Method for identifying protein-nucleic acid binding residues and related device

Abstract

Embodiments of the present application relate to the technical field of artificial intelligence, and in particular to a method for identifying protein-nucleic acid binding residues and a related device. The method comprises: acquiring the amino acid sequence and three-dimensional structural data of a protein to be identified; inputting the amino acid sequence into a trained protein language model for a sequence feature extraction operation so as to obtain sequence feature data; inputting the three-dimensional structural data, in the form of a point cloud, into a sparse convolutional network for a structural feature extraction operation so as to obtain structural feature data; performing a weighted fusion operation on the sequence feature data and the structural feature data so as to obtain a weighted fusion feature; and inputting the weighted fusion feature into a trained multi-layer perception classifier for an identification operation so as to obtain probability information indicating whether each residue is a binding site. The present application can achieve accurate and reliable identification of binding sites of a protein.

Description

A method and related equipment for identifying protein-nucleic acid binding residues Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and related equipment for identifying protein-nucleic acid binding residues. Background Technology

[0002] Sequence-based protein-nucleic acid binding residue identification methods rely on the primary structure of proteins (i.e., the amino acid sequence). These methods typically predict residues by extracting local or global features from the protein sequence. In recent years, with the construction of large-scale protein sequence databases and advancements in deep learning technology, sequence-based prediction methods have developed rapidly.

[0003] Early sequence analysis methods were typically based on traditional machine learning models such as Support Vector Machines (SVM), Random Forests (RF), and Extreme Gradient Boosting (XGBoost). These methods primarily rely on the physicochemical properties and amino acid characteristics of proteins to construct feature vectors and then use supervised learning to classify these features. For example, Yan et al. proposed an SVM model based on amino acid physicochemical properties that can identify key binding residues in proteins. The advantages of these methods are their relative simplicity and the fact that they do not require three-dimensional structural information of proteins, resulting in lower computational costs. Furthermore, sequence data is easier to obtain than three-dimensional structural data because it can be directly obtained through sequencing technology. However, the main limitation of these methods is their inability to capture the three-dimensional spatial information of proteins, neglecting the spatial location of residues and their interactions with nucleic acids. Therefore, their predictive performance is often insufficient in scenarios involving complex binding sites.

[0004] With the improvement of large-scale data and computing power, deep learning technology has begun to be widely used in sequence analysis tasks. For example, large-scale pre-trained models proposed in recent years, such as ESM (Evolutionary Scale Modeling) and ProtBERT[8], have significantly improved the performance of sequence analysis. These models can better identify important features in the sequence by learning deep contextual information from a large number of protein sequences. The ESM model proposed by Rao et al. uses the Transformer structure to model protein sequences and learns rich semantic information from millions of protein sequences, thereby effectively capturing the positional features of binding residues. Similarly, ProtBERT enables the model to understand the interaction between different amino acids by encoding the context of amino acid sequences. However, these sequence-based deep learning methods still face some limitations. First, they cannot make full use of the three-dimensional structural information of proteins, while the three-dimensional structure is crucial for the spatial relationship of residues. Second, in some cases, it is difficult to distinguish the small differences in binding sites based on sequence information alone, especially in highly conserved sequence regions. Therefore, although sequence-based deep learning methods perform well in specific scenarios, they still have bottlenecks when predicting complex binding sites.

[0005] To improve predictive performance, many researchers have attempted to enhance sequence-based models by incorporating protein evolutionary information (such as PSSM and HMM profiles). Evolutionary information reflects the conservation of proteins across different species and can help identify important binding sites. Jones et al. used evolutionary information to improve sequence-based residue identification models and significantly improved prediction accuracy by combining sequence features and evolutionary features. However, obtaining evolutionary information typically relies on multiple sequence alignments (MSAs), which is not only computationally expensive but also insufficient in some cases (such as when there are few protein families). Furthermore, evolutionary information is primarily based on sequence similarity and has a weak ability to capture spatial geometric features.

[0006] This shows that traditional sequence-based protein-nucleic acid binding residue identification methods suffer from low identification accuracy. Summary of the Invention

[0007] The purpose of this application is to propose a method and related equipment for identifying protein-nucleic acid binding residues, so as to solve the problem of low identification accuracy in traditional sequence-based protein-nucleic acid binding residue identification methods.

[0008] To address the aforementioned technical problems, this application provides a method for recognizing protein-nucleic acid binding residues, employing the technical solution described below:

[0009] Obtain the amino acid sequence and three-dimensional structure data of the protein to be identified;

[0010] The amino acid sequence is input into a trained protein language model to perform sequence feature extraction to obtain sequence feature data.

[0011] The three-dimensional structural data is input into a sparse convolutional network in the form of point cloud to perform structural feature extraction, thereby obtaining structural feature data.

[0012] A weighted fusion operation is performed on the sequence feature data and the structural feature data to obtain a weighted fusion feature;

[0013] The weighted fusion features are input into a trained multilayer perceptron classifier for recognition, and the probability information of each residue as a binding site is obtained.

[0014] Furthermore, after the step of inputting the three-dimensional structural data in the form of a point cloud into a sparse convolutional network for structural feature extraction to obtain structural feature data, the following steps are also included:

[0015] The structural feature data is adaptively adjusted based on an attention mechanism, wherein the attention mechanism is expressed as follows:

[0016] Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

[0017] Furthermore, before the step of inputting the weighted fused features into the trained multilayer perceptron classifier for recognition to obtain the probability information of each residue as a binding site, the following step is also included:

[0018] The initial multilayer perceptron classifier is invoked, and the initial multilayer perceptron classifier is trained using a self-supervised learning method to obtain the trained multilayer perceptron classifier.

[0019] Furthermore, the loss function for the model training operation is expressed as:

[0020] Among them, z i and z j This represents the protein feature vector after different data augmentations, where τ is the temperature parameter.

[0021] Furthermore, the weighted fusion feature is represented as: F fusion =α·F seq +β·F struct

[0022] Where α and β are learnable weight parameters that control the proportion of contribution of sequence features and structural features to the final prediction, respectively. seq F represents the sequence feature data. struct This represents the structural feature data.

[0023] Furthermore, the probability information is represented as: p i =σ(W·F fusion +b)

[0024] Where W is the weight matrix, b is the bias term, and σ is the Sigmoid activation function, which is used to normalize the output value to the interval [0,1][0,1][0,1].

[0025] To address the aforementioned technical problems, this application also provides a protein-nucleic acid binding residue recognition device, which employs the following technical solution:

[0026] The data acquisition module is used to acquire the amino acid sequence and three-dimensional structure data of the protein to be identified.

[0027] The sequence feature extraction module is used to input the amino acid sequence into a trained protein language model to perform sequence feature extraction operations and obtain sequence feature data.

[0028] The structural feature extraction module is used to input the three-dimensional structural data in the form of point cloud into a sparse convolutional network to perform structural feature extraction operations and obtain structural feature data.

[0029] The weighted fusion module is used to perform a weighted fusion operation on the sequence feature data and the structural feature data to obtain weighted fused features;

[0030] The recognition module is used to input the weighted fusion features into a trained multilayer perceptron classifier for recognition operations, and obtain the probability information of each residue as a binding site.

[0031] Furthermore, the device also includes:

[0032] An adaptive adjustment module is used to perform adaptive adjustment operations on the structural feature data according to an attention mechanism, wherein the attention mechanism is expressed as:

[0033] Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

[0034] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0035] It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the protein-nucleic acid binding residue recognition method as described above.

[0036] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0037] The computer-readable storage medium stores computer-readable instructions that, when executed by a processor, implement the steps of the protein-nucleic acid binding residue recognition method as described above.

[0038] This application provides a method for identifying protein-nucleic acid binding residues, comprising: acquiring the amino acid sequence and three-dimensional structural data of the protein to be identified; inputting the amino acid sequence into a trained protein language model for sequence feature extraction to obtain sequence feature data; inputting the three-dimensional structural data in point cloud form into a sparse convolutional network for structural feature extraction to obtain structural feature data; performing a weighted fusion operation on the sequence feature data and the structural feature data to obtain weighted fusion features; and inputting the weighted fusion features into a trained multilayer perceptron classifier for identification to obtain probability information of each residue as a binding site. Compared with existing technologies, this application significantly improves prediction accuracy by combining protein sequence and three-dimensional structural information and introducing a cross-modal learning framework. Furthermore, this application introduces an atomic-level attention mechanism, enabling the model to dynamically capture geometric changes in the protein, thereby improving adaptability and robustness in complex scenarios. This application also enhances the model's generalization ability on small datasets through a self-supervised learning strategy and effectively addresses the challenges posed by data imbalance. In summary, this application achieves an overall performance improvement in protein-nucleic acid binding residue recognition through multimodal fusion, self-supervised learning, and dynamic structure adjustment techniques. Attached Figure Description

[0039] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 is an exemplary system architecture diagram in which this application can be applied;

[0041] Figure 2 is a flowchart illustrating the implementation of the protein-nucleic acid binding residue recognition method provided in the embodiments of this application;

[0042] Figure 3 is a schematic diagram of the overall flow of the model provided in the embodiments of this application;

[0043] Figure 4 is a schematic diagram of an embodiment of the quantitative performance comparison of the model provided in this application;

[0044] Figure 5 is a schematic diagram of an embodiment comparing the performance improvements of various parts of the model provided in this application;

[0045] Figure 6 is a schematic diagram of an embodiment of the performance comparison of the protein language model provided in this application;

[0046] Figure 7 is a schematic diagram of an embodiment of the comparison of atomic-level attention mechanism improvement provided in this application;

[0047] Figure 8 is a schematic diagram of one embodiment provided in this application.

[0048] Figure 9 is a schematic diagram of the protein-nucleic acid binding residue recognition device provided in an embodiment of this application;

[0049] Figure 10 is a schematic diagram of the structure of an embodiment of a computer device according to this application. Detailed Implementation

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0051] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0052] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0053] As shown in Figure 1, the system architecture 100 may include a terminal device 101, a network 102, and a server 103. The terminal device 101 may be a laptop computer 1011, a tablet computer 1012, or a mobile phone 1013. The network 102 is used as a medium to provide a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0054] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0055] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0056] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0057] It should be noted that the protein-nucleic acid binding residue recognition method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the protein-nucleic acid binding residue recognition device is generally set in the server / terminal device.

[0058] It should be understood that the number of terminal devices, networks, and servers shown in Figure 1 is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0059] Referring again to Figure 2, a flowchart of an embodiment of the protein-nucleic acid binding residue recognition method according to this application is shown. The protein-nucleic acid binding residue recognition method includes: steps S201, S202, S203, S204, and S205.

[0060] In step S201, the amino acid sequence and three-dimensional structure data of the protein to be identified are obtained.

[0061] In this embodiment of the application, the input data for the protein-nucleic acid binding residue prediction task mainly includes the amino acid sequence and three-dimensional structural information of the protein. The amino acid sequence of the protein is described by its primary structure, usually represented as a string sequence, where each character represents a specific amino acid. Specifically, the sequence of a protein is S = {a1, a2, ..., a...} L} is used to describe the amino acid sequence of a protein, where L represents the sequence length, a i This represents the i-th amino acid.

[0062] In this embodiment, the three-dimensional structure of the protein is described using a point cloud, where the three-dimensional coordinates of each atom are defined as (x, y, z). i ,y i ,z i ), where i represents the atom number. The three-dimensional structural information of proteins is crucial for understanding protein function and recognizing their binding with other molecules, and this representation can effectively capture the spatial relationships between protein residues.

[0063] In step S202, the amino acid sequence is input into the trained protein language model to perform sequence feature extraction to obtain sequence feature data.

[0064] In this embodiment, the pre-trained protein language model ESM2 (Evolutionary Scale Modeling) is used for protein sequence feature extraction. The ESM2 model is based on the Transformer architecture and captures context-dependent features in the protein sequence through its multi-layer self-attention mechanism. This mechanism enables the model to effectively extract deep information from the sequence, especially long-range dependencies. Given a protein sequence S = {a1, a2, ..., a...} L The feature vector generated by the model is represented as: F seq ={f1,f2,...,f L}

[0065] in, Let d represent the feature vector of the i-th amino acid, where d is the dimension of the feature vector. This feature extraction method allows the model to capture rich information from the sequence, providing abundant information for downstream binding site identification. In this way, the model can generate feature representations with high generalization ability for identifying protein functional regions, making it particularly suitable for detecting nucleic acid binding sites.

[0066] In step S203, the three-dimensional structural data is input into a sparse convolutional network in the form of point cloud to perform structural feature extraction, thereby obtaining structural feature data.

[0067] In this embodiment, to extract important features from the three-dimensional structure of the protein, a sparse convolutional network (SVC) is used to process the protein's point cloud data. Each atom in the point cloud is represented by a three-dimensional coordinate (x, y). i ,y i ,z i The ) symbol is used to describe the three-dimensional conformation of a protein. This representation allows the model to capture fine geometric information in localized regions of the protein.

[0068] Point cloud data processing can generate the following structural feature representations: F struct ={g1,g2,...,g N}

[0069] in, Let N represent the feature of the i-th atom, and N be the number of atoms in the protein. Sparse convolutional networks can efficiently capture the geometric characteristics of protein molecules by performing local convolution operations on point cloud data, playing a particularly important role in understanding the binding of proteins with other molecules.

[0070] In step S204, a weighted fusion operation is performed on the sequence feature data and the structural feature data to obtain the weighted fused features.

[0071] In this embodiment of the application, after extracting sequence features F seq and structural features F struct Subsequently, this application proposes a cross-modal feature fusion strategy. By weightedly fusing features from these two different modalities, the model can effectively combine information from the sequence and the three-dimensional structure to generate a comprehensive feature representation: F fusion =α·F seq +β·F struct

[0072] Here, α and β are learnable weight parameters that control the proportion of contribution of sequence features and structural features to the final prediction, respectively. Through this weighted fusion method, the model can dynamically adjust the balance between sequence and structural features according to the needs of different tasks, making the prediction results more accurate.

[0073] In the embodiments of this application, cross-modal feature fusion not only improves the model's efficiency in utilizing sequence and structural information, but also captures the interactions between these features, making the model more robust when dealing with complex protein-nucleic acid interactions.

[0074] In step S205, the weighted fusion features are input into the trained multilayer perceptron classifier for recognition to obtain the probability information of each residue as a binding site.

[0075] In this embodiment of the application, the fused feature F fusion The input is fed into a multilayer perceptron (MLP) for predicting whether a residue is a nucleic acid binding site. The MLP uses multiple nonlinear transformations to probabilistically predict whether each residue is a nucleic acid binding site. The specific prediction formula is: p i =σ(W·F fusion +b)

[0076] Where W is the weight matrix, b is the bias term, and σ is the sigmoid activation function, used to normalize the output value to the interval [0,1]. The output result p i This represents the probability that residue i is a binding site. This nonlinear transformation enables the model to effectively utilize the information obtained from feature fusion, thereby achieving better prediction results in complex biological data.

[0077] In this embodiment, referring to the technical flowchart shown in Figure 3, the amino acid sequence and three-dimensional structure data of the protein are input first, and sequence features and structural features are extracted respectively. Specifically, the amino acid sequence is processed by a pre-trained protein language model (such as ESM2) to generate context-dependent features for each amino acid; simultaneously, the three-dimensional structure of the protein is input into a sparse convolutional network in the form of a point cloud to extract three-dimensional geometric features. The extracted sequence features and structural features are weighted and fused through a cross-modal feature fusion module to form a unified feature representation. The fused feature vector is input into a multilayer perceptron (MLP) classifier to predict the probability that each residue is a binding site.

[0078] This application provides a method for identifying protein-nucleic acid binding residues, comprising: acquiring the amino acid sequence and three-dimensional structural data of the protein to be identified; inputting the amino acid sequence into a trained protein language model for sequence feature extraction to obtain sequence feature data; inputting the three-dimensional structural data in point cloud form into a sparse convolutional network for structural feature extraction to obtain structural feature data; performing a weighted fusion operation on the sequence feature data and structural feature data to obtain weighted fusion features; and inputting the weighted fusion features into a trained multilayer perceptron classifier for identification to obtain the probability information of each residue as a binding site. Compared with the prior art, this application introduces a cross-modal learning framework by combining the sequence and three-dimensional structural information of the protein, which significantly improves the prediction accuracy; in addition, this application introduces an atomic-level attention mechanism, enabling the model to dynamically capture the geometric changes of the protein, thereby improving its adaptability and robustness in complex scenarios; this application also enhances the model's generalization ability on small-scale datasets through a self-supervised learning strategy and effectively solves the challenge caused by data imbalance. In summary, this application achieves an overall performance improvement in protein-nucleic acid binding residue recognition through multimodal fusion, self-supervised learning, and dynamic structure adjustment techniques.

[0079] In some optional implementations of the embodiments of this application, after the step of inputting the three-dimensional structural data in the form of a point cloud into a sparse convolutional network for structural feature extraction to obtain structural feature data, the following steps are also included:

[0080] An adaptive adjustment operation is performed on the structural feature data based on an attention mechanism, where the attention mechanism is expressed as follows:

[0081] Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

[0082] In this embodiment, to further improve the prediction accuracy of the model, an atomic-wise attention (AWA) mechanism is introduced. The AWA mechanism enhances the recognition ability of binding sites by dynamically focusing on the interactions between different atoms in the protein molecule. This mechanism adaptively adjusts the weight of each atom in the model based on its position and features.

[0083] The attention calculation formula for AWA is as follows:

[0084] Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively.k This is the scaling factor. Through this attention mechanism, the model can focus on local structural changes at the binding site, thereby improving the accuracy and stability of residue prediction.

[0085] In some optional implementations of the embodiments of this application, before the step of inputting the weighted fused features into a trained multilayer perceptron classifier for recognition to obtain the probability information of each residue as a binding site, the following step is further included:

[0086] The initial multilayer perceptron classifier is invoked, and the model is trained using a self-supervised learning approach to obtain a trained multilayer perceptron classifier.

[0087] In some optional implementations of the embodiments of this application, the loss function for model training is expressed as:

[0088] Among them, z i and z j This represents the protein feature vector after different data augmentations, where τ is the temperature parameter.

[0089] In this application embodiment, to enhance the model's generalization ability, a self-supervised learning (SSL) strategy is introduced. SSL, by training on unlabeled data, can improve the model's adaptability to small sample sizes and imbalanced data conditions. Through random data augmentation (such as rotation, translation, scaling, etc.), the model generates protein samples from different perspectives and uses contrastive learning to maintain feature consistency.

[0090] The loss function for contrastive learning is defined as:

[0091] Among them, z i and z j This represents the protein feature vector after different data augmentations, where τ is the temperature parameter. By maximizing the similarity between similar samples, the model can effectively learn important protein structural features in small sample environments and improve its adaptability to new samples.

[0092] In practical applications, this application achieves significant breakthroughs in protein-nucleic acid binding residue identification by combining protein sequence features and three-dimensional structural features through a cross-modal learning framework. As shown in Figure 4, this application demonstrates significant advantages on the DNA 129 dataset. CrossBind achieves an F1-score of 0.602 on this dataset, while the current state-of-the-art method, GraphBind, only achieves an F1-score of 0.512, representing an improvement of 17.6%. Furthermore, CrossBind also shows a significant improvement in the Matthews correlation coefficient (MCC), increasing from 0.484 in GraphBind to 0.581, an increase of 20%. This series of significant performance improvements is mainly attributed to the multimodal feature fusion strategy adopted in this application. By integrating protein sequence information and three-dimensional structural features, the model can comprehensively capture key regions of protein-nucleic acid interactions. The AUPR (area under the precision-recall curve) has also been significantly improved, further reflecting the model's powerful processing capabilities in complex residue prediction tasks.

[0093] As shown in Figure 5, ablation experiments on the DNA 129 dataset validated the importance of each module in this application for the protein-nucleic acid binding residue identification task. Using the protein language model alone on this task was not ideal, and using only the atomic point cloud segmentation encoder resulted in even worse results. However, when sequence and structural features were combined, AUPRC improved by 5.7%, and MCC improved by 7.8%, significantly outperforming state-of-the-art methods. These results fully demonstrate the effectiveness of the proposed cross-modal modules in feature fusion. The AWA module plays a crucial role in incorporating local geometric knowledge from atoms to amino acids, significantly improving model performance. Removing the AWA module significantly reduced prediction results, further validating its importance in accurately locating binding sites. Furthermore, the introduction of the SSL module improved the robustness of the segmentation encoder in handling imbalanced data; removing this module led to a 2.2% decrease in AUPRC, demonstrating the necessity of self-supervised learning in enhancing data processing capabilities. The RPF module leverages prior biological knowledge to optimize cross-modal prediction results. Introducing traditional co-evolutionary features did not improve performance; in fact, it caused it to decline. This is because the pre-trained language model already covers most of the important feature information. Overall, the ablation study results, as shown in Figure 6, further confirm the irreplaceability and technical effectiveness of each module in the CrossBind method of this application.

[0094] The study also showed that using a smaller LLM (language model) structure significantly impacts task performance; for example, reducing the 33-layer Transformer to 12 layers leads to a 2.4% decrease in AUC. In contrast, the fine-tuned 33-layer LLM model improved AUPRC by 3.1%, demonstrating that fine-tuning a larger-scale LLM model helps to better capture linguistic information in protein sequences, which is beneficial for the accuracy of residue prediction.

[0095] In the design of the atomic attention mechanism, this application employs an MLP (Multilayer Perceptron)-based attention mechanism that integrates fine-grained local geometric knowledge between atoms and amino acids. As shown in Figure 7, simply averaging or stacking atomic features leads to a significant performance degradation because valuable local geometric knowledge is lost. Instead, atomic attention is replaced with a single self-attention layer on a cross-modal module. Although performance is slightly reduced, it still outperforms the method of simply stacking all atomic features. This demonstrates that single-layer self-attention may over-smooth the local geometric relationships between atoms, resulting in a less effective feature capture compared to a refined MLP mechanism.

[0096] Furthermore, this application also presents the qualitative results of its proposed model in predicting binding residues, as shown in Figure 8. This application visualizes three cases predicted on DNA 129 by GraphSite, GraphBind, and the model proposed in this application. This application selected the example protein 6YMW_B. This protein contains 668 residues, of which 13 are binding residues. As shown in Figure 8, the proposed model CrossBind predicted 10 true binding residues and 4 false positive residues, with a Rec of 0.64, a Pre of 0.53, and an F1 score of 0.58. In contrast, GraphSite predicted 8 true binding residues and 13 false positive residues, with a Rec of 0.60, a Pre of 0.47, and an F1 score of 0.52. On the other hand, GraphBind only predicted 6 true binding residues and 5 false positive residues, with a Rec of 0.40, a Pre of 0.26, and an F1 score of 0.32. While GraphSite also predicts a sufficient number of true binding residues, it has a high false positive rate, while CrossBind exhibits higher accuracy.

[0097] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0098] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0099] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0100] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0101] Referring further to Figure 9, as an implementation of the method shown in Figure 2 above, this application provides an embodiment of a protein-nucleic acid binding residue recognition device, which corresponds to the method embodiment shown in Figure 2, and the device can be specifically applied to various electronic devices.

[0102] As shown in Figure 9, the protein-nucleic acid binding residue recognition device 200 of this application embodiment includes:

[0103] The data acquisition module 210 is used to acquire the amino acid sequence and three-dimensional structure data of the protein to be identified;

[0104] The sequence feature extraction module 220 is used to input the amino acid sequence into the trained protein language model to perform sequence feature extraction operations and obtain sequence feature data.

[0105] The structural feature extraction module 230 is used to input three-dimensional structural data in the form of point cloud into a sparse convolutional network to perform structural feature extraction operations and obtain structural feature data.

[0106] The weighted fusion module 240 is used to perform weighted fusion operations on sequence feature data and structural feature data to obtain weighted fused features.

[0107] The recognition module 250 is used to input the weighted fusion features into the trained multilayer perceptron classifier for recognition operations, and obtain the probability information of each residue as a binding site.

[0108] In this embodiment, a protein-nucleic acid binding residue identification device 200 is provided, comprising: a data acquisition module 210 for acquiring the amino acid sequence and three-dimensional structural data of the protein to be identified; a sequence feature extraction module 220 for inputting the amino acid sequence into a trained protein language model for sequence feature extraction to obtain sequence feature data; a structural feature extraction module 230 for inputting the three-dimensional structural data in point cloud form into a sparse convolutional network for structural feature extraction to obtain structural feature data; a weighted fusion module 240 for performing a weighted fusion operation on the sequence feature data and the structural feature data to obtain weighted fusion features; and an identification module 250 for inputting the weighted fusion features into a trained multilayer perceptron classifier for identification to obtain probability information of each residue as a binding site. Compared with existing technologies, this application significantly improves prediction accuracy by introducing a cross-modal learning framework that combines protein sequence and three-dimensional structural information. Furthermore, it introduces an atomic-level attention mechanism, enabling the model to dynamically capture geometric changes in proteins, thereby improving adaptability and robustness in complex scenarios. This application also enhances the model's generalization ability on small datasets through a self-supervised learning strategy and effectively addresses the challenges posed by data imbalance. In summary, this application achieves an overall performance improvement in protein-nucleic acid binding residue recognition through multimodal fusion, self-supervised learning, and dynamic structure adjustment techniques.

[0109] In some optional implementations of the embodiments of this application, the protein-nucleic acid binding residue recognition device 200 further includes:

[0110] The adaptive adjustment module is used to perform adaptive adjustment operations on the structural feature data according to the attention mechanism, where the attention mechanism is represented as:

[0111] Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

[0112] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to Figure 10 for details; Figure 10 is a basic structural block diagram of the computer device according to an embodiment of this application.

[0113] The computer device 300 includes a memory 310, a processor 320, and a network interface 330 that are interconnected via a system bus. It should be noted that only the computer device 300 with components 310-330 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0114] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0115] The memory 310 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 310 may be an internal storage unit of the computer device 300, such as the hard disk or memory of the computer device 300. In other embodiments, the memory 310 may also be an external storage device of the computer device 300, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Of course, the memory 310 may also include both internal storage units and external storage devices of the computer device 300. In this embodiment, the memory 310 is typically used to store the operating system and various application software installed on the computer device 300, such as computer-readable instructions for protein-nucleic acid binding residue recognition methods. Furthermore, the memory 310 can also be used to temporarily store various types of data that have been output or will be output.

[0116] In some embodiments, the processor 320 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 320 is typically used to control the overall operation of the computer device 300. In this embodiment, the processor 320 is used to execute computer-readable instructions stored in the memory 310 or to process data, for example, to execute computer-readable instructions for the protein-nucleic acid binding residue recognition method.

[0117] The network interface 330 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 300 and other electronic devices.

[0118] The computer device provided in this application significantly improves prediction accuracy by combining protein sequence and three-dimensional structural information and introducing a cross-modal learning framework. Furthermore, it introduces an atomic-level attention mechanism, enabling the model to dynamically capture geometric changes in proteins, thereby improving adaptability and robustness in complex scenarios. It also enhances the model's generalization ability on small datasets through a self-supervised learning strategy and effectively addresses the challenges posed by data imbalance. In summary, through multimodal fusion, self-supervised learning, and dynamic structure adjustment techniques, the overall performance of protein-nucleic acid binding residue recognition is improved.

[0119] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the protein-nucleic acid binding residue recognition method as described above.

[0120] The computer-readable storage medium provided in this application significantly improves prediction accuracy by incorporating protein sequence and three-dimensional structural information into a cross-modal learning framework. Furthermore, it introduces an atomic-level attention mechanism, enabling the model to dynamically capture geometric changes in proteins, thereby enhancing its adaptability and robustness in complex scenarios. Additionally, it enhances the model's generalization ability on small datasets through a self-supervised learning strategy and effectively addresses the challenges posed by data imbalance. In summary, through multimodal fusion, self-supervised learning, and dynamic structure adjustment techniques, the overall performance of protein-nucleic acid binding residue recognition is improved.

[0121] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0122] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method for recognizing protein-nucleic acid binding residues, characterized in that, Includes the following steps: Obtain the amino acid sequence and three-dimensional structure data of the protein to be identified; The amino acid sequence is input into a trained protein language model to perform sequence feature extraction to obtain sequence feature data. The three-dimensional structural data is input into a sparse convolutional network in the form of point cloud to perform structural feature extraction, thereby obtaining structural feature data. A weighted fusion operation is performed on the sequence feature data and the structural feature data to obtain a weighted fusion feature; The weighted fusion features are input into a trained multilayer perceptron classifier for recognition, and the probability information of each residue as a binding site is obtained.

2. The protein-nucleic acid binding residue recognition method according to claim 1, characterized in that, After the step of inputting the three-dimensional structural data in the form of a point cloud into a sparse convolutional network for structural feature extraction to obtain structural feature data, the following steps are also included: The structural feature data is adaptively adjusted based on an attention mechanism, wherein the attention mechanism is expressed as follows: Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

3. The protein-nucleic acid binding residue recognition method according to claim 1, characterized in that, Before the step of inputting the weighted fused features into a trained multilayer perceptron classifier for recognition to obtain the probability information of each residue as a binding site, the following step is also included: The initial multilayer perceptron classifier is invoked, and the initial multilayer perceptron classifier is trained using a self-supervised learning method to obtain the trained multilayer perceptron classifier.

4. The protein-nucleic acid binding residue recognition method according to claim 3, characterized in that, The loss function for the model training operation is expressed as: Among them, z i and z j This represents the protein feature vector after different data augmentations, where τ is the temperature parameter.

5. The protein-nucleic acid binding residue recognition method according to claim 1, characterized in that, The weighted fusion feature is represented as: F fusion =α·F seq +β·F struct Where α and β are learnable weight parameters that control the proportion of contribution of sequence features and structural features to the final prediction, respectively. seq F represents the sequence feature data. struct This represents the structural feature data.

6. The protein-nucleic acid binding residue recognition method according to claim 1, characterized in that, The probability information is represented as: p i =σ(W·F fusion +b) Where W is the weight matrix, b is the bias term, and σ is the Sigmoid activation function, which is used to normalize the output value to the interval [0,1][0,1][0,1].

7. A protein-nucleic acid binding residue recognition device, characterized in that, include: The data acquisition module is used to acquire the amino acid sequence and three-dimensional structure data of the protein to be identified. The sequence feature extraction module is used to input the amino acid sequence into a trained protein language model to perform sequence feature extraction operations and obtain sequence feature data. The structural feature extraction module is used to input the three-dimensional structural data in the form of point cloud into a sparse convolutional network to perform structural feature extraction operations and obtain structural feature data. The weighted fusion module is used to perform a weighted fusion operation on the sequence feature data and the structural feature data to obtain weighted fused features; The recognition module is used to input the weighted fusion features into a trained multilayer perceptron classifier for recognition operations, and obtain the probability information of each residue as a binding site.

8. The protein-nucleic acid binding residue recognition device according to claim 7, characterized in that, The device further includes: An adaptive adjustment module is used to perform adaptive adjustment operations on the structural feature data according to an attention mechanism, wherein the attention mechanism is expressed as: Where, q i and k j Let d represent the query vector and bond vector of atom i and its neighboring atom j, respectively. k This is the scaling factor.

9. A computer device, comprising a memory and a processor, characterized in that, The memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, it implements the steps of the protein-nucleic acid binding residue recognition method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions that, when executed by a processor, implement the steps of the protein-nucleic acid binding residue recognition method as described in any one of claims 1 to 6.

Images