Training method and device of total atomic structure prediction model and electronic equipment
By encoding and decoding the structural information and dynamic trajectories of biological molecules, and using differential training of the whole-atom structure prediction model, the shortcomings of deep learning in the prediction of biological molecules' dynamic trajectories are solved, and prediction accuracy and model generalization capabilities are improved.
Patent Information
- Application Number
- CN202411331123.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2024-12-31
AI Technical Summary
Existing deep learning methods are difficult to accurately predict the full atomic structure of biological molecules, especially in terms of dynamic trajectory prediction, which affects the generalization ability and accuracy of the model.
By obtaining the structural information and dynamic trajectories of biological molecules, adding noise, encoding and decoding, using differential training of the full-atom structure prediction model, improving the model's dynamic trajectory prediction ability.
It improves the accuracy of prediction of dynamic trajectory of biomolecules and the generalization ability of models, and enhances the understanding and prediction of the dynamic structure of biomolecules.
Smart Images
Figure CN119229950A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, in particular to artificial intelligence fields such as deep learning and biological computing, and specifically to a training method, device and electronic equipment for a full-atom structure prediction model. Background Art
[0002] With the development of artificial intelligence, the use of artificial intelligence technology to predict the all-atom structure of biological molecules has received more attention than experimental methods. How to use deep learning methods to promote the accurate prediction of the all-atom structure of biological molecules has become increasingly important. Summary of the invention
[0003] The present application provides a training method, device and electronic equipment for an all-atom structure prediction model.
[0004] According to one aspect of the present application, a training method for an all-atom structure prediction model is provided, comprising:
[0005] Acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule; wherein the first dynamic trajectory includes position information of atoms in the biomolecule at different time points;
[0006] adding noise to the first dynamic trajectory to obtain a second dynamic trajectory;
[0007] Encoding the structural information to obtain encoding features;
[0008] Decoding the coding feature and the second dynamic trajectory to obtain a target dynamic trajectory;
[0009] According to the difference between the target dynamic trajectory and the first dynamic trajectory, the initial all-atom structure prediction model is trained to obtain the all-atom structure prediction model.
[0010] According to another aspect of the present application, a method for predicting an all-atom structure is provided, comprising:
[0011] Obtain structural information of biomolecules;
[0012] The structural information is encoded by an all-atom structure prediction model to obtain encoding features; wherein the all-atom structure prediction model can be trained using the method described above;
[0013] The all-atom structure prediction model is used to decode the coding features and the dynamic trajectory noise to obtain the dynamic trajectory of the biological molecule; wherein the dynamic trajectory includes the position information of atoms in the biological molecule at different time points.
[0014] According to another aspect of the present application, a training device for an all-atom structure prediction model is provided, comprising:
[0015] An acquisition module, used to acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule; wherein the first dynamic trajectory includes position information of atoms in the biomolecule at different time points;
[0016] A noise adding module, used for adding noise to the first dynamic trajectory to obtain a second dynamic trajectory;
[0017] An encoding module, used for encoding the structural information to obtain encoding features;
[0018] A decoding module, used for decoding the coding feature and the second dynamic trajectory to obtain a target dynamic trajectory;
[0019] A training module is used to train an initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain a all-atom structure prediction model.
[0020] According to another aspect of the present application, a full-atom structure prediction device is provided, comprising:
[0021] An acquisition module, used to obtain the structural information of biomolecules;
[0022] An encoding module, used for inputting the structural information into an all-atom structure prediction model, encoding the structural information, and obtaining encoding features; wherein the all-atom structure prediction model can be trained using the above-mentioned device;
[0023] A decoding module is used to decode the encoding features and dynamic trajectory noise through the all-atom structure prediction model to obtain the dynamic trajectory of the biological molecule; wherein the dynamic trajectory includes the position information of atoms in the biological molecule at different time points.
[0024] According to another aspect of the present application, there is provided an electronic device, including:
[0025] at least one processor; and
[0026] a memory communicatively connected to the at least one processor; wherein,
[0027] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can perform the method described in the above embodiment.
[0028] According to another aspect of the present application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method according to the above embodiment.
[0029] According to another aspect of the present application, a computer program product is provided, including a computer program, wherein the computer program implements the steps of the method described in the above embodiment when executed by a processor.
[0030] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present application.
[0032] Figure 1 A schematic diagram of a flow chart of a method for training an all-atom structure prediction model provided in one embodiment of the present application;
[0033] Figure 2 A schematic diagram of a flow chart of a method for training an all-atom structure prediction model provided in another embodiment of the present application;
[0034] Figure 3 A schematic diagram of a flow chart of a method for training an all-atom structure prediction model provided in another embodiment of the present application;
[0035] Figure 4 A schematic diagram of a process of dynamic trajectory prediction using an all-atom structure prediction model provided in an embodiment of the present application;
[0036] Figure 5 A schematic diagram of a flow chart of an all-atom structure prediction method provided in one embodiment of the present application;
[0037] Figure 6 A schematic diagram of the structure of a training device for an all-atom structure prediction model provided in one embodiment of the present application;
[0038] Figure 7 A schematic diagram of the structure of an all-atom structure prediction device provided in one embodiment of the present application;
[0039] Figure 8 A block diagram of an electronic device used to implement the training method of the all-atom structure prediction model of an embodiment of the present application. DETAILED DESCRIPTION
[0040] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.
[0041] The following describes the training method, device, electronic device and storage medium of the all-atom structure prediction model of the embodiments of the present application with reference to the accompanying drawings.
[0042] Figure 1 A schematic flow chart of a method for training an all-atom structure prediction model provided in one embodiment of the present application.
[0043] The training method of the all-atom structure prediction model of the embodiment of the present application can be executed by the training device of the all-atom structure prediction model of the embodiment of the present application, and the device can be configured in an electronic device.
[0044] Among them, the electronic device can be any device with computing capabilities, such as a personal computer, a mobile terminal, a server, etc. The mobile terminal can be, for example, a vehicle-mounted device, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.
[0045] like Figure 1 As shown, the training method of the all-atom structure prediction model includes:
[0046] Step 101, obtaining structural information of a biomolecule and a first dynamic trajectory of the biomolecule.
[0047] In the present application, biological molecules can be any one of proteins, small molecules, RNA (Ribonucleic Acid), DNA (Deoxyribonucleic Acid), and ions. Biomolecules can also be complexes of at least two of proteins, small molecules, RNA, DNA, and ions, etc., without limitation.
[0048] The structural information of a biomolecule may refer to sequence information, molecular formula, etc. For example, if a biomolecule is a protein, RNA, DNA, etc., the structural information may refer to sequence information. For another example, if a biomolecule is a complex of a protein and a molecule, the structural information may include the sequence information of the protein and the sequence information of the DNA. For another example, if a biomolecule is a complex of a protein and a small molecule, the structural information may include the sequence information of the protein and the molecular formula of the small molecule.
[0049] The first dynamic trajectory of the biomolecule may include position information of atoms in the biomolecule at different time points. The position information here may refer to three-dimensional coordinate information. Since the three-dimensional structure of the biomolecule is determined by the position of atoms in the biomolecule, the first dynamic trajectory can be used to characterize the three-dimensional structure of the biomolecule at different time points.
[0050] For example, the shape of the first dynamic trajectory can be F1*S1*3, where F1 represents the number of frames, S1 represents the number of atoms in the biological molecule, and 3 represents the three dimensions in the three-dimensional coordinate information. It can be considered that one frame corresponds to one time point, and the first dynamic trajectory includes the position information of atoms in the biological molecule at F1 time points, that is, includes the three-dimensional structure of the biological molecule at F1 time points.
[0051] Among them, the structural information of the biological molecule corresponds to the first dynamic trajectory. Taking the sequence information of the protein as an example, the sequence information of the protein refers to the arrangement order of the amino acids in the protein molecule. The atoms in each frame of the first dynamic trajectory are arranged according to the arrangement order of the amino acids, that is, the atoms in each amino acid are sorted in sequence according to the arrangement order of the amino acids. It should be noted that the arrangement order of atoms at different time points remains unchanged, but the position information of the atoms may change.
[0052] Illustratively, the first dynamic trajectory may be based on the three-dimensional structure of the biomolecule at different time points obtained during the experimental phase, and may be obtained by other methods, which is not limited to this.
[0053] Step 102: adding noise to the first dynamic trajectory to obtain a second dynamic trajectory.
[0054] In the present application, adding noise to the first dynamic trajectory can obtain a noisy dynamic trajectory, that is, the second dynamic trajectory. Exemplarily, the position information of atoms in biological molecules at some time points can be selected from multiple time points to add noise, or the position information of some atoms at different time points can be selected from the first dynamic trajectory to add noise.
[0055] The noise used may be Gaussian noise, random noise, or other types of noise.
[0056] Step 103: Encode the structural information to obtain encoding features.
[0057] In the present application, the initial all-atom structure prediction model may include an encoder and a decoder, and the structural information of the biological molecule may be input into the encoder for encoding to obtain encoding features.
[0058] Step 104 , decoding the coding feature and the second dynamic trajectory to obtain the target dynamic trajectory.
[0059] In the present application, the encoding feature and the second dynamic trajectory may be input into a decoder in the initial all-atom structure prediction model, and the encoding feature and the second dynamic trajectory may be decoded by the decoder to obtain a target dynamic trajectory.
[0060] The target dynamic trajectory may include the predicted position information of atoms in biological molecules at different time points.
[0061] Step 105 , training the initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain the all-atom structure prediction model.
[0062] In the present application, the parameters of the initial all-atom structure prediction model can be adjusted according to the position information of atoms in the biomolecule at each time point in the target dynamic trajectory and the difference between the position information of atoms in the biomolecule at the same time point in the first dynamic trajectory to obtain the all-atom structure prediction model after parameter adjustment. If the training end condition is not met, the all-atom structure prediction model after parameter adjustment can be continuously trained until the training end condition is met to obtain the all-atom structure prediction model. Among them, the all-atom structure prediction model can be used to predict the dynamic trajectory of the biomolecule based on the structural information of the biomolecule.
[0063] It is understandable that the position information of atoms in a biological molecule at a certain point in time can also be used to train a full-atom structure prediction model that can predict the position information of atoms in a biological molecule at a certain point in time, and a full-atom structure prediction model can also be obtained for predicting the static conformation of a biological molecule.
[0064] In the embodiment of the present application, the structural information of the biomolecule is encoded to obtain a coding feature, and the target dynamic trajectory is obtained by decoding the coding feature and the second dynamic trajectory obtained by adding noise to the first dynamic trajectory, and then the all-atom structure prediction model is obtained through training based on the difference between the target dynamic trajectory and the first dynamic trajectory. Thus, based on the structural information of the biomolecule and the first dynamic trajectory of the biomolecule, a full-atom structure prediction model that can predict the dynamic trajectory of the biomolecule is obtained through training, and the prediction of the dynamic trajectory of the biomolecule is realized, which improves the generalization ability of the model.
[0065] Figure 2 A schematic flow chart of a method for training an all-atom structure prediction model provided in another embodiment of the present application.
[0066] like Figure 2 As shown, the training method of the all-atom structure prediction model includes:
[0067] Step 201 : Acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule.
[0068] In the present application, step 201 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0069] Step 202: adding noise to the first dynamic trajectory to obtain a second dynamic trajectory.
[0070] In the present application, step 202 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0071] Step 203: Encode the structural information to obtain encoding features.
[0072] In the present application, step 203 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0073] Step 204: Perform dimensionality reduction processing on the second dynamic trajectory to obtain a third dynamic trajectory.
[0074] In the present application, the position information of a plurality of adjacent atoms in the second dynamic trajectory may be fused to reduce the dimension of the second dynamic trajectory to obtain a third dynamic trajectory.
[0075] Exemplarily, the second dynamic trajectory may be cut into blocks to obtain a third dynamic trajectory. The third dynamic trajectory may include multiple first sub-block trajectories, each of which may be regarded as an element of the third dynamic trajectory, and each of which may be obtained by fusing position information of multiple adjacent atoms in the second dynamic trajectory, wherein the fusion may be calculated by averaging, or weighted averaging, or randomly selecting the position information of any atom from the position information of multiple adjacent atoms.
[0076] For example, the shape of the dynamic trajectory can be expressed as F*S*3, where F represents the number of frames, S represents the number of atoms in the biological molecule, and 3 represents the three dimensions in the three-dimensional coordinate information. The number of frames of the first dynamic trajectory is 4, and the number of atoms is 6, which can be regarded as a 4*6 matrix. The matrix can be cut into small blocks of 2*2 to obtain a 2*3 matrix, that is, the number of frames of the second dynamic trajectory is 2, and the number of atoms is 3. There are 6 first sub-block trajectories in total, thereby reducing the shape of the first dynamic trajectory from 4*6*3 to 2*3*3.
[0077] Step 205 , decoding the coded feature and the third dynamic trajectory to obtain the target dynamic trajectory.
[0078] In the present application, the coded features and the third dynamic trajectory may be decoded to obtain an intermediate dynamic trajectory, and then the intermediate dynamic trajectory may be interpolated to obtain a target dynamic trajectory.
[0079] Exemplarily, if the decoder of the model processes one-dimensional data, the third dynamic trajectory can be converted into a one-dimensional vector and then input into the decoder together with the encoded features for decoding to obtain a new one-dimensional vector. The new one-dimensional vector is then rearranged to obtain an intermediate dynamic trajectory, and the intermediate dynamic trajectory is then interpolated to obtain the target dynamic trajectory.
[0080] Exemplarily, if the first dynamic trajectory includes multiple first sub-block trajectories, the multiple first sub-block trajectories can be sorted in a preset order to obtain a first trajectory sequence, and then the coded features and the first trajectory sequence are decoded to obtain the target dynamic trajectory. Thus, by sorting the multiple first sub-block trajectories in a preset order, flattening the third dynamic trajectory into a one-dimensional first trajectory sequence and then decoding, the processing requirements of the model can be met.
[0081] The preset order may be from left to right and from top to bottom, and the length of the first track sequence is the same as the number of the first sub-block tracks.
[0082] Exemplarily, when decoding the coding feature and the first trajectory sequence, the coding feature and the first trajectory sequence can be decoded to obtain a second trajectory sequence, and multiple second sub-block trajectories in the second trajectory sequence are rearranged to obtain a fourth dynamic trajectory, and then the fourth dynamic trajectory is interpolated to obtain a target dynamic trajectory.
[0083] The second trajectory sequence is one-dimensional data, and the length of the second trajectory sequence is the same as the number of the second sub-block trajectories.
[0084] Therefore, by decoding the coding features and the first trajectory sequence, a one-dimensional second trajectory sequence is obtained, and then the fourth dynamic trajectory is obtained by rearrangement, and then the target dynamic trajectory is obtained by interpolation processing. While reducing the amount of model calculation, the accuracy of dynamic trajectory prediction can be improved.
[0085] Step 206 , training the initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain the all-atom structure prediction model.
[0086] Exemplarily, the first loss corresponding to any time point is determined based on the difference between the position information of the atom at any time point in the target dynamic trajectory and the position information of the same atom at any time point in the first dynamic trajectory, and the second loss is obtained based on the sum of the first losses corresponding to different time points. Then, based on the second loss, the initial all-atom structure prediction model is trained to obtain the all-atom structure prediction model.
[0087] Among them, the first loss corresponding to any time point can be the sum of the sub-losses corresponding to each atom, and the sub-loss corresponding to each atom can be determined based on the difference between the position information of each atom in the target dynamic trajectory and the position information of the same atom in the first dynamic trajectory.
[0088] Therefore, by training the initial all-atom structure prediction model according to the difference between the position information of atoms at each time point in the target dynamic trajectory and the position information of the same atoms at each time point in the first dynamic trajectory, a all-atom structure prediction model is obtained, thereby improving the accuracy of the model.
[0089] In the embodiment of the present application, the second dynamic trajectory is subjected to dimensionality reduction processing to obtain the third dynamic trajectory, and the coded features and the third dynamic trajectory are decoded to obtain the target dynamic trajectory. Thus, the second dynamic trajectory is subjected to dimensionality reduction before being used for decoding, which can reduce the amount of calculation and improve the model processing speed.
[0090] Figure 3 A schematic flow chart of a method for training an all-atom structure prediction model provided in another embodiment of the present application.
[0091] like Figure 3 As shown, the training method of the all-atom structure prediction model includes:
[0092] Step 301, obtaining structural information of biological molecules.
[0093] In the present application, step 301 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0094] Step 302 , generating multiple static conformations of the biomolecules by calculating the interaction energy between the biomolecules based on the structural information of the biomolecules.
[0095] In this application, physics-based docking software is used to generate multiple static conformations of biomolecules by calculating the interaction energy between molecules based on the structural information of biomolecules. These static conformations cover different binding modes and orientations, greatly enriching the diversity of data.
[0096] For example, if a biological molecule is a complex of a protein and a small molecule, docking software can be used to predict and generate multiple static conformations of the complex of the protein and the small molecule by calculating the interaction energy between molecules based on the sequence information of the protein and the small molecule information.
[0097] Step 303 , simulating multiple static conformations to capture a first dynamic trajectory of the biomolecule.
[0098] In this application, molecular dynamics simulation software can be used to simulate multiple static conformations for a long time to capture the first dynamic trajectory of the biomolecule. The first dynamic trajectory not only demonstrates the flexibility of the biomolecule, but also provides rich dynamic interaction information, which helps the model learn more realistic biomolecular behavior.
[0099] For example, multiple static conformations of a complex of multiple biomolecules can be simulated for a long time to capture the dynamic behaviors of the complex of biomolecules in a solvent environment, such as conformational changes, dynamic trajectories, etc.
[0100] Step 304: adding noise to the first dynamic trajectory to obtain a second dynamic trajectory.
[0101] In the present application, step 304 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0102] Step 305: Encode the structural information to obtain encoding features.
[0103] In the present application, step 305 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0104] Step 306: decode the encoded feature and the second dynamic trajectory to obtain the target dynamic trajectory.
[0105] In the present application, step 306 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0106] Exemplarily, the position information of each atom in the second dynamic trajectory is sorted in a preset order to obtain a third trajectory sequence, and the coded feature and the third trajectory sequence are decoded to obtain a fourth trajectory sequence, and then the position information of the atoms in the fourth trajectory sequence is rearranged to obtain the target dynamic trajectory. Thus, converting the second dynamic trajectory into a one-dimensional third trajectory sequence and then decoding it can meet the processing requirements of the model.
[0107] Step 307 , training the initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain the all-atom structure prediction model.
[0108] In the present application, step 307 can be implemented in any of the embodiments of the present application, so it will not be described in detail here.
[0109] In an embodiment of the present application, multiple static conformations of the biological molecules are generated by calculating the interaction energy between biological molecules. Since the static conformations cover different binding modes and orientations, the diversity of the data is greatly enriched. The first dynamic trajectory obtained based on multiple static conformations not only demonstrates the flexibility of the biological molecules, but also provides rich dynamic interaction information, which helps the model learn more realistic biological molecular behaviors. Therefore, the all-atom structure prediction model is trained based on the first dynamic trajectory. The model can capture time dependence and complex interactions from the first dynamic trajectory, thereby improving the accuracy of structure prediction and achieving high-precision all-atom structure prediction.
[0110] For ease of understanding, the following Figure 4 To explain, Figure 4 A schematic diagram of the process of dynamic trajectory prediction using an all-atom structure prediction model provided in an embodiment of the present application.
[0111] like Figure 4 As shown, the all-atom structure prediction model includes an encoder and a decoder. Taking the complex of biological molecules such as protein, DNA, and RNA as an example, the encoder can be used to encode the sequence information of protein, DNA, RNA, etc. to obtain coding features, and the coding features are input into the decoder.
[0112] In addition, the dynamic trajectory samples of the complex are denoised to obtain the noisy dynamic trajectory, and the noisy dynamic trajectory of the complex is cut into pieces to reduce the dimension of the noisy dynamic trajectory, and then flattened into a one-dimensional vector and input into the decoder. The decoder decodes the encoded features and the one-dimensional vector to obtain a new one-dimensional vector, and rearranges the new one-dimensional vector, and then interpolates to obtain the predicted dynamic trajectory.
[0113] The shape of the noisy dynamic trajectory can be expressed as the number of frames*the number of atoms*3, where 3 represents the three dimensions in the three-dimensional coordinate information. Figure 4 In the dynamic trajectory with noise, each square can represent the position information of an atom, and the sequence information corresponds to the dynamic trajectory. For example, the sequence information of a protein refers to the order of amino acids in the protein molecule, and the atoms in each frame of the dynamic trajectory are arranged according to the order of amino acids. If it is a complex, the atoms of one biological molecule can be sorted after the atoms of another biological molecule are sorted, and the order of the biological molecules is not limited.
[0114] During the model training stage, the initial all-atom structure prediction model can be trained according to the difference between the predicted dynamic trajectory and the dynamic trajectory sample to obtain the all-atom structure prediction model.
[0115] In order to implement the above embodiment, the embodiment of the present application also proposes a full-atom structure prediction method. Figure 5 A schematic diagram of the flow chart of the all-atom structure prediction method provided in one embodiment of the present application.
[0116] like Figure 5 As shown, the all-atom structure prediction method includes:
[0117] Step 501, obtaining structural information of biological molecules.
[0118] In the present application, the explanation of the structural information of biological molecules can be found in the above embodiments, so it will not be repeated here.
[0119] Step 502: Encode the structural information through the all-atom structure prediction model to obtain encoding features.
[0120] The all-atom structure prediction model may be obtained by training using the training method described in the above embodiment.
[0121] Among them, the all-atom structure prediction model can include an encoder and a decoder. The structural information can be input into the encoder, and the structural information of the biological molecule can be encoded by the encoder to obtain the encoding features.
[0122] Step 503 , decoding is performed based on the coding features and dynamic trajectory noise using the all-atom structure prediction model to obtain the dynamic trajectory of the biomolecule.
[0123] The dynamic trajectory noise may include noise at different time points, such as noise of multiple frames.
[0124] In the present application, the coding features and dynamic trajectory noise can be input into a decoder for decoding, and the noise can be gradually removed by decoding to obtain the dynamic trajectory of the biomolecule. The dynamic trajectory can include the position information of atoms in the biomolecule at different time points.
[0125] In the embodiments of the present application, the all-atom structure prediction model trained by the training method described in the above embodiments can be used to predict the dynamic trajectory of biological molecules, thereby improving the accuracy of the prediction.
[0126] In order to implement the above embodiment, the embodiment of the present application also proposes a training device for a full-atom structure prediction model. Figure 6 A schematic diagram of the structure of a training device for an all-atom structure prediction model provided in one embodiment of the present application.
[0127] like Figure 6 As shown, the training device 600 of the all-atom structure prediction model includes:
[0128] An acquisition module 610 is used to acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule; wherein the first dynamic trajectory includes position information of atoms in the biomolecule at different time points;
[0129] A noise adding module 620, configured to add noise to the first dynamic trajectory to obtain a second dynamic trajectory;
[0130] An encoding module 630, used to encode the structural information to obtain encoding features;
[0131] A decoding module 640, configured to decode the coding feature and the second dynamic trajectory to obtain a target dynamic trajectory;
[0132] The training module 650 is used to train the initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain the all-atom structure prediction model.
[0133] Optionally, the decoding module 640 is used to:
[0134] Performing dimensionality reduction processing on the second dynamic trajectory to obtain a third dynamic trajectory;
[0135] The coding feature and the third dynamic trajectory are decoded to obtain the target dynamic trajectory.
[0136] Optionally, the decoding module 640 is used to:
[0137] Sorting the plurality of first sub-block trajectories according to a preset order to obtain a first trajectory sequence;
[0138] The encoded feature and the first trajectory sequence are decoded to obtain the target dynamic trajectory.
[0139] Optionally, the decoding module 640 is used to:
[0140] Decoding the coding feature and the first trajectory sequence to obtain a second trajectory sequence;
[0141] Rearranging a plurality of second sub-block trajectories in the second trajectory sequence to obtain a fourth dynamic trajectory;
[0142] The fourth dynamic trajectory is interpolated to obtain the target dynamic trajectory.
[0143] Optionally, the acquisition module 610 is used to:
[0144] generating a plurality of static conformations of the biomolecule by calculating the interaction energy between the biomolecules according to the structural information of the biomolecule;
[0145] The plurality of static conformations are simulated to capture a first dynamic trajectory of the biomolecule.
[0146] Optionally, the decoding module 640 is used to:
[0147] sorting the position information of each atom in the second dynamic trajectory according to a preset order to obtain a third trajectory sequence;
[0148] Decoding the coding feature and the third trajectory sequence to obtain a fourth trajectory sequence;
[0149] The position information of atoms in the fourth trajectory sequence is rearranged to obtain the target dynamic trajectory.
[0150] Optionally, the training module 650 is used to:
[0151] Determine a first loss corresponding to any time point according to a difference between position information of an atom at any time point in the target dynamic trajectory and position information of the same atom at any time point in the first dynamic trajectory;
[0152] According to the first losses corresponding to different time points, the second loss is obtained;
[0153] The initial all-atom structure prediction model is trained according to the second loss to obtain the all-atom structure prediction model.
[0154] It should be noted that the explanation of the aforementioned embodiment of the training method of the all-atom structure prediction model is also applicable to the training device of the all-atom structure prediction model of this embodiment, so it will not be repeated here.
[0155] In the embodiment of the present application, the structural information of the biomolecule is encoded to obtain a coding feature, and the target dynamic trajectory is obtained by decoding the coding feature and the second dynamic trajectory obtained by adding noise to the first dynamic trajectory, and then the all-atom structure prediction model is obtained through training based on the difference between the target dynamic trajectory and the first dynamic trajectory. Thus, based on the structural information of the biomolecule and the first dynamic trajectory of the biomolecule, a full-atom structure prediction model that can predict the dynamic trajectory of the biomolecule is obtained through training, and the prediction of the dynamic trajectory of the biomolecule is realized, which improves the generalization ability of the model.
[0156] In order to implement the above embodiment, the embodiment of the present application also proposes a full-atom structure prediction device. Figure 7 A schematic diagram of the structure of an all-atom structure prediction device provided in one embodiment of the present application.
[0157] like Figure 7 As shown, the all-atom structure prediction device 700 includes:
[0158] An acquisition module 710, used to acquire structural information of biomolecules;
[0159] An encoding module 720 is used to encode the structural information through an all-atom structure prediction model to obtain encoding features; wherein the all-atom structure prediction model can be trained using the device described in the above embodiment;
[0160] The decoding module 730 is used to decode the encoding features and dynamic trajectory noise through the all-atom structure prediction model to obtain the dynamic trajectory of the biological molecule; wherein the dynamic trajectory includes the position information of atoms in the biological molecule at different time points.
[0161] It should be noted that the explanation of the aforementioned all-atom structure prediction method embodiment is also applicable to the all-atom structure prediction device of this embodiment, so it will not be repeated here.
[0162] In the embodiments of the present application, the all-atom structure prediction model obtained by training based on the training device described in the above embodiments can be used to predict the dynamic trajectory of biological molecules, thereby improving the accuracy of the prediction.
[0163] According to an embodiment of the present application, the present application also provides an electronic device, a readable storage medium and a computer program product.
[0164] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0165] like Figure 8As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 802 or a computer program loaded from a storage unit 808 to a RAM (Random Access Memory) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An I / O (Input / Output) interface 805 is also connected to the bus 804.
[0166] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0167] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, CPU (Central Processing Unit), GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the training method of the all-atom structure prediction model. For example, in some embodiments, the training method of the all-atom structure prediction model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the training method of the all-atom structure prediction model described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the training method of the all-atom structure prediction model in any other appropriate manner (for example, by means of firmware).
[0168] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general-purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0169] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0170] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only-Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0171] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0172] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.
[0173] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services (Virtual Private Server). The server may also be a server of a distributed system, or a server combined with a blockchain.
[0174] It should be noted that the above electronic device can also implement the all-atom structure prediction method described in the above embodiment.
[0175] According to an embodiment of the present application, the present application also provides a computer program product. When the instruction processor in the computer program product is executed, it executes the training method of the all-atom structure prediction model or the all-atom structure prediction method proposed in the above embodiments of the present application.
[0176] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.
[0177] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.
Claims
1. A training method for an all-atom structure prediction model, comprising: Acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule; wherein the first dynamic trajectory includes position information of atoms in the biomolecule at different time points; adding noise to the first dynamic trajectory to obtain a second dynamic trajectory; Encoding the structural information to obtain encoding features; Decoding the coding feature and the second dynamic trajectory to obtain a target dynamic trajectory; According to the difference between the target dynamic trajectory and the first dynamic trajectory, the initial all-atom structure prediction model is trained to obtain the all-atom structure prediction model.
2. The method of claim 1, wherein: The decoding of the coding feature and the second dynamic trajectory to obtain the target dynamic trajectory includes: Performing dimensionality reduction processing on the second dynamic trajectory to obtain a third dynamic trajectory; The coding feature and the third dynamic trajectory are decoded to obtain the target dynamic trajectory.
3. The method of claim 2, wherein: The third dynamic trajectory includes a plurality of first sub-block trajectories, and decoding the encoding feature and the third dynamic trajectory to obtain the target dynamic trajectory includes: Sorting the plurality of first sub-block trajectories according to a preset order to obtain a first trajectory sequence; The encoded feature and the first trajectory sequence are decoded to obtain the target dynamic trajectory.
4. The method of claim 3, wherein: The decoding of the coding feature and the first trajectory sequence to obtain the target dynamic trajectory includes: Decoding the coding feature and the first trajectory sequence to obtain a second trajectory sequence; Rearranging a plurality of second sub-block trajectories in the second trajectory sequence to obtain a fourth dynamic trajectory; The fourth dynamic trajectory is interpolated to obtain the target dynamic trajectory.
5. The method of claim 1, wherein: Acquiring a first dynamic trajectory of the biomolecule includes: generating a plurality of static conformations of the biomolecule by calculating the interaction energy between the biomolecules according to the structural information of the biomolecule; The plurality of static conformations are simulated to capture a first dynamic trajectory of the biomolecule.
6. The method of claim 1, wherein: The decoding of the coding feature and the second dynamic trajectory to obtain the target dynamic trajectory includes: sorting the position information of each atom in the second dynamic trajectory according to a preset order to obtain a third trajectory sequence; Decoding the coding feature and the third trajectory sequence to obtain a fourth trajectory sequence; The position information of atoms in the fourth trajectory sequence is rearranged to obtain the target dynamic trajectory.
7. The method according to any one of claims 1 to 6, wherein: The step of training the initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain the all-atom structure prediction model comprises: Determine a first loss corresponding to any time point according to a difference between position information of an atom at any time point in the target dynamic trajectory and position information of the same atom at any time point in the first dynamic trajectory; According to the first losses corresponding to different time points, the second loss is obtained; The initial all-atom structure prediction model is trained according to the second loss to obtain the all-atom structure prediction model.
8. A method for predicting an all-atom structure, comprising: Obtain structural information of biomolecules; The structural information is encoded by an all-atom structure prediction model to obtain encoding features; wherein the all-atom structure prediction model can be trained by the method described in any one of claims 1 to 7; The all-atom structure prediction model is used to decode the coding features and the dynamic trajectory noise to obtain the dynamic trajectory of the biological molecule; wherein the dynamic trajectory includes the position information of atoms in the biological molecule at different time points.
9. A training device for an all-atom structure prediction model, comprising: An acquisition module, used to acquire structural information of a biomolecule and a first dynamic trajectory of the biomolecule; wherein the first dynamic trajectory includes position information of atoms in the biomolecule at different time points; A noise adding module, used for adding noise to the first dynamic trajectory to obtain a second dynamic trajectory; An encoding module, used for encoding the structural information to obtain encoding features; A decoding module, used for decoding the coding feature and the second dynamic trajectory to obtain a target dynamic trajectory; A training module is used to train an initial all-atom structure prediction model according to the difference between the target dynamic trajectory and the first dynamic trajectory to obtain a all-atom structure prediction model.
10. The device of claim 9, wherein: The decoding module is used for: Performing dimensionality reduction processing on the second dynamic trajectory to obtain a third dynamic trajectory; The coding feature and the third dynamic trajectory are decoded to obtain the target dynamic trajectory.
11. The device of claim 10, wherein: The decoding module is used for: Sorting the plurality of first sub-block trajectories according to a preset order to obtain a first trajectory sequence; The encoded feature and the first trajectory sequence are decoded to obtain the target dynamic trajectory.
12. The device of claim 11, wherein: The decoding module is used for: Decoding the coding feature and the first trajectory sequence to obtain a second trajectory sequence; Rearranging a plurality of second sub-block trajectories in the second trajectory sequence to obtain a fourth dynamic trajectory; The fourth dynamic trajectory is interpolated to obtain the target dynamic trajectory.
13. The device of claim 9, wherein: The acquisition module is used to: generating a plurality of static conformations of the biomolecule by calculating the interaction energy between the biomolecules according to the structural information of the biomolecule; The plurality of static conformations are simulated to capture a first dynamic trajectory of the biomolecule.
14. The device of claim 9, wherein: The decoding module is used for: sorting the position information of each atom in the second dynamic trajectory according to a preset order to obtain a third trajectory sequence; Decoding the coding feature and the third trajectory sequence to obtain a fourth trajectory sequence; The position information of atoms in the fourth trajectory sequence is rearranged to obtain the target dynamic trajectory.
15. The device according to any one of claims 9 to 14, wherein: The training module is used to: Determine a first loss corresponding to any time point according to a difference between position information of an atom at any time point in the target dynamic trajectory and position information of the same atom at any time point in the first dynamic trajectory; According to the first losses corresponding to different time points, the second loss is obtained; The initial all-atom structure prediction model is trained according to the second loss to obtain the all-atom structure prediction model.
16. An all-atom structure prediction device, comprising: An acquisition module, used to obtain the structural information of biomolecules; An encoding module, used to encode the structural information through an all-atom structure prediction model to obtain encoding features; wherein the all-atom structure prediction model can be trained using the device described in any one of claims 9 to 15; A decoding module is used to decode the encoding features and dynamic trajectory noise through the all-atom structure prediction model to obtain the dynamic trajectory of the biological molecule; wherein the dynamic trajectory includes the position information of atoms in the biological molecule at different time points.
17. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method of any one of claims 1 to 7, or to execute the method of claim 8.
18. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 7, or to execute the method according to claim 8.
19. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 7, or implements the steps of the method according to claim 8.