Protein residue motion coordinate prediction method, equipment, medium and program product

Abstract

The invention discloses a protein residue motion coordinate prediction method and device, a medium and a program product, and relates to the technical field of proteodynamics, and the protein residue motion coordinate prediction method comprises the steps: obtaining a first frequency spectrum corresponding to a coordinate sequence of a protein residue, and obtaining a feature component in a preset frequency range from the first frequency spectrum, a frequency domain feature vector is determined, and the preset frequency range is lower than a preset frequency threshold value; inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining a predicted frequency domain vector; and converting the predicted frequency domain vector into a corresponding second frequency spectrum, and reducing the second frequency spectrum into a predicted coordinate sequence of the protein residues. The technical problems that a traditional protein residue motion trail prediction scheme is large in calculation amount and low in efficiency are solved.

Description

Technical Field

[0001] The present application relates to the field of protein dynamics technology, and in particular to a method, device, computer-readable storage medium, and computer program product for predicting protein residue motion coordinates. Background Art

[0002] Currently, in the field of protein dynamics prediction, when faced with the need to model and predict residue motion trajectories, the traditional approach is to predict the three-dimensional coordinates of protein residues at each time point, frame by frame, in the time domain. However, as the number of prediction steps increases, the model's input and output dimensions grow linearly, resulting in a dramatic increase in computational effort and significantly prolonged inference time. This approach is inefficient and makes it difficult to meet the requirements for long-term conformational changes.

[0003] The above information disclosed in this Background section is only for understanding the background of the present invention and therefore it may contain information that does not constitute prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a method, device, medium and program product for predicting the motion coordinates of protein residues, aiming to solve the technical problems of large computational complexity and low efficiency of traditional protein residue motion trajectory prediction schemes.

[0005] To achieve the above objectives, the present application provides a method for predicting protein residue motion coordinates, the method comprising:

[0006] Obtaining a first frequency spectrum corresponding to the coordinate sequence of the protein residues, and obtaining characteristic components within a preset frequency range from the first frequency spectrum to determine a frequency domain characteristic vector, wherein the preset frequency range is lower than a preset frequency threshold;

[0007] Inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining a predicted frequency domain vector;

[0008] The predicted frequency domain vector is converted into a corresponding second spectrum, and the second spectrum is restored to a predicted coordinate sequence of protein residues.

[0009] In one embodiment, the steps of obtaining a first frequency spectrum corresponding to the coordinate sequence of protein residues, obtaining characteristic components within a preset frequency range from the first frequency spectrum, and determining a frequency domain characteristic vector include:

[0010] Performing a fast Fourier transform on the coordinate sequence of the protein residues to obtain a first frequency spectrum of the time series signal corresponding to each of the three-dimensional directions;

[0011] Cutting off a preset number of low-frequency components with the largest amplitudes from each of the first frequency spectra, or selecting low-frequency components within a preset frequency range from each of the first frequency spectra;

[0012] The low-frequency components are concatenated into a frequency domain feature vector based on the order of protein residues and three-dimensional coordinate axes.

[0013] In one embodiment, the potential diffusion model includes a normalization module and a flow network module;

[0014] The step of inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining the predicted frequency domain vector comprises:

[0015] After the structural feature vector, sequence feature vector and frequency domain feature vector of the protein residues are input into a preset potential diffusion model, the frequency domain feature vector is normalized by the normalization module to obtain a standard frequency domain vector;

[0016] Processing the standard frequency domain vector through the residual block in the stream network module, and then fusing the processed standard frequency domain vector, the sequence feature vector and the structural feature vector into the attention calculation to obtain a multidimensional frequency domain feature vector;

[0017] The multi-dimensional frequency domain feature vector is initialized based on a normal distribution, and denoising iteration is performed on the multi-dimensional frequency domain feature vector to obtain a predicted feature vector.

[0018] In one embodiment, before the step of inputting the structural feature vector, the sequence feature vector, and the frequency domain feature vector of the protein residues into a preset latent diffusion model, the method further comprises:

[0019] Obtain the initial static structure and amino acid sequence of protein residues;

[0020] Encoding the initial static structure through a preset variational autoencoder to obtain a structural feature vector of the protein residues;

[0021] Extracting a sequence feature vector corresponding to the amino acid sequence.

[0022] In one embodiment, the protein residue motion coordinate prediction method further comprises:

[0023] obtaining degree features of protein residues, and determining the degree features as reconstruction content targets of an initial diffusion model;

[0024] Training the initial diffusion model based on the reconstructed content target to obtain a pre-trained model;

[0025] The pre-trained model is migrated to a preset spectrum reconstruction task, and the pre-trained model is trained with minimizing the frequency domain reconstruction error as a training goal to obtain a potential diffusion model.

[0026] In one embodiment, the step of converting the predicted frequency domain vector into a corresponding second spectrum and restoring the second spectrum to a predicted coordinate sequence of protein residues includes:

[0027] Performing high-frequency zero padding on the predicted frequency domain vector to obtain a second spectrum;

[0028] The second spectrum is subjected to inverse fast Fourier transform to obtain a predicted coordinate sequence of protein residues.

[0029] In one embodiment, before the step of obtaining a first spectrum corresponding to the coordinate sequence of protein residues, the method further comprises:

[0030] Obtain original three-dimensional structure data and motion trajectory data of proteins;

[0031] Based on the motion trajectory data, the three-dimensional coordinates of each protein residue are recorded in a time series form to obtain a coordinate sequence of the protein residue.

[0032] In addition, to achieve the above-mentioned purpose, the present application also provides a protein residue motion coordinate prediction device, which at least includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program is configured to implement the steps of the protein residue motion coordinate prediction method as described above.

[0033] In addition, to achieve the above-mentioned purpose, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the protein residue motion coordinate prediction method as described above are implemented.

[0034] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned protein residue motion coordinate prediction method when executed by a processor.

[0035] The present application provides a method for predicting the motion coordinates of protein residues. First, a first spectrum corresponding to the coordinate sequence of the protein residues is obtained, and characteristic components within a preset frequency range are obtained from the first spectrum to determine a frequency domain feature vector. By using the low-frequency frequency domain vector as the target representation of the diffusion model, the original multidimensional time series signal can be significantly compressed, the scale of the model input and output can be reduced, and the computational overhead of training and inference can be significantly reduced. Then, the structural feature vector, sequence feature vector, and frequency domain feature vector of the protein residue are input into a preset latent diffusion model. The latent diffusion model performs diffusion prediction on the protein residue to determine a predicted frequency domain vector. The predicted frequency domain vector is then converted into a corresponding second spectrum, and the second spectrum is restored to the predicted coordinate sequence of the protein residue. In the technical solution of the present application, in order to avoid multiple iterations when restoring the time domain trajectory, the one-time generation advantage of the latent diffusion model in the field of image generation is utilized. The frequency domain feature components can be mapped to a compact latent space, allowing the diffusion model to complete the reconstruction of all multiple low-frequency coefficients in this space at one time, thereby obtaining a complete frequency domain representation, i.e., the predicted frequency domain. Finally, the predicted coordinate sequence of the protein residue can be obtained by restoring it back to the time domain. The protein residue motion coordinate prediction method of the embodiment of the present application not only avoids repeated calculations of frame-by-frame predictions and reduces the amount of calculation, but also achieves rapid generation of motion trajectory coordinates by leveraging the characteristics of fewer parameters and more concentrated expression in the potential diffusion model, thereby improving the prediction efficiency of protein residue motion trajectory coordinates. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0037] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0038] Figure 1 Schematic diagram of the principle of traditional protein residue motion trajectory prediction scheme;

[0039] Figure 2 This is a schematic diagram of a process for predicting protein residue motion coordinates according to an embodiment of the present application;

[0040] Figure 3 Schematic diagram of the principle of the method for predicting protein residue motion coordinates in the embodiments of this application;

[0041] Figure 4This is a data flow diagram of a feasible protein residue motion coordinate prediction solution in an embodiment of the present application;

[0042] Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the protein residue motion coordinate prediction method in the embodiment of the present application.

[0043] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0044] To make the above-mentioned purposes, features, and advantages of the present application more clearly understood, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0045] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0046] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0047] In the field of protein dynamics prediction, the present invention proposes a new approach for efficient modeling and reasoning of residue motion trajectories. Figure 1 As shown, the input protein structure is denoised based on a single-frame trajectory generation model to obtain a noisy protein structure, and then a single-frame protein structure is reconstructed through a denoising network (ScoreNetwork), and the frame-by-frame calculation is continued N times to finally obtain a reconstructed trajectory. This shows that traditional prediction schemes often predict the three-dimensional coordinates of each time point frame by frame in the time domain. As the number of prediction steps increases, the model input and output dimensions grow linearly, which will lead to a sharp increase in the amount of calculation and a significant extension of the reasoning time, making it difficult to meet the demand for long-term conformational changes. The core of the embodiment of the present application is to extract the low-frequency components of protein residues that change with time from a frequency domain perspective through Fourier transform - these low-frequency components can reflect the main trend of the overall conformation of the protein, while the high-frequency details contribute relatively little to the macroscopic motion. Therefore, in the technical solution of the embodiment of the present application, the modeling object of the residue trajectory is converted from hundreds to thousands of coordinate points in the original space to a small number of low-frequency coefficients that only need to be predicted (for example, it can be recorded as K dimension, which is much smaller than the time step T), thereby fundamentally reducing the dimension that the model needs to process and reducing the amount of calculation.

[0048] The present invention provides a method for predicting the motion coordinates of protein residues. Figure 2 , Figure 2 This is a flow chart of an embodiment of a method for predicting protein residue motion coordinates of the present application. The method for predicting protein residue motion coordinates includes:

[0049] Step S10, obtaining a first spectrum corresponding to the coordinate sequence of the protein residues, and obtaining characteristic components within a preset frequency range from the first spectrum to determine a frequency domain characteristic vector, wherein the preset frequency range is lower than a preset frequency threshold;

[0050] The protein residue coordinate sequence refers to a three-dimensional coordinate sequence of the protein residues that changes over time, while the first spectrum reflects the frequency domain characteristics of the coordinate sequence. The first spectrum can be obtained by converting the three-dimensional coordinate sequence into the time-frequency domain. In the embodiments of the present application, the frequency domain feature vectors in the first spectrum are used to replace the original protein residue coordinate sequence to predict the motion trajectory coordinates. This can reduce the high-dimensional problem that originally grows linearly with time steps to a fixed K-dimensional space, effectively reducing the computational complexity.

[0051] For example, all structural coordinates need to be centered and normalized before input to eliminate the impact of differences in protein size and coordinate scale on model training and prediction.

[0052] In addition, the feature classification of the preset frequency range in the first spectrum refers to the low-frequency component in the first spectrum, whose frequency is greater than the pre-occupied frequency threshold. For example, for the first spectrum, several characteristic components with the lowest frequency can reflect the macroscopic motion of the protein residues, and other characteristic components only reflect the microscopic motion characteristics to a certain extent, such as small swings or noise, which can be ignored. In the embodiment of the present application, only the low-frequency components in the first spectrum are extracted to obtain the frequency feature vector, which has the effect of compressing the dimension of the motion trajectory characteristics of the protein residues, and can also reduce the scale of the input and output data of the model, thereby reducing the computational overhead of model training and inference.

[0053] The embodiments of the present application mainly break the limitations of high-dimensional time-domain modeling through the idea of ​​dimensionality reduction. The embodiments of the present application specifically transform the spatial coordinate sequence of protein residues over time in the frequency domain, and only select low-frequency feature components containing major conformational information for learning. This can reduce the modeling dimension from T to a dimension K that is much smaller than T while maintaining prediction accuracy, thereby greatly reducing the scale of model input and output, and thus significantly reducing the computational overhead of training and inference.

[0054] Step S20, inputting the structural feature vector, sequence feature vector, and frequency domain feature vector of the protein residue into a preset latent diffusion model, performing diffusion prediction on the protein residue using the latent diffusion model, and determining a predicted frequency domain vector;

[0055] The present application further incorporates the Latent Diffusion Model (LDM) technology, a technology that has demonstrated strong capabilities in high-dimensional generation tasks in recent years. Unlike direct point-by-point generation based on the original coordinate sequence of protein residues, LDM can only iteratively add noise and denoise the low-frequency coefficients (i.e., frequency domain feature vectors) to generate a complete frequency domain representation (i.e., predicted frequency domain vector) in one go. By completing the diffusion process in the latent space, the model can effectively capture the global laws of residue motion while avoiding redundant calculations for independent reasoning of each frame, significantly shortening the overall reasoning time.

[0056] In addition, because the three-dimensional arrangement of protein residues has complex spatial relationships, the diffusion generation process requires the use of the invariant point attention (IPA) mechanism to incorporate the spatial prior of the initial structure (specifically, this can include the structural feature vectors and sequence feature vectors of the protein residues obtained after encoding) into the model to ensure that the generated predicted frequency domain vectors can maintain reasonable conformational coherence when reconstructed into three-dimensional coordinates. In this way, the model can significantly reduce the dimensionality and accelerate inference while still outputting physically self-consistent and biologically interpretable dynamic trajectories. Therefore, the model simultaneously refers to the initial geometric prior in each denoising iteration to ensure that the generated low-frequency coefficients conform to the frequency domain trend while maintaining spatial structural consistency. The diffusion process completes the random noise injection and denoising restoration of all K low-frequency coefficients in the latent space at one time, without the need for frame-by-frame iteration.

[0057] Step S30: converting the predicted frequency domain vector into a corresponding second spectrum, and restoring the second spectrum into a predicted coordinate sequence of protein residues.

[0058] After the latent diffusion model predicts the low-frequency predicted frequency domain vector, it needs to be restored to obtain the motion trajectory in the form of a coordinate sequence. Specifically, the predicted frequency domain vector can be converted into the corresponding second spectrum by referring to the coordinate sequence conversion process described above. This is then further restored to the predicted coordinate sequence of the protein residues. The predicted coordinate sequence can intuitively reflect the motion trajectory of the protein residues.

[0059] The technical solution of the embodiment of the present application maps the motion trajectory of protein residues to the frequency domain, retaining only a few low-frequency frequency domain feature vectors and generating a complete trajectory in the latent space at one time, thereby significantly reducing the model dimension and the number of iterations. Compared with the traditional time-domain frame-by-frame prediction scheme, the inference efficiency is improved several times, and the time overhead increases smoothly with the growth of the trajectory length, without sudden increases, thus achieving efficient expansion; in addition, the embodiment of the present application also continuously integrates three-dimensional spatial constraints in the diffusion model with IPA attention, ensuring that the trajectory obtained after the generated prediction domain vector is restored can not only truly reflect the global motion trend of the protein, but also avoid structural drift and numerical instability.

[0060] Furthermore, the steps of obtaining a first frequency spectrum corresponding to the coordinate sequence of the protein residues, obtaining characteristic components within a preset frequency range from the first frequency spectrum, and determining a frequency domain characteristic vector may include:

[0061] Step S11, performing fast Fourier transform on the coordinate sequence of the protein residues to obtain a first spectrum of the time series signal corresponding to each of the three-dimensional directions;

[0062] Step S12, extracting a preset number of low-frequency components with the largest amplitudes from each first spectrum, or selecting low-frequency components within a preset frequency range from each first spectrum;

[0063] In step S13, the low-frequency components are concatenated into a frequency domain feature vector based on the order of the protein residues and the three-dimensional coordinate axes.

[0064] In the embodiment of the present application, a Fast Fourier Transform (FFT) is first performed on each residue coordinate sequence to calculate the first spectrum corresponding to the three time series signals in the three-dimensional directions x, y, and z, and the K low-frequency components with the largest amplitude (or the low-frequency components within a uniformly selected frequency range) are intercepted. Then, they are spliced ​​into a length of 3·K·N according to the order of the residues and the coordinate axes. e Frequency domain eigenvector of N e The frequency domain feature vector is used as the target representation of the potential diffusion model, which significantly compresses the original T·3·N e dimensional timing signal.

[0065] Through the above steps, the embodiment of the present application reduces the modeling dimension from T·3·N to e Compressed to 3·K·N e , which greatly reduces the amount of computation and uses the potential diffusion model to achieve one-time complete trajectory generation, thereby keeping the inference time within an acceptable range even in trajectory prediction tasks of hundreds of steps, thereby improving processing efficiency.

[0066] In the embodiment of the present application, the three-dimensional coordinate sequence of protein residues over time is first mapped to the frequency domain through discrete Fourier transform, and only the low-frequency coefficients containing macroscopic motion information are retained, thereby reducing the high-dimensional problem that originally grows linearly with time steps to a fixed K-dimensional space.

[0067] In a feasible embodiment, the step of converting the predicted frequency domain vector into a corresponding second spectrum and restoring the second spectrum to a predicted coordinate sequence of protein residues may include:

[0068] Step S31, performing high-frequency zero padding on the predicted frequency domain vector to obtain a second spectrum;

[0069] Step S32: performing an inverse fast Fourier transform on the second spectrum to obtain a predicted coordinate sequence of the protein residues.

[0070] The process of performing high-frequency zero padding (zero-padding) on ​​the predicted frequency domain vector in the frequency domain to achieve spectrum completion is mainly based on the characteristics of the Fourier transform in signal processing and its time-frequency domain correspondence. Its mathematical essence is to insert zero values ​​in the middle (high frequency band) of the frequency domain signal, which is equivalent to expanding the original spectrum. Due to the symmetry of the Fourier transform, frequency domain zero padding does not change the actual frequency component of the signal, but it will produce a higher number of sampling points in the time domain through the inverse Fourier transform (IFFT). For example, if the length of the original frequency domain signal is N, it becomes M after zero padding (M>N), and the number of time domain signal points also becomes M, which is equivalent to interpolating the original time domain signal.

[0071] After restoring the reconstructed low-frequency coefficients to a complete spectrum, the frequency domain data is restored to a T-step coordinate sequence in the time domain through inverse fast Fourier transform, which intuitively reflects the motion trajectory information of the protein residues.

[0072] In combination with the contents of the aforementioned application embodiments, the present application embodiment essentially provides a frequency domain shallow space trajectory generation model, such as Figure 3 As shown in the figure, the input protein structure information is first Fourier transformed to obtain K-dimensional low-frequency coefficients (noise) with a size of ×2×K. Then, the structural feature vector and sequence feature vector after the Encoder encoder encodes the initial structure of the protein and the aforementioned low-frequency coefficients (equivalent to the frequency domain feature vector) are input into the latent space diffusion model LDM, and then decoded by the Decoder decoder to output the predicted K-dimensional low-frequency coefficients with a size of ×2×K. Finally, the trajectory is reconstructed by inverse Fourier transform, thereby realizing the trajectory reconstruction of the input protein structure based on low-frequency dynamic motion.

[0073] In a possible embodiment, Figure 4As shown, the potential diffusion model includes a normalization module and a flow network module;

[0074] The step of inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining the predicted frequency domain vector comprises:

[0075] Step S21, after the structural feature vector, sequence feature vector, and frequency domain feature vector of the protein residues are input into a preset potential diffusion model, the frequency domain feature vector is normalized by a normalization module to obtain a standard frequency domain vector;

[0076] like Figure 4 As shown in the figure, the latent space of the potential diffusion model includes two parts: the standardization module and the flow network module. The frequency domain feature components obtained from the motion spectrum are encoded by the encoder to obtain z0 and then standardized to obtain the standard frequency domain vector z t , then the standard frequency domain vector z t Input into the flow network, the flow network outputs the predicted frequency domain vector z * 0 to the decoder, and after completing the spectrum reconstruction, the trajectory and protein structure are reconstructed through IFFT (Inverse Fast Fourier Transform). Among them, the normalization process can include centering and normalization to eliminate the impact of different protein sizes and coordinate scale differences on model training and prediction.

[0077] Step S22: Process the standard frequency domain vector through the residual block in the flow network module, and then fuse the processed standard frequency domain vector, sequence feature vector and structural feature vector into the attention calculation to obtain a multi-dimensional frequency domain feature vector;

[0078] Specifically, the core module based on the latent diffusion model (LDM) uses a U-Net (a convolutional neural network architecture for image segmentation) style denoising network combined with IPA. The network input of the flow network module consists of three parts: the frequency domain feature z after the current perturbation t(i.e., standard frequency domain vectors), structural latent vectors of protein residues generated based on the encoder, and one-hot (unique hot encoding) corresponding to the amino acid sequence of the protein residues or sequence feature vectors obtained through pre-training Embedding. The flow network first performs preliminary processing on the standard frequency domain vectors through several residual blocks (ResidualBlock), and then in multiple IPA sub-layers, the three-dimensional geometric information of the coordinate system (including the aforementioned three parts of the network input) is integrated into the attention calculation to ensure that the overall spatial consistency is preserved during the denoising process. Specifically, each diffusion time step t corresponds to a set of denoising parameters, the noise scheduling adopts linear or cosine scheduling, and the denoising target is the reconstruction estimate of the frequency domain feature vector.

[0079] Step S23 : Initialize the multi-dimensional frequency domain feature vector based on the normal distribution, and perform denoising iteration on the multi-dimensional frequency domain feature vector to obtain a predicted feature vector.

[0080] During the inference phase of the flow network module, full diffusion sampling is performed on the K-dimensional frequency-domain feature vector (i.e., the multidimensional frequency-domain feature vector) in the latent space. Specifically, the frequency-domain coefficient noise is initialized from a standard normal distribution. After L steps of denoising iterations, a clear frequency-domain vector is obtained. It is important to note that this inference process does not require a frame-by-frame loop; instead, a complete frequency-domain representation is generated all at once, resulting in a corresponding predicted feature vector that reflects the predicted trajectory of the protein residues.

[0081] Furthermore, in a feasible embodiment, before the step of inputting the structural feature vector, sequence feature vector, and frequency domain feature vector of the protein residues into a preset potential diffusion model, the method may further include:

[0082] Step A10, obtaining the initial static structure and amino acid sequence of the protein residues;

[0083] Step A20, encoding the initial static structure through a preset variational autoencoder to obtain a structural feature vector of the protein residues;

[0084] Step A30: extracting the sequence feature vector corresponding to the amino acid sequence.

[0085] In an embodiment of the present application, the initial static structure (equivalent to the step 0 structure) and amino acid sequence of the protein residues are first obtained. The preset variational autoencoder can be a GearBind (a pre-trained deep neural network) structure VAE (Variational Auto Encoder), which can encode the step 0 structure of the protein residues to obtain the corresponding structural feature vector. The VAE represents the protein structure as graph data, the node features include residue or atomic coordinates and chemical properties, and the edge features represent the interaction between residues. The encoder part consists of a multi-layer graph convolutional network (GraphConv) and a variational layer (Variational Layer), and finally outputs a potential vector z, whose dimension is generally much smaller than the number of input nodes.

[0086] In another possible embodiment, Figure 4 As shown, the entire frequency-domain shallow spatial trajectory generation model also includes a decoder structure (i.e., a decoder). The decoder structure is symmetrical to the encoder and is used for structural reconstruction and latent space regularization during training to reconstruct the spectrum. In the embodiments of this application, GearBind VAE pre-training or fine-tuning can allow the latent vector to carry rich geometric priors.

[0087] In addition, when extracting the sequence feature vector corresponding to the amino acid sequence, the amino acid sequence can be encoded by the One-hot encoding method, or it can be processed by pre-training Embedding to obtain the corresponding sequence feature vector to input into the flow network module.

[0088] In the embodiment of the present application, by adopting GearBind VAE to encode the initial static structure, a high-quality geometric prior can be provided for LDM, and a compact geometric potential identifier can be obtained, which is input into the potential diffusion model together with the low-frequency feature vector in the frequency domain. At the same time, the invariant point attention (IPA) mechanism continuously introduces three-dimensional structural constraints in the diffusion denoising process to ensure that the generated trajectory maintains physical rationality and spatial consistency outside the global motion trend. It not only takes into account the prediction accuracy, but also maximizes the efficiency advantages of frequency domain dimensionality reduction and potential space generation. Therefore, by combining frequency domain low-frequency modeling with potential diffusion generation, the technical solution of the embodiment of the present application solves the problem of low efficiency and poor scalability of traditional protein dynamic prediction in high-dimensional time domain iteration, which can not only ensure the prediction accuracy of long trajectories, but also control the inference time within an acceptable range, thereby meeting the demand for efficient simulation of large-scale and long-time resolution motion of proteins.

[0089] In one embodiment, the protein residue motion coordinate prediction method may further include:

[0090] Step B10, obtaining the degree features of the protein residues, and determining the degree features as the reconstruction content target of the initial diffusion model;

[0091] Step B20, training the initial diffusion model based on the reconstructed content target to obtain a pre-trained model;

[0092] In step B30, the pre-trained model is transferred to a preset spectrum reconstruction task, and the pre-trained model is trained with minimizing the frequency domain reconstruction error as the training goal to obtain a potential diffusion model.

[0093] The embodiment of the present application provides a method for training a potential diffusion model. During the model training process, the frequency domain reconstruction error (MSE Loss) is used as the main target, so that the potential diffusion model can learn to accurately restore the low-frequency components and obtain a more accurate and stable predicted frequency domain vector.

[0094] Optionally, a geometric consistency loss or skeleton preservation loss can be added to the training objective during training to further constrain the physical continuity of the denoised result and the initial structure. The entire model training is performed in parallel on a GPU (Graphics Processing Unit) cluster.

[0095] It should be noted that before pre-training, a degree feature is constructed as the pre-training reconstruction content, and then the pre-trained model is transferred to the spectrum reconstruction task. The batch size during pre-training can be dynamically adjusted according to protein size. The learning rate is adjusted using the Adam (Adaptive Moment Estimation optimizer) with a cosine annealing strategy and a warm-up strategy.

[0096] In another embodiment, it is necessary to apply KL (Kullback-Leibler, relative entropy) divergence regularization to the pre-trained VAE (Variational Autoencoder) part to make the potential space distribution tend to standard normal in order to diffuse the model conditioning input. Figure 4 As shown in the figure, during the structure pre-training process, the graph features of the protein structure are mainly extracted, and then encoded and decoded by the Incoder encoder and the Decoder decoder to complete the training process.

[0097] In a feasible embodiment, before the step of obtaining the first spectrum corresponding to the coordinate sequence of the protein residues, the method may further include:

[0098] Step C10, obtaining original three-dimensional structure data and motion trajectory data of the protein;

[0099] Step C20 , based on the motion trajectory data, the three-dimensional coordinates of each protein residue are recorded in a time series format to obtain a coordinate sequence of the protein residue.

[0100] When implementing the technical solutions of the embodiments of this application, Figure 4 As shown, the data required for training and inference is first prepared, including the original 3D protein structure coordinates and the corresponding motion trajectory data (i.e., dataset trajectory). Protein sequences include sequences such as "HE...GTQ." Before formally predicting protein residue motion coordinates, complete T-step trajectories are obtained from molecular dynamics simulations or high-quality experimental data, recording the x, y, and z coordinates of each residue in a time series format. During model inference, only the static structure and amino acid sequence at step 0 are required, where step 0 refers to the initial state.

[0101] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the protein residue motion coordinate prediction method of the present application. More forms of simple transformations based on this technical concept are all within the scope of protection of the present application.

[0102] An embodiment of the present application also provides a protein residue motion coordinate prediction device, which includes at least: at least one processor; and 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 protein residue motion coordinate prediction method in the above embodiment.

[0103] Reference below Figure 5 , which shows a structural schematic diagram of a control unit suitable for implementing an embodiment of the present application. Figure 5 The control unit shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0104] like Figure 5As shown, the control unit may include a processing device 101 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 102 or programs loaded from a storage device 103 into a random access memory (RAM) 104. RAM 104 also stores various programs and data required for the operation of the control unit. Processing device 101, ROM 102, and RAM 104 are interconnected via a bus 105. An input / output (I / O) interface 106 is also connected to the bus. Typically, the following systems may be connected to I / O interface 106: input devices 107 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 108 including, for example, a liquid crystal display (LCD), speakers, vibrator, etc.; storage device 103 including, for example, a magnetic tape, hard disk, etc.; and communication device 109. Communication device 109 may allow the control unit to communicate with other devices wirelessly or by wire to exchange data. Although the figures show a control unit with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented or have instead.

[0105] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 103, or installed from a ROM 102. When the computer program is executed by the processing device 101, the above-mentioned functions defined in the method of the embodiment of the present application are performed.

[0106] The protein residue motion coordinate prediction device provided in the embodiments of the present application utilizes the protein residue motion coordinate prediction method described in the aforementioned embodiments, thereby resolving the technical issues of high computational complexity and low efficiency associated with conventional protein residue motion trajectory prediction schemes. Compared to the prior art, the beneficial effects of the protein residue motion coordinate prediction device provided in the embodiments of the present application are the same as those of the protein residue motion coordinate prediction method described in the aforementioned embodiments. Other technical features of the protein residue motion coordinate prediction device are the same as those disclosed in the aforementioned embodiments and are not further elaborated upon here.

[0107] It should be understood that the various parts of the embodiments of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0108] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the above claims.

[0109] An embodiment of the present application further provides a computer-readable storage medium storing a computer program executable on a processor, wherein the computer program is used to execute the protein residue motion coordinate prediction method in the above embodiment.

[0110] The computer-readable storage medium provided in the embodiments of the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0111] The computer-readable storage medium may be included in the protein residue motion coordinate prediction device; or it may exist independently without being assembled into the protein residue motion coordinate prediction device.

[0112] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the protein residue motion coordinate prediction device, the protein residue motion coordinate prediction device performs the following steps: obtaining a first spectrum corresponding to the coordinate sequence of the protein residue, and obtaining characteristic components within a preset frequency range from the first spectrum to determine a frequency domain feature vector, wherein the preset frequency range is lower than a preset frequency threshold; inputting the structural feature vector, sequence feature vector and frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model to determine a predicted frequency domain vector; converting the predicted frequency domain vector into a corresponding second spectrum, and restoring the second spectrum to a predicted coordinate sequence of the protein residue.

[0113] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0114] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0115] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0116] The computer-readable storage medium provided in the embodiments of the present application stores computer-readable program instructions for executing the above-described method for predicting protein residue motion coordinates. This method can address the technical issues of high computational complexity and low efficiency associated with conventional protein residue motion trajectory prediction methods. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in the embodiments of the present application are similar to those of the method for predicting protein residue motion coordinates provided in the above-described embodiments, and are not further elaborated here.

[0117] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned protein residue motion coordinate prediction method.

[0118] The computer program product provided in the embodiments of this application can address the technical issues of high computational complexity and low efficiency associated with conventional protein residue motion trajectory prediction methods. Compared to the prior art, the computer program product provided in the embodiments of this application offers the same beneficial effects as the protein residue motion coordinate prediction method provided in the aforementioned embodiments, and will not be further elaborated here.

[0119] The above is only an exemplary solution of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A method for predicting protein residue motion coordinates, characterized in that: The protein residue motion coordinate prediction method comprises: Obtaining a first frequency spectrum corresponding to the coordinate sequence of the protein residues, and obtaining characteristic components within a preset frequency range from the first frequency spectrum to determine a frequency domain characteristic vector, wherein the preset frequency range is lower than a preset frequency threshold; Inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining a predicted frequency domain vector; The predicted frequency domain vector is converted into a corresponding second spectrum, and the second spectrum is restored to a predicted coordinate sequence of protein residues.

2. The method for predicting protein residue motion coordinates according to claim 1, wherein: The steps of obtaining a first spectrum corresponding to the coordinate sequence of the protein residues, obtaining characteristic components within a preset frequency range from the first spectrum, and determining a frequency domain characteristic vector include: Performing a fast Fourier transform on the coordinate sequence of the protein residues to obtain a first frequency spectrum of the time series signal corresponding to each of the three-dimensional directions; Cutting off a preset number of low-frequency components with the largest amplitudes from each of the first frequency spectra, or selecting low-frequency components within a preset frequency range from each of the first frequency spectra; The low-frequency components are concatenated into a frequency domain feature vector based on the order of protein residues and three-dimensional coordinate axes.

3. The method for predicting protein residue motion coordinates according to claim 1, wherein: The potential diffusion model includes a normalization module and a flow network module; The step of inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residue into a preset potential diffusion model, performing diffusion prediction on the protein residue by the potential diffusion model, and determining the predicted frequency domain vector comprises: After the structural feature vector, sequence feature vector and frequency domain feature vector of the protein residues are input into a preset potential diffusion model, the frequency domain feature vector is normalized by the normalization module to obtain a standard frequency domain vector; Processing the standard frequency domain vector through the residual block in the stream network module, and then fusing the processed standard frequency domain vector, the sequence feature vector and the structural feature vector into the attention calculation to obtain a multidimensional frequency domain feature vector; The multi-dimensional frequency domain feature vector is initialized based on a normal distribution, and denoising iteration is performed on the multi-dimensional frequency domain feature vector to obtain a predicted feature vector.

4. The method for predicting protein residue motion coordinates according to claim 3, wherein: Before the step of inputting the structural feature vector, the sequence feature vector and the frequency domain feature vector of the protein residues into a preset potential diffusion model, the method further comprises: Obtain the initial static structure and amino acid sequence of protein residues; Encoding the initial static structure through a preset variational autoencoder to obtain a structural feature vector of the protein residues; Extracting a sequence feature vector corresponding to the amino acid sequence.

5. The method for predicting protein residue motion coordinates according to claim 3, wherein: The protein residue motion coordinate prediction method further comprises: obtaining degree features of protein residues, and determining the degree features as reconstruction content targets of an initial diffusion model; Training the initial diffusion model based on the reconstructed content target to obtain a pre-trained model; The pre-trained model is migrated to a preset spectrum reconstruction task, and the pre-trained model is trained with minimizing the frequency domain reconstruction error as a training goal to obtain a potential diffusion model.

6. The method for predicting protein residue motion coordinates according to claim 1, wherein: The step of converting the predicted frequency domain vector into a corresponding second spectrum and restoring the second spectrum to a predicted coordinate sequence of protein residues comprises: Performing high-frequency zero padding on the predicted frequency domain vector to obtain a second spectrum; The second spectrum is subjected to inverse fast Fourier transform to obtain a predicted coordinate sequence of protein residues.

7. The method for predicting protein residue motion coordinates according to any one of claims 1 to 6, wherein: Before the step of obtaining a first frequency spectrum corresponding to the coordinate sequence of protein residues, the method further comprises: Obtain original three-dimensional structure data and motion trajectory data of proteins; Based on the motion trajectory data, the three-dimensional coordinates of each protein residue are recorded in a time series form to obtain a coordinate sequence of the protein residue.

8. A protein residue motion coordinate prediction device, characterized in that: The protein residue motion coordinate prediction device comprises at least: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the protein residue motion coordinate prediction method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program for implementing a method for predicting protein residue motion coordinates, and the program for implementing a method for predicting protein residue motion coordinates is executed by a processor to implement the steps of the method for predicting protein residue motion coordinates as described in any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method for predicting protein residue motion coordinates are implemented.

Images