Cas protein recognition method and system based on protein language model and multi-scale convolutional neural network

Abstract

The present application relates to a Cas protein recognition method and system based on a protein language model and a multi-scale convolutional neural network. The method comprises: obtaining different Cas protein data sets collected, the data set containing protein sequences; encoding the input protein sequence through an AATP feature channel, representing the protein sequence as AATP features; encoding the input protein sequence through an ESM-2 feature channel, extracting protein embedding features from the protein sequence; fusing the compressed AATP features and ESM-2 features to obtain fused features; and training a Cas protein recognition model based on the fused features. By using two feature channels to encode the input protein sequence respectively, extracting two features and then performing compression and dimensionality reduction fusion processing, the trained Cas protein recognition model can quickly and efficiently screen out possible Cas proteins with low cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bioinformatics, in particular to a Cas protein recognition method and system based on a protein language model and a multi-scale convolutional neural network. BACKGROUND

[0002] Cas protein is a key component of the CRISPR / Cas system, which is a natural adaptive immune system widely existing in prokaryotes. The CRISPR / Cas system is composed of CRISPR sequences and various Cas proteins, and Cas proteins are responsible for performing specific biological functions. The recognition of Cas protein mainly relies on traditional methods based on sequence homology, such as BLAST and HMMER. These methods identify Cas proteins by detecting conserved regions similar to known sequences, which are suitable for analyzing proteins closely related to sequences in the database. In addition to relying on traditional methods based on sequence homology, other traditional machine learning methods still rely on manually designed features, such as CASPredict, CRISPRCasStack, and CASPredict-SVM, which are difficult to fully exploit deep features in sequences.

[0003] However, for Cas proteins or new Cas proteins found in poorly studied bacteria and archaea, these methods show obvious limitations, with long time consumption, limited recognition accuracy, and difficulty in achieving high-throughput applications. CASPredict, CRISPRCasStack, and CASPredict-SVM, which recognize Cas proteins, cannot achieve end-to-end automatic feature extraction and classification prediction, increasing the complexity of practical applications.

[0004] Therefore, the traditional Cas protein recognition method often has the problems of low recognition efficiency and high recognition cost. SUMMARY

[0005] Therefore, in order to solve the above technical problems, a Cas protein recognition method and system based on a protein language model and a multi-scale convolutional neural network are provided, which can quickly and efficiently screen possible Cas proteins with low cost.

[0006] A Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network, the method comprising:

[0007] Collect different Cas protein datasets and divide the Cas protein datasets into two target datasets, and the target datasets contain protein sequences;

[0008] determining an AATP feature channel and an ESM-2 feature channel, encoding the input protein sequence through the AATP feature channel to represent the protein sequence as an AATP feature, and encoding the input protein sequence through the ESM-2 feature channel to extract a protein embedding feature from the protein sequence;

[0009] compressing the AATP feature using a neural network full connection layer to obtain a compressed AATP feature, and performing deep extraction on the protein embedding feature using a multi-scale convolutional neural network to obtain an ESM-2 feature;

[0010] fusing the compressed AATP feature and the ESM-2 feature to obtain a fused feature, training a Cas protein recognition model based on the fused feature, and performing Cas protein recognition through the trained Cas protein recognition model.

[0011] In one embodiment, different Cas protein datasets collected are obtained, and the Cas protein datasets are divided into two target datasets, including:

[0012] Different Cas protein datasets collected are obtained, and the Cas protein datasets are divided into a CRISPRCasStack dataset and a CASPredict dataset.

[0013] The protein sequences in the CRISPRCasStack dataset are clustered by a first similarity threshold;

[0014] The protein sequences in the CASPredict dataset are clustered by a second similarity threshold; wherein the CRISPRCasStack dataset is used for hyperparameter optimization of the model, and the CASPredict dataset and the CRISPRCasStack dataset are used together to evaluate the performance of the model under different training conditions.

[0015] In one embodiment, encoding the input protein sequence through the AATP feature channel to represent the protein sequence as an AATP feature includes:

[0016] The position-specific scoring matrix in the AATP feature channel captures the conservation and evolutionary information of different positions in the protein sequence;

[0017] Based on the values in the position-specific scoring matrix, the preference information of different types of amino acids in the protein sequence is extracted;

[0018] According to the preference information, the conservation and evolution information, AATP features are extracted from the position-specific scoring matrix and output.

[0019] In one of the embodiments, the input protein sequence is encoded by the ESM-2 feature channel, and protein embedding features in the protein sequence are extracted, including:

[0020] The protein sequence is input into the ESM-2 feature channel, and a protein language model ESM-2 is arranged in the ESM-2 feature channel;

[0021] The protein embedding features are extracted from the protein sequence by the protein language model.

[0022] In one of the embodiments, the AATP features are compressed by using a neural network full connection layer to obtain compressed AATP features, including:

[0023] The AATP features are extracted by using a two-layer neural network full connection layer;

[0024] The AATP features are processed by the neural network full connection layer to obtain compressed AATP features.

[0025] In one of the embodiments, the protein embedding features are deeply extracted by using a multi-scale convolutional neural network to obtain ESM-2 features, including:

[0026] The protein embedding features are input into a multi-scale convolutional neural network, and the protein embedding features are deeply extracted by different convolution scales in the multi-scale convolutional neural network to obtain deep features;

[0027] The deep features are compressed and dimensionally reduced to obtain ESM-2 features.

[0028] In one of the embodiments, the compressed AATP features and the ESM-2 features are fused to obtain fusion features, including:

[0029] The feature dimensions of the compressed AATP features and the feature dimensions of the ESM-2 features are converted to the same feature dimensions;

[0030] The compressed AATP features and the ESM-2 features after the feature dimension conversion are standardized to obtain target AATP features and target ESM-2 features;

[0031] The target AATP features and the target ESM-2 features are fused according to the feature dimensions to obtain fusion features.

[0032] In one of the embodiments, the Cas protein recognition model is trained based on the fusion features, including:

[0033] The fusion features are input into the Cas protein recognition model, and the loss is calculated based on the target data set using a cross-entropy loss function and a direction propagation strategy to determine the loss function;

[0034] The Cas protein recognition model is verified using a five-fold cross-validation method to obtain a verification result; and the hyperparameters of the Cas protein recognition model are determined according to the verification result;

[0035] The Cas protein recognition model is updated based on the loss function, and the Cas protein recognition model is trained in combination with the hyperparameters.

[0036] A Cas protein recognition system based on a protein language model and a multi-scale convolutional neural network, the system comprising:

[0037] A data collection module for collecting different Cas protein data sets and dividing the Cas protein data sets into two target data sets, and the target data sets containing protein sequences;

[0038] A protein sequence representation module for determining an AATP feature channel and an ESM-2 feature channel, encoding the input protein sequence through the AATP feature channel, and representing the protein sequence as AATP features; and encoding the input protein sequence through the ESM-2 feature channel and extracting protein embedding features from the protein sequence;

[0039] A feature extraction module for compressing the AATP features using a neural network full connection layer to obtain compressed AATP features; and using a multi-scale convolutional neural network to deeply extract the protein embedding features to obtain ESM-2 features;

[0040] A feature fusion and model training module for fusing the compressed AATP features and ESM-2 features to obtain fusion features; training a Cas protein recognition model based on the fusion features, and performing Cas protein recognition through the trained Cas protein recognition model.

[0041] The Cas protein recognition method and system based on the protein language model and the multi-scale convolutional neural network described above, by using a double feature channel to encode the input protein sequence, extracting two features, and then performing compression, dimension reduction and fusion processing, the trained Cas protein recognition model can quickly and efficiently screen out possible Cas proteins with low cost. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 An application environment diagram of a Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network in an embodiment;

[0043] Figure 2 A flowchart of a Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network in an embodiment;

[0044] Figure 3 A structural block diagram of a Cas protein recognition system based on a protein language model and a multi-scale convolutional neural network in an embodiment;

[0045] Figure 4 A structural framework diagram of a Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network in an embodiment;

[0046] Figure 5 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0047] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0048] It can be understood that the terms "first", "second", and the like used in the present application can be used herein to describe similarity thresholds, but these similarity thresholds are not limited by these terms. These terms are only used to distinguish the first similarity threshold from the other similarity threshold. For example, without departing from the scope of the present application, the first similarity threshold can be referred to as the second similarity threshold, and similarly, the second similarity threshold can be referred to as the first similarity threshold. Both the first similarity threshold and the second similarity threshold are similarity thresholds, but they are not the same similarity threshold.

[0049] The Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network provided by the embodiments of the present application can be applied in an application environment as shown in Figure 1 , for example. As shown in Figure 1As shown, the application environment includes a computer device 110. The computer device 110 can obtain a collected different Cas protein dataset, and divide the Cas protein dataset into two target datasets, and the target dataset contains a protein sequence; the computer device 110 can determine an AATP feature channel and an ESM-2 feature channel, encode the input protein sequence through the AATP feature channel, and represent the protein sequence as an AATP feature; encode the input protein sequence through the ESM-2 feature channel, and extract a protein embedding feature in the protein sequence; the computer device 110 can use a neural network full connection layer to compress the AATP feature, and obtain a compressed AATP feature; and use a multi-scale convolutional neural network to deeply extract the protein embedding feature, and obtain an ESM-2 feature; the computer device 110 can fuse the compressed AATP feature and the ESM-2 feature, and obtain a fused feature; train a Cas protein recognition model based on the fused feature, and perform Cas protein recognition through the trained Cas protein recognition model. The computer device 110 can be, but is not limited to, various personal computers, notebook computers, smart phones, robots, tablet computers and the like.

[0050] In one embodiment, as shown in Figure 2 A Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network is provided, including the following steps:

[0051] Step 202, obtaining a collected different Cas protein dataset, and dividing the Cas protein dataset into two target datasets, and the target dataset contains a protein sequence.

[0052] Among them, the different Cas protein datasets can be collected from previous studies. Specifically, the collected different Cas protein datasets are two different Cas protein datasets collected from two previous studies, which are CRISPRCasStack dataset and CASPredict dataset.

[0053] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network further includes a process of collecting a data set, and the specific process includes: obtaining different Cas protein data sets collected, dividing the Cas protein data sets into a CRISPRCasStack data set and a CASPredict data set; the protein sequences in the CRISPRCasStack data set are clustered by a first similarity threshold; the protein sequences in the CASPredict data set are clustered by a second similarity threshold; wherein the CRISPRCasStack data set is used for hyperparameter optimization of the model, and the CASPredict data set is used together with the CRISPRCasStack data set to evaluate the performance of the model under different training conditions.

[0054] The CRISPRCasStack data set contains 209 positive samples and artificially screened non-Cas protein negative samples, and the protein sequence clustering processing is performed by the CD-HIT tool, and the first similarity threshold is set to 70%, so as to ensure the low similarity of the sequence and exclude non-standard amino acid sequences, finally 150 positive samples and 150 negative samples are obtained for the training set, and 59 positive samples and 59 negative samples are obtained for the test set.

[0055] The CASPredict data set is based on the UniProt database, 293 Cas protein sequences are initially extracted, after removing the sequences containing ambiguous residues by manual inspection, further screening is performed by the CD-HIT tool with a second similarity threshold of 30%, and finally 155 high-quality positive samples are reserved. At the same time, non-Cas proteins are artificially screened from the UniProt database as negative samples, and CD-HIT is used to remove redundancy with a second similarity threshold of 40%, so as to ensure the diversity and length distribution of the negative samples, finally 155 positive samples and 155 negative samples are obtained as the training set, and 64 positive samples and 64 negative samples are obtained as the test set.

[0056] In this embodiment, the CRISPRCasStack data set is mainly used for hyperparameter optimization of the model, and the CASPredict data set is used together with the CRISPRCasStack data set to comprehensively evaluate the performance of the model under different training conditions. These data sets are strictly processed and screened, and high-quality and well-balanced training and test data are provided.

[0057] In step 204, the AATP feature channel and the ESM-2 feature channel are determined, the input protein sequence is encoded through the AATP feature channel, and the protein sequence is represented as AATP features; the input protein sequence is encoded through the ESM-2 feature channel, and the protein embedding features in the protein sequence are extracted.

[0058] In order to comprehensively capture the functional and evolutionary information of protein sequences, a dual-channel feature representation method is adopted in this embodiment to encode the input protein sequences through AATP feature channels and ESM-2 feature channels respectively.

[0059] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network can further include a process of extracting AATP features, and the specific process includes: capturing the conservation and evolutionary information of different positions in the protein sequence through the position-specific scoring matrix in the AATP feature channel; extracting the preference information of different types of amino acids in the protein sequence based on the values in the position-specific scoring matrix; and extracting the AATP features from the position-specific scoring matrix according to the preference information, the conservation and the evolutionary information and outputting the AATP features.

[0060] The position-specific scoring matrix (PSSM) is a matrix representing the evolutionary information of a protein sequence. The PSSM captures the conservation and evolutionary information of different positions in the sequence, and the AATP features can be derived from the PSSM through different calculation methods. The core concept of AATP is to analyze and summarize the values in the PSSM matrix to extract the preference information of different types of amino acids in the sequence. By statistically capturing the trend of the occurrence of a specific amino acid, AATP reduces data redundancy while retaining key sequence information related to amino acid preference. In the AATP feature channel, the protein sequence to be recognized is represented as 420-dimensional AATP features.

[0061] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network can further include a process of extracting protein embedding features, and the specific process includes: inputting the protein sequence into the ESM-2 feature channel, and setting a protein language model ESM-2 in the ESM-2 feature channel; and extracting protein embedding features from the protein sequence through the protein language model.

[0062] In the ESM-2 feature channel, the protein sequence is processed by the ESM-2 (Evolutionary Scale Modeling 2) protein large language model to generate 320-dimensional ESM-2 features. ESM-2 is a large-scale pre-training model based on the Transformer architecture, which is used to extract deep embedding features from protein sequences. In bioinformatics, protein sequences have structural similarity with text data, so the representation method based on language models can be directly applied to protein sequences. The pre-training model can learn semantic information from large-scale protein sequences to generate rich protein embeddings.

[0063] In step 206, the AATP features are compressed using a neural network full connection layer to obtain compressed AATP features, and the protein embedding features are deeply extracted using a multi-scale convolutional neural network to obtain ESM-2 features.

[0064] In the two feature channels, different feature extraction strategies and different feature processing methods are used for feature processing.

[0065] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network further includes a process of processing the AATP features, and the specific process includes: extracting the AATP features using two layers of neural network full connection layers; and performing dimension reduction processing on the AATP features through the neural network full connection layers to obtain compressed AATP features.

[0066] For the generated 420-dimensional AATP features, the AATP features are further extracted and dimensionally reduced using two layers of neural network full connection layers to compress the AATP features to 64 dimensions, thereby obtaining compressed AATP features.

[0067] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network further includes a process of extracting the protein embedding features, and the specific process includes: inputting the protein embedding features into the multi-scale convolutional neural network, deeply extracting the protein embedding features through different convolution scales in the multi-scale convolutional neural network to obtain deep features; and performing compression and dimension reduction processing on the deep features to obtain ESM-2 features.

[0068] For the 320-dimensional protein embedding features, the multi-scale convolutional neural network can be used to perform deep feature extraction on the protein embedding features through different convolution scales (9, 11, and 13), and finally convert the features to 256-dimensional features, i.e., ESM-2 features.

[0069] In step 208, the compressed AATP features and the ESM-2 features are fused to obtain fused features, a Cas protein recognition model is trained based on the fused features, and Cas protein recognition is performed through the trained Cas protein recognition model.

[0070] The two features of the two channels extracted through the feature extraction are fused and combined into 300-dimensional fused features, and then the features are input into two layers of full connection layers for prediction, and finally the prediction results of different protein sequences can be obtained.

[0071] Specifically, in one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network further includes a feature fusion process, and the specific process includes: converting the feature dimensions of the compressed AATP features and the ESM-2 features to the same feature dimension; performing standardization processing on the compressed AATP features and the ESM-2 features after the conversion of the feature dimensions, to obtain target AATP features and target ESM-2 features; and fusing the target AATP features and the target ESM-2 features according to the feature dimensions to obtain fused features.

[0072] In one embodiment, the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network further includes a model training process, and the specific process includes: inputting the fused features into the Cas protein recognition model, calculating a loss based on a target data set using a cross-entropy loss function and a direction propagation strategy, determining the loss function; verifying the Cas protein recognition model using a five-fold cross-validation method to obtain a verification result; determining the hyperparameters of the Cas protein recognition model according to the verification result; and updating the parameters of the Cas protein recognition model based on the loss function, and training the Cas protein recognition model in combination with the hyperparameters.

[0073] In this embodiment, on the basis of the target data set, the five-fold cross-validation method is used to update the function to calculate the loss in combination with the cross-entropy loss function and the direction propagation strategy, the hyperparameters of the Cas protein recognition model are determined by observing the performance of the verification set divided in the five-fold cross-validation, and then the training set of the target data set is used for complete model training to update the model parameters, and finally the Cas protein recognition model is determined. Then, the Cas protein recognition model is set to a verification mode and no longer participates in the direction propagation, different protein sequences are input, and the prediction of the Cas protein is completed.

[0074] In one embodiment, the same test set is used on the two target data sets respectively, and the effects obtained on different models are compared as follows:

[0075] On the CRISPRCasStack data set, the same test set is used, and the effects of the HMMCAS, CASPredict and CRISPRCasStack models are compared, and the effect data are shown in the following table:

[0076]

[0077] On the CASPredict data set, the same test set is used, and the CASPredict-SVM model and the Cas protein recognition model are compared, and the comparison results are as follows:

[0078]

[0079] Benefiting from the pre-training characteristics of the protein language model ESM-2 and the multi-scale feature learning ability of the multi-scale convolutional neural network MSCNN and the performance ability of the generated amino acid type feature AATP in the Cas protein, the Cas protein recognition model in the present application has a significant recognition advantage in recognizing Cas proteins.

[0080] It should be understood that although each step in the above flowchart is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in each of the above flowcharts can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0081] In one embodiment, as shown in Figure 3 A Cas protein recognition system based on a protein language model and a multi-scale convolutional neural network is provided, comprising: a data collection module 310, a protein sequence representation module 320, a feature extraction module 330, and a feature fusion and model training module 340, wherein:

[0082] The data collection module 310 is configured to obtain different Cas protein data sets collected, and divide the Cas protein data sets into two target data sets, and the target data sets contain protein sequences.

[0083] The protein sequence representation module 320 is configured to determine AATP feature channels and ESM-2 feature channels, encode the input protein sequence through the AATP feature channel, and represent the protein sequence as AATP features; and encode the input protein sequence through the ESM-2 feature channel, and extract protein embedding features from the protein sequence.

[0084] The feature extraction module 330 is configured to compress the AATP features using a neural network full connection layer to obtain compressed AATP features; and use a multi-scale convolutional neural network to deeply extract the protein embedding features to obtain ESM-2 features.

[0085] The feature fusion and model training module 340 is configured to fuse the compressed AATP features and the ESM-2 features to obtain fused features, train a Cas protein recognition model based on the fused features, and perform Cas protein recognition through the trained Cas protein recognition model.

[0086] In one embodiment, the Cas protein recognition system based on the protein language model and the multi-scale convolutional neural network can be applied in a structural framework as shown in Figure 4 The structural framework can include A, a Cas protein recognition model structure (Overview of CasPro-ESM2 Model), B, protein language model and multi-scale convolutional neural network feature extraction (ESM-2 MSCNN Feature Extraction), C, model interpretability (Model Interpretability), and D, web server architecture (Web Server Architecture), as shown in Figure 4 The protein sequence is encoded through an AATP feature channel to represent the protein sequence as AATP features, and is encoded through an ESM-2 feature channel to extract protein embedding features in the protein sequence. The AATP features are compressed using a neural network full connection layer to obtain compressed AATP features, and the protein embedding features are deeply extracted using a multi-scale convolutional neural network to obtain ESM-2 features. The compressed AATP features and the ESM-2 features are fused to obtain fused features, and a Cas protein recognition model is trained based on the fused features to obtain a trained Cas protein recognition model.

[0087] In one embodiment, the data collection module 310 is further configured to obtain different Cas protein data sets collected, divide the Cas protein data sets into a CRISPRCasStack data set and a CASPredict data set, cluster the protein sequences in the CRISPRCasStack data set through a first similarity threshold, cluster the protein sequences in the CASPredict data set through a second similarity threshold, wherein the CRISPRCasStack data set is used for hyperparameter optimization of the model, and the CASPredict data set and the CRISPRCasStack data set are used together to evaluate the performance of the model under different training conditions.

[0088] In an embodiment, the protein sequence representation module 320 is further configured to capture the conservation and evolutionary information of different positions in the protein sequence through a position-specific scoring matrix in the AATP feature channel; extract the preference information of different types of amino acids in the protein sequence based on the values in the position-specific scoring matrix; and extract the AATP features from the position-specific scoring matrix according to the preference information, the conservation and evolutionary information, and output the AATP features.

[0089] In an embodiment, the protein sequence representation module 320 is further configured to input the protein sequence into an ESM-2 feature channel, and the ESM-2 feature channel is provided with a protein language model ESM-2; and extract protein embedding features from the protein sequence through the protein language model.

[0090] In an embodiment, the feature extraction module 330 is further configured to extract the AATP features using a two-layer neural network full connection layer; and perform dimension reduction processing on the AATP features through the neural network full connection layer to obtain compressed AATP features.

[0091] In an embodiment, the feature extraction module 330 is further configured to input the protein embedding features into a multi-scale convolutional neural network, perform deep extraction on the protein embedding features through different convolution scales in the multi-scale convolutional neural network to obtain deep features, and perform compression and dimension reduction processing on the deep features to obtain ESM-2 features.

[0092] In an embodiment, the feature fusion and model training module 340 is further configured to convert the feature dimensions of the compressed AATP features and the ESM-2 features to the same feature dimension; perform standardization processing on the compressed AATP features and the ESM-2 features after the conversion of the feature dimensions to obtain target AATP features and target ESM-2 features; and fuse the target AATP features and the target ESM-2 features according to the feature dimensions to obtain fusion features.

[0093] In an embodiment, the feature fusion and model training module 340 is further configured to input the fusion features into a Cas protein recognition model, calculate a loss based on a target data set using a cross-entropy loss function and a direction propagation strategy to determine the loss function; verify the Cas protein recognition model using a five-fold cross-validation method to obtain a verification result; determine hyperparameters of the Cas protein recognition model according to the verification result; and perform parameter updating on the Cas protein recognition model based on the loss function, and train the Cas protein recognition model in combination with the hyperparameters.

[0094] In an embodiment, a computer device is provided, which can be a terminal, and an internal structure diagram of the computer device can be as shown in Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0095] Those skilled in the art can understand that, Figure 5 The skilled in the art can understand that,

[0096] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network.

[0097] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by the processor to implement the steps of the Cas protein recognition method based on the protein language model and the multi-scale convolutional neural network.

[0098] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, databases, or other media in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0099] The technical features of the above embodiments can be combined in any way. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist, they should be considered as the scope of the present application.

[0100] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of the patent of the present application should be subject to the appended claims.

Claims

1. A Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network, characterized by, The method comprises: acquiring different Cas protein data sets collected and dividing the Cas protein data sets into two target data sets, and the target data sets containing protein sequences; determining an AATP feature channel and an ESM-2 feature channel, encoding the input protein sequences through the AATP feature channel, and representing the protein sequences as AATP features, including: capturing the conservation and evolution information of different positions in the protein sequences through a position-specific scoring matrix in the AATP feature channel; based on the values in the position-specific scoring matrix, extracting the preference information of different types of amino acids in the protein sequences; according to the preference information, the conservation and evolution information, extracting AATP features from the position-specific scoring matrix and outputting; encoding the input protein sequences through the ESM-2 feature channel, and extracting protein embedding features in the protein sequences, including: inputting the protein sequences into the ESM-2 feature channel, and a protein language model ESM-2 is arranged in the ESM-2 feature channel; extracting protein embedding features from the protein sequences through the protein language model; using a neural network full connection layer to compress the AATP features to obtain compressed AATP features, including: using a two-layer neural network full connection layer to extract the AATP features; performing dimension reduction processing on the AATP features through the neural network full connection layer to obtain compressed AATP features; and using a multi-scale convolutional neural network to deeply extract the protein embedding features to obtain ESM-2 features, including: inputting the protein embedding features into a multi-scale convolutional neural network, deeply extracting the protein embedding features through different convolution scales in the multi-scale convolutional neural network to obtain deep features; performing compression and dimension reduction processing on the deep features to obtain ESM-2 features; fusing the compressed AATP features and the ESM-2 features to obtain fusion features; training a Cas protein recognition model based on the fusion features, and recognizing Cas proteins through the trained Cas protein recognition model.

2. The Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network according to claim 1, characterized in that, acquiring different Cas protein data sets collected and dividing the Cas protein data sets into two target data sets, including: acquiring different Cas protein data sets collected, and dividing the Cas protein data sets into a CRISPRCasStack data set and a CASPredict data set; the protein sequences in the CRISPRCasStack data set are subjected to clustering processing through a first similarity threshold; the protein sequences in the CASPredict data set are subjected to clustering processing through a second similarity threshold; The CRISPRCasStack dataset is used for hyperparameter optimization of the model, and the CASPredict dataset is used together with the CRISPRCasStack dataset to evaluate the performance of the model under different training conditions.

3. The Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network according to claim 1, characterized in that, The compressed AATP features and ESM-2 features are fused to obtain fused features, including: The feature dimensions of the compressed AATP features and the ESM-2 features are converted to the same feature dimension; The compressed AATP features and ESM-2 features after feature dimension conversion are standardized to obtain target AATP features and target ESM-2 features; The target AATP features and target ESM-2 features are fused according to the feature dimension to obtain fused features.

4. The Cas protein recognition method based on a protein language model and a multi-scale convolutional neural network according to claim 1, characterized in that, Based on the fused features, a Cas protein recognition model is trained, including: The fused features are input into the Cas protein recognition model, and the loss function is calculated based on the target dataset using the cross-entropy loss function and the direction propagation strategy to determine the loss function; The Cas protein recognition model is verified using a five-fold cross-validation method to obtain a verification result; the hyperparameters of the Cas protein recognition model are determined according to the verification result; Based on the loss function, the parameters of the Cas protein recognition model are updated, and the Cas protein recognition model is trained in combination with the hyperparameters.

5. A Cas protein recognition system based on a protein language model and a multi-scale convolutional neural network, characterized in that, The system includes: A data collection module for collecting different Cas protein datasets and dividing the Cas protein datasets into two target datasets, and the target datasets containing protein sequences; A protein sequence representation module for determining AATP feature channels and ESM-2 feature channels, encoding the input protein sequence through the AATP feature channels, and representing the protein sequence as AATP features, including: capturing the conservation and evolution information of different positions in the protein sequence through the position-specific scoring matrix in the AATP feature channel; based on the values in the position-specific scoring matrix, extracting the preference information of different types of amino acids in the protein sequence; according to the preference information, conservation and evolution information, extracting AATP features from the position-specific scoring matrix and outputting; encoding the input protein sequence through the ESM-2 feature channel, and extracting protein embedding features in the protein sequence, including: inputting the protein sequence into the ESM-2 feature channel, and setting a protein language model ESM-2 in the ESM-2 feature channel; extracting protein embedding features from the protein sequence through the protein language model; The feature extraction module is configured to compress the AATP features using a neural network full connection layer to obtain compressed AATP features, including: extracting the AATP features using a two-layer neural network full connection layer; performing dimension reduction processing on the AATP features through the neural network full connection layer to obtain compressed AATP features; and performing deep extraction on the protein embedding features using a multi-scale convolutional neural network to obtain ESM-2 features, including: inputting the protein embedding features into a multi-scale convolutional neural network, performing deep extraction on the protein embedding features through different convolution scales in the multi-scale convolutional neural network to obtain deep features; and performing compression and dimension reduction processing on the deep features to obtain ESM-2 features. The feature fusion and model training module is configured to fuse the compressed AATP features and the ESM-2 features to obtain fused features, train a Cas protein recognition model based on the fused features, and perform Cas protein recognition through the trained Cas protein recognition model.

Images