target mining and other fields, the accurate prediction of protein binding conformation is increasingly urgent. Precise binding conformation prediction is the core support for promoting the efficiency of targeted drug design, revealing the mechanism of protein interaction, and clarifying the pathogenesis of diseases. Especially in the dynamic environment of protein binding process, the capture accuracy of dynamic change characteristics of the binding site, the reliability of conformation prediction and the scene adaptability put forward higher requirements.
[0003] However, the current protein binding conformation prediction technology still has significant limitations. Most methods rely on protein static structure data for analysis. Such data can only record the three-dimensional coordinates of atoms at a specific moment of the protein, and cannot reflect the dynamic conformation fluctuation and interaction change of residues in the binding process, resulting in false negative results in prediction, which is difficult to meet the demand of dynamic feature description for drug design and function research. SUMMARY
[0004] Therefore, it is necessary to solve the above problems, and the present application provides a protein binding conformation prediction method and system.
[0005] In a first aspect, the present application provides a protein binding conformation prediction method, which comprises:
[0006] obtaining a static structure file of a target protein; constructing a residue index table based on atomic information extracted from the static structure file; performing standardization processing on atomic coordinates in the residue index table to obtain standardized atomic coordinates;
[0007] calculating a distance matrix between residue pairs based on the standardized atomic coordinates;
[0008] obtaining residue motion freedom degrees based on conformation sampling data in the static structure file and the residue index table through adaptive weighting calculation and correlation correction;
[0009] obtaining a dynamic distance threshold of residue pairs based on the residue motion freedom degrees; and converting the distance matrix into a binary contact map based on the dynamic distance threshold;
[0010] obtaining contact frequencies of residue pairs based on the binary contact map; and constructing a contact frequency matrix based on the contact frequencies;
[0011] extending the residue motion freedom degrees into a symmetric matrix;
[0012] The contact frequency matrix and the symmetric matrix are fused to obtain the initial dynamic probability matrix;
[0013] The initial probability matrix is then normalized and subjected to probability propagation in sequence to obtain the dynamic probability matrix;
[0014] The protein binding conformation is predicted based on the dynamic probability matrix.
[0015] In some embodiments, a residue index table is constructed based on the atomic information extracted from the static structure file, including:
[0016] The static structure file is parsed to obtain information about all atoms; the information includes the atom name, chain identifier, residue number, insertion code, residue name, coordinates, and temperature factor for each atom.
[0017] The chain identifier, residue number, and insertion code are used as residue identifiers. Based on the residue identifiers, all atoms are grouped to obtain the set of atoms for each residue.
[0018] The atomic set of each residue is separated into main chain and side chain, and the main chain coordinates and side chain coordinates are obtained;
[0019] Extract the residue-level attributes of each group of atoms; the residue-level attributes include chain identifier, residue number, insertion code, residue name, and temperature factor;
[0020] A residue index table is constructed based on the residue identifiers, main chain coordinates, side chain coordinates, and residue-level attributes.
[0021] In some embodiments, the method of calculating the distance matrix between residue pairs based on the normalized atomic coordinates includes:
[0022] The distance D between each pair of residues is calculated based on the following formula. ij ,include:
[0023]
[0024] Among them, D ij The Cα atom distance between each pair of residues i and j; (x i ,y i ,z i (x) represents the Cα atomic coordinates of residue i; j ,y j ,z j () represents the Cα atomic coordinates of residue j;
[0025] The distance D between all residue pairs ij Arrange them to obtain the distance matrix between residue pairs.
[0026] In some embodiments, the degrees of freedom for residue motion are obtained through adaptive weighted calculation and correlation correction based on conformational sampling data in the static structure file and the residue index table, including:
[0027] Principal component analysis is performed on the conformational sampling data in the static structure file to obtain the motion components of the residues.
[0028] The temperature factor of the residues in the residue index table is normalized.
[0029] The motion components and the temperature factor of the normalized residues are used to calculate the mixed motion degrees of freedom through an adaptive weighting strategy;
[0030] The hybrid motion degrees of freedom are corrected for correlation to obtain the residue motion degrees of freedom.
[0031] In some embodiments, a dynamic distance threshold for residue pairs is obtained based on the degrees of freedom of residue motion; the distance matrix is converted into a binary contact map based on the dynamic distance threshold, including:
[0032] The dynamic distance threshold R is obtained based on the following formula. ij :
[0033]
[0034] Among them, R ij d is the dynamic distance threshold for residue pairs (i, j); i d represents the degrees of freedom for the motion of residue i; j is the degree of freedom of motion of residue j; 0.2 and 0.8 are fixed coefficients;
[0035] The distance matrix is converted into a binary contact map based on the following formula:
[0036]
[0037] Among them, CM ij D represents the element corresponding to residue i and residue j in the binary contact diagram, with a value of 1 or 0; ij R represents the Cα atom distance between each pair of residues i and j; ij is the dynamic distance threshold between residue i and residue j.
[0038] In some embodiments, the contact frequency of residue pairs is obtained based on the binary contact map, including:
[0039] The basic contact frequency is calculated based on the binary contact diagram.
[0040] The base contact frequency is Gaussian smoothed to obtain the desired contact frequency.
[0041] In some embodiments, the contact frequency matrix and the symmetric matrix are fused to obtain an initial dynamic probability matrix, including:
[0042] The contact frequency matrix and the symmetric matrix are fused using the Hadamard product;
[0043] The fusion results are logarithmically transformed to obtain the initial dynamic probability matrix.
[0044] In some embodiments, the initial probability matrix is sequentially normalized and probability propagated to obtain a dynamic probability matrix, including:
[0045] The initial probability matrix is normalized using a dual normalization strategy;
[0046] The PageRank algorithm is used to propagate the normalized initial probability matrix to obtain the dynamic probability matrix.
[0047] In some embodiments, the normalized initial probability matrix is propagated based on the following formula:
[0048]
[0049] in, P represents the interaction probability of residue pair (i,j) after probability propagation; ij λ represents the original probability of residue pair (i,j); λ is the retention coefficient; P ik P represents the probability of residue pair (i,k); kj Let be the probability of residue pair (k,j).
[0050] Secondly, this application also provides a protein binding conformation prediction system, the protein binding conformation prediction system comprising:
[0051] The data acquisition module is used to acquire the static structure file of the target protein.
[0052] The data preprocessing module is used to construct a residue index table based on the atomic information extracted from the static structure file; and to perform normalization processing on the atomic coordinates in the residue index table to obtain normalized atomic coordinates.
[0053] The feature extraction module is used to calculate the distance matrix between residue pairs based on the standardized atomic coordinates; obtain the residue motion degrees of freedom through adaptive weighting calculation and correlation correction based on the conformation sampling data in the static structure file and the residue index table; obtain the dynamic distance threshold of residue pairs based on the residue motion degrees of freedom; convert the distance matrix into a binary contact map based on the dynamic distance threshold; and obtain the contact frequency of residue pairs based on the binary contact map.
[0054] A dynamic probability matrix generation module is used to construct a contact frequency matrix based on the contact frequency; extend the residue motion degrees of freedom into a symmetric matrix; fuse the contact frequency matrix and the symmetric matrix to obtain an initial dynamic probability matrix; and perform normalization and probability propagation on the initial probability matrix in sequence to obtain the dynamic probability matrix.
[0055] The prediction module is used to predict the protein binding conformation based on the dynamic probability matrix.
[0056] The protein binding conformation prediction method and system of this application constructs a dynamic probability matrix by extracting the joint probability distribution of residue contact frequency and motion degrees of freedom from the static structure, and indirectly infers the possibility of conformational changes during the binding process. This improves the accuracy of binding conformation prediction, while avoiding the high computational cost of directly using molecular dynamics simulations, and effectively improves the overall operating efficiency and performance of the system. Attached Figure Description
[0057] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0058] Figure 1 This is a flowchart of a protein binding conformation prediction method provided in one embodiment of this application;
[0059] Figure 2 This is a structural block diagram of a protein binding conformation prediction system provided in another embodiment of this application;
[0060] Figure labeling: 10, Data acquisition module; 20, Data preprocessing module; 30, Feature extraction module; 40, Dynamic probability matrix generation module; 50, Prediction module. Detailed Implementation
[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0062] In one embodiment, see Figure 1 This application provides a method for predicting protein binding conformation, which includes the following steps: S11 to S19.
[0063] S11: Obtain the static structure file of the target protein; construct a residue index table based on the atomic information extracted from the static structure file; perform normalization processing on the atomic coordinates in the residue index table to obtain normalized atomic coordinates.
[0064] S12: Calculate the distance matrix between residue pairs based on the standardized atomic coordinates.
[0065] S13: Based on the conformation sampling data in the static structure file and the residue index table, the degrees of freedom of residue motion are obtained through adaptive weighted calculation and correlation correction.
[0066] S14: Obtain the dynamic distance threshold of the residue pair based on the degrees of freedom of the residue motion; convert the distance matrix into a binary contact map based on the dynamic distance threshold.
[0067] S15: Obtain the contact frequency of residue pairs based on the binary contact diagram; construct a contact frequency matrix based on the contact frequency.
[0068] S16: Extend the degrees of freedom of the residue motion into a symmetric matrix.
[0069] S17: The contact frequency matrix and the symmetric matrix are fused to obtain the initial dynamic probability matrix.
[0070] S18: The initial probability matrix is normalized and probability propagated sequentially to obtain the dynamic probability matrix.
[0071] S19: Predict the protein binding conformation based on the dynamic probability matrix.
[0072] The protein binding conformation prediction method of this application constructs a dynamic probability matrix by extracting the joint probability distribution of residue contact frequency and motion degrees of freedom from the static structure, and indirectly infers the possibility of conformational changes during the binding process. This improves the accuracy of binding conformation prediction, while avoiding the high computational cost of directly using molecular dynamics simulation, and effectively improves the overall operating efficiency and performance of the system.
[0073] In step S11, please refer to Figure 1In step S11, a static structure file of the target protein is obtained; a residue index table is constructed based on the atomic information extracted from the static structure file; and the atomic coordinates in the residue index table are normalized to obtain normalized atomic coordinates.
[0074] As an example, constructing a residue index table based on the atomic information extracted from the static structure file may include the following:
[0075] The static structure file is parsed to obtain information about all atoms; the information includes the atom name, chain identifier, residue number, insertion code, residue name, coordinates, and temperature factor for each atom.
[0076] The chain identifier, residue number, and insertion code are used as residue identifiers. Based on the residue identifiers, all atoms are grouped to obtain the set of atoms for each residue.
[0077] The atomic set of each residue is separated into main chain and side chain, and the main chain coordinates and side chain coordinates are obtained;
[0078] Extract the residue-level attributes of each group of atoms; the residue-level attributes include chain identifier, residue number, insertion code, residue name, and temperature factor;
[0079] A residue index table is constructed based on the residue identifiers, main chain coordinates, side chain coordinates, and residue-level attributes.
[0080] Specifically, a standard amino acid atom data (ATOM) and residue sequence information (SEQRES) are read line by line from the protein database using a standard format parser for protein database files. Water molecules (HOH), ions (such as CA, MG), ligand molecules, and irrelevant annotations are excluded to generate a static structure file.
[0081] Specifically, the static structure file is parsed to extract detailed information for all atoms, including atom name, chain identifier, residue number, insertion code, residue name, three-dimensional coordinates, and temperature factor. Using the chain identifier, residue number, and insertion code as unique residue identifiers, all atoms are grouped according to these identifiers to obtain the atom set corresponding to each residue. Then, the main chain and side chains of each residue's atom set are separated (main chain atoms are N, CA, C, O, etc., and the rest are side chain atoms), and the main chain coordinates and side chain coordinates are extracted separately. Simultaneously, residue-level attributes for each group of atoms are extracted, including chain identifier, residue number, insertion code, residue name, and the average temperature factor of all atoms in that residue. Finally, the residue identifier, main chain coordinates, side chain coordinates, and residue-level attributes are integrated to construct a residue index table, achieving a structured association between residue and atom data.
[0082] As an example, normalizing the atomic coordinates in the residue index table includes:
[0083] The original coordinates can be translated to the geometric center of the protein to unify the coordinate system based on the following formula:
[0084]
[0085] Among them, (x i ,y i ,z i (x′) represents the original coordinates of the i-th atom; i ,y′ i ,z′ i (i) represents the i-th normalized coordinate; These represent the sum of the original coordinates of all N atoms on the corresponding X, Y, and Z axes; The average coefficient is denoted by N; N is the total number of atoms.
[0086] will Convert to nanometers (nm), and divide all coordinate values by 10 to unify the units;
[0087] Least squares fitting is performed on the main chain atoms (N, CA, C) of each residue to align them with the ideal coordinates in the standard amino acid conformation library, thus eliminating the deviation caused by the crystal stacking effect.
[0088] In step S12, please refer to Figure 1 In step S12, the distance matrix between residue pairs is calculated based on the standardized atomic coordinates.
[0089] As an example, calculating the distance matrix between residue pairs based on the normalized atomic coordinates includes:
[0090] Calculate the distance D between each pair of residues. ij The calculation formula can be:
[0091]
[0092] Among them, D ij The Cα atom distance between each pair of residues i and j; (x i ,y i ,z i (x) represents the Cα atomic coordinates of residue i; j ,y j ,z j () represents the Cα atomic coordinates of residue j;
[0093] The distance D between all residue pairs ij Arrange them to obtain the distance matrix between residue pairs.
[0094] Specifically, the normalized Cα atom coordinates of each residue are extracted from the main chain coordinate field of the residue index table (Cα atoms are main chain atoms common to all amino acid residues, with stable coordinates and located at the center of the residue structure, effectively reflecting the spatial position of the residue). The distance between each pair of residues is calculated using the Euclidean distance formula. Taking advantage of the symmetry of the distance values (the distance between residue i and j is equal to the distance between j and i), the distances of all residue pairs are arranged according to the residue index to form an N×N symmetric distance matrix (N is the total number of residues), where the matrix element [i,j] is the distance between residue i and j, and the diagonal elements are 0 (the distance between a residue and itself is 0).
[0095] In step S13, please refer to Figure 1 In step S13, the degrees of freedom of residue motion are obtained through adaptive weighted calculation and correlation correction based on the conformation sampling data in the static structure file and the residue index table.
[0096] As an example, step S13 may include the following steps: S131 to S134.
[0097] S131: Perform principal component analysis on the conformational sampling data in the static structure file to obtain the motion components of the residues.
[0098] S132: Normalize the temperature factor of the residues in the residue index table.
[0099] S133: Calculate the mixed motion degrees of freedom by using an adaptive weighting strategy to combine the motion components and the temperature factor of the normalized residues.
[0100] S134: The hybrid motion degrees of freedom are corrected for correlation to obtain the residue motion degrees of freedom.
[0101] Specifically, the Cα atom coordinates of all residues can be extracted from the conformational sampling data associated with the static structure file (each conformation corresponds to an N×3 coordinate matrix, where N is the total number of residues), and used to construct the conformational space coordinate matrix X∈R. 3N×M : Where N is the total number of residues, M is the number of conformational samples, and x k Let be the expanded vector of the Cα coordinates of all residues in the k-th conformation; then calculate the covariance matrix, which can be expressed as: The eigenvalue decomposition of the covariance matrix can be expressed as: C = UΛU T , where U is an orthogonal matrix composed of eigenvectors and Λ is a diagonal matrix; calculating the covariance matrix and performing eigenvalue decomposition can extract a few principal components from complex data, which can be used to summarize the main trends of change in the original data.
[0102] Specifically, the first three principal components (PC1-PC3) can be taken to calculate the motion component of residue i. The expression can be:
[0103]
[0104] Among them, u 3i-2,k u 3i-1,k and u 3i,k These are the x, y, and z coordinate components corresponding to the i-th residue in the eigenvector matrix, respectively; k is the summation index, with values of 1, 2, and 3 (selecting the first 3 principal components).
[0105] As an example, the normalization expression for the temperature factor of the residues in the residue index table can be:
[0106]
[0107] in, B is the temperature factor value obtained after normalizing residue i; i For the temperature factor of residue i in the residue index table; μ B σ is the average temperature factor of all residues in the protein; B This represents the standard deviation of the temperature factor of all residues in the protein.
[0108] Specifically, the average temperature factor of each residue is obtained from the temperature factor field of the residue index table (the temperature factors of all atoms in a residue are taken as an arithmetic average). Normalizing the average temperature factor can eliminate the absolute numerical differences between different crystal structures.
[0109] As an example, the mixed motion degrees of freedom d are calculated using an adaptive weighting strategy by combining the motion components and the temperature factor of the normalized residues. i The expression can be:
[0110]
[0111] Where α is the weighting coefficient; The solvent-accessible surface area (SASA) of residues is dynamically adjusted. Exposed residues with a surface area value >0.5 are more dependent on the PCA component, while buried residues are more dependent on the temperature factor. i Let be the solvent-accessible surface area of residue i.
[0112] As an example, the hybrid motion degrees of freedom are corrected for correlation using a Markov random field (MRF) model to obtain the residue motion degrees of freedom, and the expression can be:
[0113]
[0114] in, denoted as denoted as the final degree of freedom of residue i; Z is the normalization constant; exp is the natural exponential function; β is the weighting coefficient of the mixed degree of freedom of residue i itself; γ is the weighting coefficient of the motion information of neighboring residues. w is the set of neighboring residues of residue i; ij d represents the coupling weight between residue j and residue i; j For residue j, represent the mixed motion degrees of freedom without correlation correction.
[0115] In step S14, please refer to Figure 1 In step S14, the dynamic distance threshold of the residue pair is obtained based on the degrees of freedom of the residue motion; the distance matrix is converted into a binary contact map based on the dynamic distance threshold.
[0116] As an example, the expression for calculating the dynamic distance threshold can be:
[0117]
[0118] Among them, R ij d is the dynamic distance threshold for residue pairs (i, j); i d represents the degrees of freedom for the motion of residue i; j is the degree of freedom of motion of residue j; 0.2 and 0.8 are fixed coefficients.
[0119] Specifically, if the degrees of freedom of movement of residues in residue pair (i,j) are both large (flexible residues): the spatial position fluctuates drastically, and even if the instantaneous distance is slightly far, contact may be formed due to overlapping movements. Therefore, the dynamic distance threshold needs to be relaxed (take a larger value); if the degrees of freedom of movement of residues in residue pair (i,j) are both small (rigid residues): the position is stable, and stable contact is only formed when the distance is close enough. Therefore, the dynamic distance threshold needs to be tightened (take a smaller value).
[0120] As an example, the distance matrix is converted into a binary contact map based on the following formula:
[0121]
[0122] Among them, CM ij D represents the element corresponding to residue i and residue j in the binary contact diagram, with a value of 1 or 0; ij R represents the Cα atom distance between each pair of residues i and j; ij is the dynamic distance threshold between residue i and residue j.
[0123] Specifically, the symmetry of the distance matrix (D) can be utilized. ij =D jiOnly the upper triangular (or lower triangular) region needs to be calculated, avoiding redundant calculations; then the distance D between each residue pair (i,j) in the distance matrix is compared. ij With dynamic distance threshold R ij If D ij ≤R ij If it is determined to be dynamic contact, the corresponding position in the binary contact diagram is marked as 1; if D ij >R ij If the condition is determined to be non-contact, the corresponding position is recorded as 0.
[0124] As an example, graph convolutional networks (GCNs) can be used to optimize contact graphs, including:
[0125] The graph Laplacian matrix L can be constructed using the following expression:
[0126] L = D - CM
[0127] Where D is the degree matrix, D ii =∑ j CM ij CM ij The connection weight between residue i and residue j; ∑ j CM ij Sum all elements in the i-th row of the contact matrix CM; CM is the adjacency matrix.
[0128] The expression for updating the contact map via spectral convolution can be:
[0129] CM (t+1) =ReLU(CM) (t) ·W0+L·CM (t) ·W1)
[0130] Among them, CM (t) Let CM be the contact matrix at step t; (t+1) t is the contact matrix updated at step t+1; ReLU is the rectified linear unit, an activation function; W0 and W1 are trainable parameters, and t is the number of iterations.
[0131] In step S15, please refer to Figure 1 In step S15, the contact frequency of residue pairs is obtained based on the binary contact map; and a contact frequency matrix is constructed based on the contact frequency.
[0132] As an example, obtaining the contact frequency of residue pairs based on the binary contact map includes:
[0133] The basic contact frequency is calculated based on the binary contact diagram.
[0134] The base contact frequency is Gaussian smoothed to obtain the desired contact frequency.
[0135] As an example, under a single static structure, the expression for calculating the foundation contact frequency can be:
[0136]
[0137] in, Let (i, j) be the fundamental contact frequency of residue pair (i, j).
[0138] As an example, the Gaussian smoothing of the base contact frequency can be expressed as:
[0139]
[0140] Among them, f ij CM represents the contact frequency of residue pairs (i, j) after Gaussian smoothing (i.e., the locally smoothed contact frequency of a single structure); N represents the total number of residues; kl σ represents the contact state of residue pair (k, l); (i, j) represents the target residue pair whose smooth contact frequency is to be calculated; (k, l) represents the neighboring residue pair of (i, j); σ represents the standard deviation of the Gaussian kernel; Gaussian smoothing kernel processing, through its local weighting properties, accurately integrates the inherent local correlation information in protein contact patterns, enhances the consistency of contact patterns in local regions, and can effectively capture the local correlation of contact patterns.
[0141] As an example, the contact frequency matrix can be constructed using the following expression:
[0142]
[0143] Among them, f ij ′ represents the multiconformation statistical contact frequency. W represents the window size. This represents the contact state of the k-th neighboring conformation.
[0144] In step S16, please refer to Figure 1 In step S16, the degrees of freedom of the residue motion are extended into a symmetric matrix.
[0145] As an example, extending the residue motion degrees of freedom into a motion degree-of-freedom matrix, the expression can be:
[0146]
[0147] Among them, D′ ij d represents the degrees of freedom of residue i and residue j in the motion degree-of-freedom matrix; i d represents the degree of freedom of residue i after correlation correction; j σ represents the degree of freedom of residue j after correlation correction; σ is the Euclidean distance between the spatial coordinates of residues i and j; and σ is the standard deviation of the Gaussian kernel.
[0148] In step S17, please refer to Figure 1 In step S17, the contact frequency matrix and the symmetric matrix are fused to obtain the initial dynamic probability matrix.
[0149] As an example, step S17 may include the following steps: S171 to S172.
[0150] S171: The contact frequency matrix and the symmetric matrix are fused by the Hadamard product.
[0151] S172: Perform a logarithmic transformation on the fusion result to obtain the initial dynamic probability matrix.
[0152] As an example, by fusing F and D through the Hadamard product, the expression can be:
[0153]
[0154] Among them, F ij D represents the contact frequency of residue pair (i, j); ij Let be the Cα atomic distance between each pair of residues i and j.
[0155] As an example, the initial dynamic probability matrix is obtained by performing a logarithmic transformation on the fusion result, and the expression can be:
[0156]
[0157] in, This is the initial dynamic probability matrix.
[0158] In step S18, please refer to Figure 1 In step S18, the initial probability matrix is normalized and probability propagated sequentially to obtain the dynamic probability matrix.
[0159] As an example, step S18 may include the following steps: S181 to S182.
[0160] S181: The initial probability matrix is normalized using a dual normalization strategy.
[0161] S182: The PageRank algorithm is used to propagate the normalized initial probability matrix to obtain the dynamic probability matrix.
[0162] As an example, we can sum all elements from column 1 to column N in row i. The row normalization expression can be:
[0163]
[0164] in, The result (matrix element values) after normalizing residue pairs (i,j); To sum all elements in the i-th row of the matrix; is the matrix element; N is the total number of residues, i.e., the upper limit of the cumulative sum of column index k.
[0165] The expression for column normalization can be:
[0166]
[0167] in, The result of column normalization (matrix element values) for residue pairs (i,j); To sum all elements in the j-th column of the matrix; is the row-normalized value of residue pair (k, j).
[0168] As an example, the expression for the normalized initial probability matrix can be:
[0169]
[0170] As an example, the normalized initial probability matrix is propagated based on the following formula:
[0171]
[0172] in, P represents the interaction probability of residue pair (i,j) after probability propagation; ij λ represents the original probability of residue pair (i,j); λ is the retention coefficient; P ik P represents the probability of residue pair (i,k); kj The probability of residue pair (k,j) is given. This breaks through the limitation that the initial probability only reflects "direct contact" and incorporates information about indirect interaction chains in the residue network (such as i→k→j) into the probability value, which is more in line with the global correlation of residue interactions in proteins.
[0173] In step S19, please refer to Figure 1 In step S19, the protein binding conformation is predicted based on the dynamic probability matrix.
[0174] Specifically, the elements in the dynamic probability matrix quantify the probability of residue pairs (i, j) being in close contact in the bound state (the higher the probability, the more likely residues i and j are to be adjacent in the bound conformation); the probability of close contact is converted into spatial distance constraints (high probability corresponds to close contact, low probability corresponds to loose constraints) to assist in structural modeling, and finally generate protein binding conformations that conform to the residue contact rules.
[0175] In another embodiment, this application also provides a protein binding conformation prediction system, which may include a data acquisition module 10, a data preprocessing module 20, a feature extraction module 30, a dynamic probability matrix generation module 40, and a prediction module 50; wherein the data acquisition module 10 is used to acquire a static structure file of the target protein; the data preprocessing module 20 is used to construct a residue index table based on the atomic information extracted from the static structure file; perform normalization processing on the atomic coordinates in the residue index table to obtain normalized atomic coordinates; the feature extraction module 30 is used to calculate the distance matrix between residue pairs based on the normalized atomic coordinates; and perform prediction based on the conformation sampling data in the static structure file and the... The residue index table obtains the residue motion degrees of freedom through adaptive weighted calculation and correlation correction; based on the residue motion degrees of freedom, a dynamic distance threshold for residue pairs is obtained; based on the dynamic distance threshold, the distance matrix is converted into a binary contact map; based on the binary contact map, the contact frequency of residue pairs is obtained; the dynamic probability matrix generation module 40 is used to construct a contact frequency matrix based on the contact frequency; the residue motion degrees of freedom are extended into a symmetric matrix; the contact frequency matrix and the symmetric matrix are fused to obtain an initial dynamic probability matrix; the initial probability matrix is then normalized and subjected to probability propagation to obtain a dynamic probability matrix; the prediction module 50 is used to predict the protein binding conformation based on the dynamic probability matrix.
[0176] The protein binding conformation prediction system of this application can construct a dynamic probability matrix by extracting the joint probability distribution of residue contact frequency and motion degree of freedom from the static structure through the data acquisition module 10, data preprocessing module 20, feature extraction module 30, dynamic probability matrix generation module 40 and prediction module 50. This indirectly infers the possibility of conformational changes during the binding process, improves the accuracy of binding conformation prediction, and avoids the high computational cost of directly using molecular dynamics simulation, effectively improving the overall operating efficiency and performance of the system.
[0177] As an example, the protein binding conformation prediction system of this application can be used to perform, for example... Figure 1 And the protein binding conformation prediction method in the relevant embodiments.
[0178] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features of 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.
[0179] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
Claims
1. A method of predicting a protein binding conformation, characterized by, The method comprises the following steps: obtaining a static structure file of a target protein; constructing a residue index table based on atomic information extracted from the static structure file; performing standardization processing on atomic coordinates in the residue index table to obtain standardized atomic coordinates; calculating a distance matrix between residue pairs based on the standardized atomic coordinates; calculating a residue motion degree of freedom based on conformational sampling data in the static structure file and the residue index table through adaptive weighting calculation and correlation correction; obtaining a dynamic distance threshold of a residue pair based on the residue motion degree of freedom; converting the distance matrix into a binary contact map based on the dynamic distance threshold; obtaining a contact frequency of a residue pair based on the binary contact map; constructing a contact frequency matrix based on the contact frequency; extending the residue motion degree of freedom into a symmetric matrix; fusing the contact frequency matrix and the symmetric matrix to obtain an initial dynamic probability matrix; performing normalization processing and probability propagation on the initial probability matrix in sequence to obtain a dynamic probability matrix; predicting a protein binding conformation based on the dynamic probability matrix.
2. The method of claim 1, wherein, The method for constructing a residue index table based on atomic information extracted from a static structure file comprises the following steps: parsing the static structure file to obtain information of all atoms; the information comprises atomic name, chain identifier, residue sequence number, insertion code, residue name, coordinates and temperature factor of each atom; grouping all atoms based on the residue identifier to obtain an atomic set of each residue by taking the chain identifier, residue sequence number and insertion code as a residue identifier; separating the atomic set of each residue into a main chain and a side chain to obtain main chain coordinates and side chain coordinates; extracting residue-level attributes of each group of atoms; the residue-level attributes comprise chain identifier, residue sequence number, insertion code, residue name and temperature factor; constructing a residue index table based on the residue identifier, main chain coordinates, side chain coordinates and residue-level attributes.
3. The method of claim 1, wherein, The method for calculating a distance matrix between residue pairs based on standardized atomic coordinates comprises the following steps: The distance D for each pair of residues is calculated based on the following formula ij comprising: where D ij is the C alpha atom distance for each pair of residues i and j; (x i ,y i ,z i ) is the C alpha atom coordinates of residue i; (x j ,y j ,z j ) is the C alpha atom coordinates of residue j; D = D (R) = distance between all pairs of residues ij D = D (R) = distance between all pairs of residues 4. The method of claim 1, wherein, The method for calculating a residue motion degree of freedom based on conformational sampling data in the static structure file and the residue index table through adaptive weighting calculation and correlation correction comprises the following steps: performing principal component analysis on the conformational sampling data in the static structure file to obtain motion components of residues; performing normalization processing on temperature factors of residues in the residue index table; calculating a hybrid motion degree of freedom through an adaptive weighting strategy based on the motion components and the normalized temperature factors of residues; performing correlation correction on the hybrid motion degree of freedom to obtain a residue motion degree of freedom.
5. The method of claim 1, wherein, The method for obtaining a dynamic distance threshold of a residue pair based on the residue motion degree of freedom comprises the following steps: The method for converting the distance matrix into a binary contact map based on the dynamic distance threshold comprises the following steps: The dynamic distance threshold R is obtained based on the following formula ij : wherein R ij is the dynamic distance threshold for residue pair (i, j); d i is the degree of freedom of motion of residue i; d j is the degree of freedom of motion of residue j; and 0.2 and 0.8 are fixed coefficients. The method for converting the distance matrix into a binary contact map based on the following formula comprises the following steps: wherein CM ij is the element corresponding to residue i and residue j in the binary contact map, taking values of 1 or 0; D ij is the distance between the Cα atoms of each pair of residues i and j; R ij is the dynamic distance threshold between residue i and residue j.
6. The method of claim 1, wherein, The method for obtaining a contact frequency of a residue pair based on the binary contact map comprises the following steps: calculating a basic contact frequency based on the binary contact map; performing Gaussian smoothing processing on the basic contact frequency to obtain a required contact frequency.
7. The method of claim 1, wherein, The method for fusing the contact frequency matrix and the symmetric matrix to obtain an initial dynamic probability matrix comprises the following steps: fuse the contact frequency matrix and the symmetric matrix by Hadamard product; logarithmically transform the fusion result to obtain an initial dynamic probability matrix.
8. The method of claim 1, wherein, normalize and propagate probability of the initial probability matrix in sequence to obtain a dynamic probability matrix, including: normalizing the initial probability matrix by a double normalization strategy; propagating probability of the normalized initial probability matrix by a PageRank algorithm to obtain a dynamic probability matrix.
9. The method of claim 8, wherein, propagating probability of the normalized initial probability matrix by a PageRank algorithm to obtain a dynamic probability matrix based on the following formula: wherein, is the interaction probability of residue pair (i, j) after probability propagation; P ij is the original probability of residue pair (i, j); λ is a retention coefficient; P ik is the probability of residue pair (i, k); P kj is the probability of residue pair (k, j).
10. A protein binding conformation prediction system, characterized by, including: a data acquisition module configured to acquire a static structure file of a target protein; a data preprocessing module configured to construct a residue index table based on atomic information extracted from the static structure file, and perform standardization processing on atomic coordinates in the residue index table to obtain standardized atomic coordinates; a feature extraction module configured to calculate a distance matrix between residue pairs based on the standardized atomic coordinates, obtain a residue motion degree of freedom by adaptive weighting calculation and correlation correction based on conformational sampling data in the static structure file and the residue index table, and obtain a dynamic distance threshold of residue pairs based on the residue motion degree of freedom; convert the distance matrix into a binary contact map based on the dynamic distance threshold, and obtain a contact frequency of residue pairs based on the binary contact map; a dynamic probability matrix generation module configured to construct a contact frequency matrix based on the contact frequency, extend the residue motion degree of freedom into a symmetric matrix, fuse the contact frequency matrix and the symmetric matrix to obtain an initial dynamic probability matrix, and normalize and propagate probability of the initial probability matrix in sequence to obtain a dynamic probability matrix; a prediction module configured to predict a protein binding conformation based on the dynamic probability matrix.