Transmembrane protein-to-transmembrane protein interaction prediction method, device and medium

Through the transmembrane prediction neural network and attention mechanism model, the interaction between transmembrane protein and transmembrane protein is automatically predicted, which solves the problem of high human and material consumption in the existing technology, and achieves efficient and accurate protein interaction prediction.

CN120496642APending Publication Date: 2025-08-15SOMESTECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510478762.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-16
Publication Date
2025-08-15

AI Technical Summary

Technical Problem

In the prior art, the prediction of protein interactions mainly depends on biological experiments, consumes a lot of manpower and material resources, and is low in efficiency, making it difficult to efficiently and accurately determine the interaction between transmembrane proteins and transmembrane proteins.

Method used

The transmembrane prediction neural network and attention mechanism model are used to automatically determine the interaction between transmembrane protein and transmembrane protein. The transmembrane prediction neural network is used to screen out the external amino acid fragment sequence of the cell membrane, and the trained attention mechanism model is used to calculate the interaction score to improve prediction efficiency.

Benefits of technology

The interaction between transmembrane protein and transmembrane protein can be accurately predicted without a large amount of manpower and material resources, which improves the determination efficiency, reduces experimental costs, and improves the accuracy and efficiency of prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120496642A_ABST
    Figure CN120496642A_ABST
Patent Text Reader

Abstract

The invention relates to a transmembrane protein and transmembrane protein interaction prediction method, computer equipment and a storage medium, and the method comprises the steps: inputting an amino acid sequence 1 of protein 1 and an amino acid sequence 2 of protein 2 into a transmembrane prediction neural network, and when an output result identifies that the two proteins are transmembrane proteins, determining the transmembrane protein interaction between the two proteins; screening an extracellular amino acid fragment sequence 1 of the protein 1 and an extracellular amino acid fragment sequence 2 of the protein 2 from output results of the transmembrane prediction neural network; the extracellular amino acid fragment sequence 1 and the extracellular amino acid fragment sequence 2 are input into a trained attention mechanism model, an interaction score is obtained, and the interaction score is used for representing the interaction possibility of the protein 1 and the protein 2. According to the method, the interaction between the transmembrane protein and the transmembrane protein is automatically determined in a computer mode, and excessive manpower and material resources do not need to be consumed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of biological information processing technology, and in particular to a method for predicting the interaction between transmembrane proteins, a computer device, and a storage medium. Background Art

[0002] Accurate prediction of membrane protein interactions plays a crucial role in biology, drug design, and molecular biology. With the deepening of protein interaction research, predicting and understanding the interactions between membrane proteins has become a hot topic. The study of protein interactions is also of great value in drug design. Understanding the binding and interaction patterns between proteins can help design more precise and effective drugs. Research in these areas has far-reaching implications for biological research, drug development, and personalized medicine.

[0003] When discussing protein interactions, the research fields involved span biology, medicine, and drug design. Protein-protein interactions underlie cellular function, signal transduction, and numerous physiological processes within organisms. In medicine, a precise understanding of protein interactions is crucial for understanding disease mechanisms. For example, the development of certain diseases is associated with dysregulated or abnormal protein-protein interactions. These studies are not only of great significance in basic science but also have profound implications for clinical medicine and drug development.

[0004] The medical field urgently needs to accurately predict protein interactions. However, current methods for testing protein interactions rely primarily on biological experiments. This not only consumes excessive manpower and resources, but also results in low efficiency. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, computer equipment and storage medium for predicting the interaction between transmembrane proteins to address the above technical problems, so as to automatically determine the interaction between transmembrane proteins by computer without consuming too much manpower and material resources, and also to improve the efficiency of determining the interaction between transmembrane proteins.

[0006] A method for predicting interactions between transmembrane proteins comprises: obtaining an amino acid sequence of a first protein and an amino acid sequence of a second protein to be predicted; inputting the amino acid sequence of the first protein and the amino acid sequence of the second protein into a transmembrane prediction neural network respectively; when the first protein and the second protein are both identified as transmembrane proteins based on the output result of the transmembrane prediction neural network, screening out the extracellular membrane amino acid fragment sequence of the first protein from multiple amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and screening out the extracellular membrane amino acid fragment sequence of the second protein from multiple amino acid fragment sequences of the second protein output by the transmembrane prediction neural network; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the possibility of interaction between the first protein and the second protein, and the trained attention mechanism model is trained by the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the sample transmembrane proteins.

[0007] In one embodiment, the model training of the attention mechanism model includes: obtaining the extracellular membrane amino acid fragment sequence of a first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of a second sample single transmembrane protein; obtaining the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein, wherein the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein is determined by one or more biological research methods or information mining methods; inputting the extracellular membrane amino acid fragment sequence of the first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of the second sample single transmembrane protein, as well as the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein into the self-attention mechanism model to perform model training on the self-attention mechanism model.

[0008] In one embodiment, when the first protein and the second protein are both identified as single-transmembrane proteins based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is 1, and the number of extracellular membrane amino acid fragment sequences of the second protein is 1; the interaction score between the first protein and the second protein is obtained by inputting 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model.

[0009] In one embodiment, when the first protein and the second protein are both identified as multi-transmembrane proteins based on the output result of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is multiple, and the number of extracellular membrane amino acid fragment sequences of the second protein is multiple; a method for predicting the interaction between transmembrane proteins also includes: combining each extracellular membrane amino acid fragment sequence of the first protein with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, the combined data set containing multiple combination elements, each combination element containing one extracellular membrane amino acid fragment sequence of the first protein and one extracellular membrane amino acid fragment sequence of the second protein; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, including: inputting each combination element into the trained attention mechanism model to obtain the interaction score of each combination element output by the trained attention mechanism model; outputting the highest score among the interaction scores of each combination element, the highest score being used to characterize the possibility of interaction between the first protein and the second protein.

[0010] In one embodiment, a method for predicting the interaction between transmembrane proteins further includes: determining multiple target combination elements whose score values are greater than a set threshold in the interaction scores of each combination element; obtaining the position of the extracellular membrane amino acid fragment sequence of the first protein in each combination element in the first protein, and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein; displaying the interaction score of each target combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein in each target combination element.

[0011] In one embodiment, a method for predicting the interaction between transmembrane proteins further includes: when the highest score is less than a score threshold, outputting information indicating that the interaction between the first protein and the second protein is weak.

[0012] In one embodiment, the attention mechanism model includes a word embedding layer, a position encoding layer, a multi-head attention layer and a multi-layer perceptron layer connected in sequence; wherein the multi-head attention layer includes an encoding layer composed of multiple connected encoders, and the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is input into the multi-layer perceptron layer after feature fusion.

[0013] In one embodiment, the method for determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model includes: configuring the attention mechanism model to be trained through preset hyperparameters; obtaining the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein; training the attention mechanism model to be trained through the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein, and adjusting the hyperparameters of the attention mechanism model and the length of the extracellular membrane amino acid fragment sequence of the input sample transmembrane protein during the model training process; when the video memory overflow of the device running the model training is identified, determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model and the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model.

[0014] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the above-mentioned embodiments are implemented.

[0015] A computer-readable storage medium stores a computer program, which implements the steps of any of the above-mentioned method embodiments when executed by a processor.

[0016] The above-mentioned method, computer device and storage medium for predicting the interaction between transmembrane proteins obtain the amino acid sequence of the first protein to be predicted and the amino acid sequence of the second protein; the amino acid sequence of the first protein and the amino acid sequence of the second protein are respectively input into the transmembrane prediction neural network; when the first protein and the second protein are identified as transmembrane proteins based on the output result of the transmembrane prediction neural network, the extracellular membrane amino acid fragment sequence of the first protein is screened out from the multiple amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and the extracellular membrane amino acid fragment sequence of the second protein is screened out from the multiple amino acid fragment sequences of the second protein output by the transmembrane prediction neural network; the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein are input into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the possibility of the interaction between the first protein and the second protein, and the trained attention mechanism model is trained by the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein. Therefore, the interaction between transmembrane proteins can be automatically determined by computer without consuming too much manpower and material resources, and the efficiency of determining the interaction between transmembrane proteins is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG1 is a schematic diagram of a process for predicting the interaction between transmembrane proteins in one embodiment;

[0018] Figure 2 1 is a flow chart of a model training process of an attention mechanism model in one embodiment;

[0019] Figure 3 A schematic diagram of a process for applying a trained attention mechanism model to predict interaction scores of extracellular membrane amino acid fragment sequences in one embodiment;

[0020] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0022] This application provides a method for predicting the interaction between transmembrane proteins. Figure 1 As shown, in one embodiment, a method for predicting the interaction between transmembrane proteins comprises the following steps:

[0023] S102: Obtain the amino acid sequence of the first protein to be predicted and the amino acid sequence of the second protein.

[0024] In this example, a first protein and a second protein are two proteins to be tested for interaction. The amino acid sequence of the first protein and the amino acid sequence of the second protein are extracted. A protein's amino acid sequence is information that records the order of all the amino acid residues that make up a particular protein, from the N-terminus (amino terminus) to the C-terminus (carboxyl terminus). This ordered sequence of amino acids determines the primary structure of the protein and plays a decisive role in its three-dimensional structure and function.

[0025] S104, inputting the amino acid sequence of the first protein and the amino acid sequence of the second protein into the transmembrane prediction neural network respectively, and when it is identified based on the output result of the transmembrane prediction neural network that the first protein and the second protein are both transmembrane proteins, screening the extracellular membrane amino acid fragment sequence of the first protein from the multiple amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and screening the extracellular membrane amino acid fragment sequence of the second protein from the multiple amino acid fragment sequences of the second protein output by the transmembrane prediction neural network.

[0026] In this embodiment, the transmembrane prediction neural network is an open source model, which can be TMHMM software, such as DeepTMHMM software. The amino acid sequence of a first protein is input into the transmembrane prediction neural network, which outputs a transmembrane prediction result for the first protein and one or more amino acid fragment sequences based on their relative positions on the cell membrane. The amino acid sequence of a second protein is input into the transmembrane prediction neural network, which outputs a transmembrane prediction result for the second protein and one or more amino acid fragment sequences based on their relative positions on the cell membrane.

[0027] The predicted transmembrane protein is a transmembrane protein or a non-transmembrane protein. The one or more amino acid fragment sequences based on the relative position of the cell membrane include an amino acid fragment sequence outside the cell membrane, an amino acid fragment sequence inside the cell membrane, and an amino acid fragment sequence on the cell membrane.

[0028] S106: Input the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the likelihood of interaction between the first protein and the second protein. The trained attention mechanism model is trained using the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the sample transmembrane proteins.

[0029] In this embodiment, when step S104 determines that both the first protein and the second protein are transmembrane proteins, the extracellular amino acid fragment sequence of the first protein and the extracellular amino acid fragment sequence of the second protein are screened and input into the trained attention mechanism model. The trained attention mechanism model outputs an interaction score between the first protein and the second protein, and the interaction between the two is determined based on the interaction score. The trained attention mechanism model is trained using the extracellular amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein. Each sample transmembrane protein used to train the attention mechanism model is a single transmembrane protein. Since a single transmembrane protein contains only one extracellular amino acid fragment sequence, the interaction between the extracellular amino acid fragment sequences of two single transmembrane proteins can be macroscopically manifested as the interaction between two single transmembrane proteins. Based on this, the attention mechanism model can be trained using multiple sample single transmembrane proteins, so that the trained attention mechanism model can be used to identify the interaction between two extracellular amino acid fragment sequences.

[0030] In one embodiment, model training of the attention mechanism model includes: obtaining the extracellular membrane amino acid fragment sequence of a first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of a second sample single transmembrane protein; obtaining the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein, the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein being determined by one or more biological research methods or protein information inference methods; inputting the extracellular membrane amino acid fragment sequence of the first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of the second sample single transmembrane protein, as well as the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein into the self-attention mechanism model to perform model training on the self-attention mechanism model.

[0031] In this embodiment, the interaction score of the first sample single transmembrane protein and the second sample single transmembrane protein is determined by one or more biological research means or information mining means. For example, biological research means include gene co-expression, gene adjacency, gene fusion and gene co-occurrence, etc. Information mining means include text mining, etc. The interaction scores determined by various biological research means or information mining means are comprehensively calculated by setting standards, and the interaction scores of the two sample single transmembrane proteins are finally obtained. For example, the interaction score of the interaction between two single transmembrane proteins determined by the means of gene co-expression is 0.8, and the interaction score determined by the means of gene co-occurrence is 0.7, then it can be determined that the interaction score of the two single transmembrane proteins is 0.75.

[0032] Specifically, the training process of the trained attention mechanism model is as follows:

[0033] First, the amino acid sequences of multiple proteins and their experimentally demonstrated interaction scores are obtained from a protein database. The amino acid sequences are then processed using a deep neural network. This deep neural network breaks down the amino acid sequences into different fragments based on their location on the cell membrane and outputs these fragments. It also outputs a predicted likelihood of whether the protein is a transmembrane protein. For example, the output indicates that the protein is a transmembrane protein, along with the protein's amino acid sequences outside the cell membrane, inside the cell membrane, and on the cell membrane.

[0034] During the training phase of the attention mechanism model, transmembrane proteins and their extracellular amino acid fragment sequences are screened as described above. Furthermore, experimentally proven interaction scores for each transmembrane protein are collected. This yields the extracellular amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein for the attention mechanism model. Furthermore, these extracellular amino acid fragment sequences and the interaction scores between each sample transmembrane protein are used to train the attention mechanism model.

[0035] In an example of this embodiment, the attention mechanism model includes a word embedding layer, a position encoding layer, a multi-head attention layer and a multi-layer perceptron layer connected in sequence; wherein the multi-head attention layer includes an encoding layer composed of multiple connected encoders, and the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is input into the multi-layer perceptron layer after feature fusion.

[0036] Specifically, the word embedding layer processes the two input amino acid fragment sequences separately to obtain masked representations of the amino acid fragment sequences. The position encoding layer processes the masked representations of the amino acid fragment sequences to obtain a representation of the masked amino acid fragment sequences containing the amino acid position order. The encoding layer composed of multiple connected encoders in the multi-head attention layer represents the masked amino acid fragment sequences containing the amino acid position order to obtain a self-attention representation of the amino acid fragment sequences. The self-attention representations of the two amino acid fragment sequences are connected together to obtain a feature fusion representation. The multi-layer perceptron layer of the attention mechanism model processes the feature fusion representation to generate a predicted interaction possibility score.

[0037] It should be emphasized that, compared with the traditional attention mechanism model, the attention mechanism model of the present application includes an encoding layer in the multi-head attention layer but does not include a decoding layer. That is, the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is directly subjected to feature fusion, without the need for decoding by the decoding layer and then feature fusion. Since the amino acid fragment sequence does not have the complex grammatical structure of natural language, it does not need to be decoded. The self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is directly subjected to feature fusion, which can also be perceived and recognized by the multi-layer perceptron layer, and has little impact on the final interaction score. Based on this, setting an encoding layer but not a decoding layer in the multi-head attention layer can reduce the model complexity of the attention mechanism model without affecting the final prediction results. When using the attention mechanism model to predict the interaction score between the first protein and the second protein, the amount of computation of the model can be reduced, and the prediction efficiency of the interaction score of the two proteins can be improved.

[0038] In an example of this embodiment, the method for determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model includes: configuring the attention mechanism model to be trained through preset hyperparameters; obtaining the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein; training the attention mechanism model to be trained through the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein, and adjusting the hyperparameters of the attention mechanism model and the length of the extracellular membrane amino acid fragment sequence of the input sample transmembrane protein during the model training process; when the video memory overflow of the device running the model training is identified, determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model and the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model.

[0039] Specifically, the attention mechanism model is set up according to a set of hyperparameters, and the attention mechanism model is trained using the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein. The model training process adjusts the various hyperparameters within the attention mechanism model and the length of the input extracellular membrane amino acid fragment sequence. Using the video memory of the device running the model as a constraint, the hyperparameters and the length of the extracellular membrane amino acid fragment sequence are repeatedly adjusted. When a video memory overflow occurs under acceptable prediction results, the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer, and multi-layer perceptron layer in the attention mechanism model, as well as the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model, are determined. In a specific training result, the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model was determined to be 300 amino acids. In the debugging result, the hyperparameters of the attention mechanism model are as follows:

[0040]

[0041]

[0042] The hyperparameters of the attention mechanism model are designed to reduce the memory requirements and prevent overfitting because proteins do not have as complex a grammar as natural languages.

[0043] Based on the attention mechanism model given in the above embodiment, the specific training process and application process are given below:

[0044] like Figure 2As shown, the word embedding layer performs a word embedding operation on the amino acid fragment sequence. The word embedding operation can include mapping the amino acid fragment sequence into a high-dimensional vector space. For example, the amino acid fragment sequence data is a one-dimensional sequence data, where each data element is an amino acid symbol. There are 20 standard amino acid symbols in total. The word embedding operation can map each amino acid symbol into a high-dimensional vector of a specified dimension.

[0045] The position encoding layer performs a position encoding operation on the amino acid sequence vector representation output by the word embedding layer. The position encoding operation involves embedding the original position information of the sequence into the input amino acid sequence vector representation. Position encoding ensures that the model can consider the position information of elements when processing sequence data, which is crucial for understanding amino acids that are far apart in protein sequences but may have important interactions. For example, in some embodiments, the configuration can process a maximum amino acid sequence length of 10,000, thereby generating 10,000 unknown position encodings, and then embedding the position encoding information into the vector representation of the amino acid sequence based on different input data. All modules are applied sequentially to the amino acid sequence vector representation.

[0046] Multi-head self-attention is used to learn about amino acid fragment sequences, allowing the model to simultaneously focus on information in different subspaces, or "heads," of the input sequence. Each attention head independently learns and weights the correlations between amino acids at different positions. Finally, the results of all attention heads are concatenated or merged to obtain a more comprehensive and detailed understanding of the entire amino acid fragment sequence. Therefore, an encoder configured to use multi-head self-attention can efficiently discover deep patterns and complex interactions in amino acid sequences, helping to predict protein structure and function.

[0047] A multilayer perceptron (MLP) can include at least three layers of nodes: an input layer, a hidden layer, and an output layer. Because the multilayer perceptron (MLP) is fully connected, each unit in each layer is interconnected with all units in the next layer through specific weights. Except for the input layer units, each unit is essentially a neuron component that uses a nonlinear activation function. The MLP uses a supervised learning algorithm called backpropagation for model training. It is precisely because of its multilayer structure and nonlinear activation mechanism that the MLP is clearly distinguished from the linear perceptron and demonstrates effective discrimination capabilities for nonlinearly separable data sets.

[0048] If we assume that all neurons in a multilayer perceptron (MLP) use only linear activation functions, that is, weighted inputs are mapped to the outputs of each neuron through a linear function, the principles of linear algebra point out that no matter how many layers there are, such a structure is equivalent to a simple two-layer input-output model. However, in actual MLPs, some neurons use nonlinear activation functions, which were originally designed to imitate the behavior of biological neurons in stimulating action potentials or adjusting discharge frequencies. During the training process, the learning process of MLP is achieved by comparing the error between the actual output and the expected output of each training data, and adjusting the connection weights between neurons based on this. This is an example of supervised learning, in which the learning mechanism is based on the principle of backpropagation. It is actually an extended application of the minimum mean square error algorithm in the linear perceptron. In this application, if Figure 2 As shown in Figure 1, the output of the MLP is the predicted interaction score between two amino acid sequences. During model training, the loss of its output is adjusted by the true label. The true label refers to a probability value between 0 and 1.

[0049] In a specific model training process:

[0050] The amino acid sequences of proteins and protein-protein interaction scores are collected from open source databases as sample training sets. The collected data are screened by deep neural networks to select single-transmembrane extracellular amino acid fragment sequences and their related protein-protein interaction scores, which contain amino acid sequences of approximately 1,100 proteins and approximately 49,000 interaction scores. In other training processes of the model, multi-transmembrane extracellular amino acid fragment sequences can also be selected for training. What type of extracellular amino acid fragment sequence is used to train the model, the trained model is applied to the test of the interaction score of the extracellular amino acid fragment sequence of that type. For the sample training set, 80% of the data is used for model training, 10% of the data is used to verify the results of model training, and the remaining 10% of the data provides results, as shown in the table below:

[0051] MSE RMSE R2 MAE EVS MAPE Max Error 0.0145 0.121 0.670 0.092 0.681 34.187% 0.72

[0052] MSE (Mean Square Error) is the average of the sum of squared errors;

[0053] RMSE (Root Mean Square Error) is the square root of MSE;

[0054] R2 coefficient of determination - the larger it is, the better the fit;

[0055] MAE (Mean Absolute Error) is the average of the absolute values of the errors;

[0056] EVS (Explained variance score);

[0057] MAPE (Mean Absolute Percentage Error).

[0058] In the application stage of the trained attention mechanism model, the amino acid sequences of each group of two proteins are used as input and passed through the transmembrane prediction neural network. When the output prediction result is that both are transmembrane proteins, at least two extracellular membrane amino acid fragment sequences output by the transmembrane prediction neural network are used to construct a pairwise combination data set of the two groups of fragments, and each extracellular membrane amino acid fragment sequence is input into the trained attention mechanism model one by one.

[0059] like Figure 3 As shown in FIG, the trained attention mechanism model receives at least two extracellular amino acid fragment sequences and generates a predicted score of the interaction score between the two extracellular amino acid fragment sequences. The predicted score of the interaction score is a probability numerical result between 0 and 1.

[0060] For the application stage of the trained attention mechanism model, several examples are given below:

[0061] In one embodiment, when the first protein and the second protein are both identified as single-transmembrane proteins based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is 1, and the number of extracellular membrane amino acid fragment sequences of the second protein is 1; the interaction score between the first protein and the second protein is obtained by inputting 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model.

[0062] In this embodiment, the first protein and the second protein are both single-transmembrane proteins. In this case, the number of extracellular membrane amino acid fragment sequences in the output result of the transmembrane prediction neural network is 1. The interaction score between the first protein and the second protein is directly determined by 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein. Therefore, after inputting 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model respectively, the interaction between the first protein and the second protein is determined by directly referring to the interaction score output by the trained attention mechanism model. Specifically, the output interaction score can be directly used to evaluate the interaction between the first protein and the second protein.

[0063] In another embodiment, when the first protein and the second protein are both identified as multi-transmembrane proteins based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is multiple, and the number of extracellular membrane amino acid fragment sequences of the second protein is multiple. At this time, before the step of inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, it also includes: combining each extracellular membrane amino acid fragment sequence of the first protein with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, the combined data set contains multiple combination elements, each combination element contains 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, including: inputting each combination element into the trained attention mechanism model to obtain the interaction score of each combination element output by the trained attention mechanism model; outputting the highest score among the interaction scores of each combination element, and the highest score is used to characterize the possibility of interaction between the first protein and the second protein.

[0064] In this embodiment, both the first and second proteins are multi-membrane proteins. In this case, the output of the transmembrane prediction neural network contains multiple extracellular amino acid fragment sequences for the first protein and multiple extracellular amino acid fragment sequences for the second protein. The multiple extracellular amino acid fragment sequences for the first and second proteins are combined in pairs to generate a combined dataset. These paired extracellular amino acid fragment sequences are then input into the trained attention mechanism model to obtain interaction scores for these paired extracellular amino acid fragment sequences. Once the interaction scores for all paired extracellular amino acid fragment sequences are obtained, the highest score is used to predict the interaction between the first and second proteins. Specifically, the interaction score output by the attention mechanism model each time is used to characterize the likelihood of interaction between the two extracellular amino acid fragment sequences of the two proteins in that input. For multi-membrane proteins, this process is repeated until the interaction scores for all extracellular amino acid fragment sequences of the first and second proteins are obtained, ultimately outputting a list of interaction scores for the first and second proteins. The highest interaction score in the list of interaction scores is used to characterize the likelihood of interaction between the first and second proteins.

[0065] In another embodiment, when the first protein is identified as a single-transmembrane protein and the second protein is identified as a multi-transmembrane protein based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is 1, and the number of extracellular membrane amino acid fragment sequences of the second protein is multiple. At this time, before the step of inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, it also includes: combining one extracellular membrane amino acid fragment sequence of the first protein with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, the combined data set containing multiple combination elements, each combination element containing one extracellular membrane amino acid fragment sequence of the first protein and one extracellular membrane amino acid fragment sequence of the second protein; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, including: inputting each combination element into the trained attention mechanism model to obtain the interaction score of each combination element output by the trained attention mechanism model; outputting the highest score among the interaction scores of each combination element, and the highest score is used to characterize the possibility of interaction between the first protein and the second protein.

[0066] In this example, the first protein is a single transmembrane protein and the second protein is a multi-transmembrane protein. In this case, the output of the transmembrane prediction neural network contains one extracellular amino acid sequence for the first protein and multiple extracellular amino acid sequences for the second protein. The multiple extracellular amino acid sequences for the first and second proteins are combined in pairs to generate a combined dataset. Each of these paired extracellular amino acid sequences is then input into the trained attention mechanism model to obtain interaction scores for these paired extracellular amino acid sequences. Once the interaction scores for all paired extracellular amino acid sequences are obtained, the highest score is used to predict the interaction between the first and second proteins. Specifically, each interaction score output by the attention mechanism model is used to characterize the likelihood of interaction between the two extracellular amino acid sequences of the two proteins input at that time. For single and multi-transmembrane proteins, this process is repeated until the interaction scores for all extracellular amino acid sequences of the first and second proteins are obtained, ultimately outputting a list of interaction scores for the first and second proteins. The highest interaction score in the list of interaction scores is used to characterize the likelihood of interaction between the first and second proteins.

[0067] In an example of this embodiment, after the above-mentioned step of determining the interaction between the first protein and the second protein based on the highest score in the interaction scores of each combination element, the step further includes: determining multiple target combination elements whose score values in the interaction scores of each combination element are greater than a set threshold; obtaining the position of the extracellular membrane amino acid fragment sequence of the first protein in each combination element in the first protein, and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein; displaying the interaction score of each target combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein in each target combination element.

[0068] Furthermore, in another example, after the step of determining the interaction between the first protein and the second protein based on the highest score among the interaction scores of the combination elements, the method further includes: when the highest score is less than a score threshold, outputting information indicating that the interaction between the first protein and the second protein is weak.

[0069] Specifically, a threshold value for the interaction score is pre-set. The threshold value is used to reflect that the strength of the interaction between the two extracellular membrane amino acid fragment sequences has reached a certain level. When the interaction score of any combination element is greater than the threshold value, the information of the combination element is output. The information of the combination element includes the interaction score of the combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the combination element in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein. For example, as shown in the following table:

[0070]

[0071] Therefore, it is possible to provide researchers with a visualization of the interaction between multiple segments of extracellular membrane amino acid fragment sequences in the first protein and the second protein, providing researchers with analytical assistance for protein interactions. For example, if the interaction between multiple segments of extracellular membrane amino acid fragment sequences is greater than a set threshold, it means that there are multiple segments of amino acid fragments in the extracellular membrane amino acid fragment sequences between the two proteins that interact with each other. After obtaining this display information, the R&D personnel can conduct subsequent biological experiments on the two proteins based on this. For example, based on this, a new drug targeting the protein can be developed. It can also be the case that the first protein and the second protein have many transmembrane region fragments. If the number of interactions between the extracellular membrane amino acid fragment sequences in the first protein and the second protein that are greater than the set threshold is small, such as only one or two segments of extracellular membrane amino acid fragment sequences interacting with each other, the R&D personnel can conduct subsequent biological experiments on the two proteins based on this after obtaining this display information.

[0072] Alternatively, a scoring threshold can be set for the score output by the attention mechanism model. Using this scoring threshold as a reference, if the highest score is less than the scoring threshold, it indicates that the interaction between the first protein and the second protein is weak. In this case, information indicating the weak interaction between the first protein and the second protein needs to be output to directly visually remind researchers of the weak interaction. As analyzed above, the score output by the attention mechanism model is a value between 0 and 1, and a scoring threshold of 0.5 can be selected.

[0073] In the medical field, an accurate understanding of protein interactions is crucial for the analysis of disease mechanisms. For example, the development of certain diseases is related to the imbalance or abnormal interaction between proteins. These studies are not only of great significance in the field of basic science, but also have a profound impact on clinical medicine and new drug research and development. It can be seen that the demand for accurate prediction of protein interactions in the medical field is very urgent. A method for predicting the interaction between transmembrane proteins based on the present application can predict protein interactions by computer, lay a technical foundation for the design and development of drugs based on protein interactions, and improve the efficiency of drug design.

[0074] It should be understood that although the steps in the flowchart are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0075] In one embodiment, a computer device is provided. The computer device may be a server for running a method for predicting the interaction between transmembrane proteins of the present application. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. 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, a computer program and a database. 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 receive the amino acid sequence of the first protein to be predicted and the amino acid sequence of the second protein to be predicted, etc. via the network. When the computer program is executed by the processor, it implements the above-mentioned method for predicting the interaction between transmembrane proteins.

[0076] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is intended to be applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0077] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: obtaining an amino acid sequence of a first protein and an amino acid sequence of a second protein to be predicted; inputting the amino acid sequence of the first protein and the amino acid sequence of the second protein into a transmembrane prediction neural network, respectively; and when it is identified based on the output result of the transmembrane prediction neural network that both the first protein and the second protein are transmembrane proteins, screening out the extracellular membrane amino acid fragment sequence of the first protein from a plurality of amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and screening out the extracellular membrane amino acid fragment sequence of the second protein from a plurality of amino acid fragment sequences of the second protein output by the transmembrane prediction neural network; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the possibility of interaction between the first protein and the second protein, and the trained attention mechanism model is trained using the extracellular membrane amino acid fragment sequences of a plurality of sample transmembrane proteins and the interaction scores between the sample transmembrane proteins.

[0078] In one embodiment, the model training of the attention mechanism model includes: obtaining the extracellular membrane amino acid fragment sequence of a first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of a second sample single transmembrane protein; obtaining the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein, wherein the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein is determined by one or more biological research methods or information mining methods; inputting the extracellular membrane amino acid fragment sequence of the first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of the second sample single transmembrane protein, as well as the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein into the self-attention mechanism model to perform model training on the self-attention mechanism model.

[0079] In one embodiment, when the first protein and the second protein are both identified as single-transmembrane proteins based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is 1, and the number of extracellular membrane amino acid fragment sequences of the second protein is 1; the interaction score between the first protein and the second protein is obtained by inputting 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model.

[0080] In one embodiment, when the output result of the transmembrane prediction neural network is used to identify that the first protein and the second protein are both multi-transmembrane proteins, the number of extracellular membrane amino acid fragment sequences of the first protein is multiple, and the number of extracellular membrane amino acid fragment sequences of the second protein is multiple; when the processor executes the computer program, the following steps are further implemented: each extracellular membrane amino acid fragment sequence of the first protein is combined with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, the combined data set contains multiple combination elements, and each combination element contains one extracellular membrane amino acid fragment sequence of the first protein and one extracellular membrane amino acid fragment sequence of the second protein; when the processor executes the computer program to implement the above-mentioned step of inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, the following steps are specifically implemented: inputting each combination element into the trained attention mechanism model to obtain the interaction score of each combination element output by the trained attention mechanism model; outputting the highest score among the interaction scores of each combination element, and the highest score is used to characterize the possibility of interaction between the first protein and the second protein.

[0081] In one embodiment, when the processor executes the computer program, the following steps are further implemented: determining multiple target combination elements whose score values in the interaction scores of each combination element are greater than a set threshold; obtaining the position of the extracellular membrane amino acid fragment sequence of the first protein in each combination element in the first protein, and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein; displaying the interaction score of each target combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein in each target combination element.

[0082] In one embodiment, when the processor executes the computer program, the processor further implements the following step: when the highest score is less than a score threshold, outputting information indicating that the interaction between the first protein and the second protein is weak.

[0083] In one embodiment, the attention mechanism model includes a word embedding layer, a position encoding layer, a multi-head attention layer and a multi-layer perceptron layer connected in sequence; wherein the multi-head attention layer includes an encoding layer composed of multiple connected encoders, and the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is input into the multi-layer perceptron layer after feature fusion.

[0084] In one embodiment, the method for determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model includes: configuring the attention mechanism model to be trained through preset hyperparameters; obtaining the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein; training the attention mechanism model to be trained through the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein, and adjusting the hyperparameters of the attention mechanism model and the length of the extracellular membrane amino acid fragment sequence of the input sample transmembrane protein during the model training process; when the video memory overflow of the device running the model training is identified, determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model and the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model.

[0085] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: obtaining the amino acid sequence of a first protein and the amino acid sequence of a second protein to be predicted; inputting the amino acid sequence of the first protein and the amino acid sequence of the second protein into a transmembrane prediction neural network respectively; when the first protein and the second protein are both identified as transmembrane proteins based on the output result of the transmembrane prediction neural network, screening out the extracellular membrane amino acid fragment sequence of the first protein from the multiple amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and screening out the extracellular membrane amino acid fragment sequence of the second protein from the multiple amino acid fragment sequences of the second protein output by the transmembrane prediction neural network; inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the possibility of interaction between the first protein and the second protein, and the trained attention mechanism model is trained by the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the sample transmembrane proteins.

[0086] In one embodiment, the model training of the attention mechanism model includes: obtaining the extracellular membrane amino acid fragment sequence of a first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of a second sample single transmembrane protein; obtaining the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein, wherein the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein is determined by one or more biological research methods or information mining methods; inputting the extracellular membrane amino acid fragment sequence of the first sample single transmembrane protein and the extracellular membrane amino acid fragment sequence of the second sample single transmembrane protein, as well as the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein into the self-attention mechanism model to perform model training on the self-attention mechanism model.

[0087] In one embodiment, when the first protein and the second protein are both identified as single-transmembrane proteins based on the output results of the transmembrane prediction neural network, the number of extracellular membrane amino acid fragment sequences of the first protein is 1, and the number of extracellular membrane amino acid fragment sequences of the second protein is 1; the interaction score between the first protein and the second protein is obtained by inputting 1 extracellular membrane amino acid fragment sequence of the first protein and 1 extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model.

[0088] In one embodiment, when the output result of the transmembrane prediction neural network identifies that the first protein and the second protein are both multi-transmembrane proteins, the number of extracellular membrane amino acid fragment sequences of the first protein is multiple, and the number of extracellular membrane amino acid fragment sequences of the second protein is multiple; when the computer program is executed by the processor, the following steps are also implemented: each extracellular membrane amino acid fragment sequence of the first protein is combined with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, the combined data set contains multiple combination elements, each combination element contains one extracellular membrane amino acid fragment sequence of the first protein and one extracellular membrane amino acid fragment sequence of the second protein; the computer The program is executed by the processor to implement the above-mentioned step of inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into the trained attention mechanism model to obtain the interaction score output by the trained attention mechanism model, and determining the interaction between the first protein and the second protein according to the interaction score output by the trained attention mechanism model. The following steps are specifically implemented: inputting each combination element into the trained attention mechanism model to obtain the interaction score of each combination element output by the trained attention mechanism model; and outputting the highest score among the interaction scores of each combination element, where the highest score is used to characterize the possibility of interaction between the first protein and the second protein.

[0089] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented: determining multiple target combination elements whose score values in the interaction scores of each combination element are greater than a set threshold; obtaining the position of the extracellular membrane amino acid fragment sequence of the first protein in each combination element in the first protein, and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein; displaying the interaction score of each target combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein in each target combination element.

[0090] In one embodiment, when the computer program is executed by a processor, the computer program further implements the following step: when the highest score is less than a score threshold, outputting information indicating that the interaction between the first protein and the second protein is weak.

[0091] In one embodiment, the attention mechanism model includes a word embedding layer, a position encoding layer, a multi-head attention layer and a multi-layer perceptron layer connected in sequence; wherein the multi-head attention layer includes an encoding layer composed of multiple connected encoders, and the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is input into the multi-layer perceptron layer after feature fusion.

[0092] In one embodiment, the method for determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model includes: configuring the attention mechanism model to be trained through preset hyperparameters; obtaining the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein; training the attention mechanism model to be trained through the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between each sample transmembrane protein, and adjusting the hyperparameters of the attention mechanism model and the length of the extracellular membrane amino acid fragment sequence of the input sample transmembrane protein during the model training process; when the video memory overflow of the device running the model training is identified, determining the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer and multi-layer perceptron layer in the attention mechanism model and the length threshold of the extracellular membrane amino acid fragment sequence input to the attention mechanism model.

[0093] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this 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. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), 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).

[0094] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for predicting the interaction between transmembrane proteins, characterized in that: include: Obtaining an amino acid sequence of a first protein to be predicted and an amino acid sequence of a second protein; inputting the amino acid sequence of the first protein and the amino acid sequence of the second protein into a transmembrane prediction neural network respectively, and when it is identified based on the output result of the transmembrane prediction neural network that the first protein and the second protein are both transmembrane proteins, screening the cell membrane extracellular amino acid fragment sequence of the first protein from the multiple amino acid fragment sequences of the first protein output by the transmembrane prediction neural network, and screening the cell membrane extracellular amino acid fragment sequence of the second protein from the multiple amino acid fragment sequences of the second protein output by the transmembrane prediction neural network; The extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein are input into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model, wherein the interaction score is used to characterize the possibility of interaction between the first protein and the second protein, and the trained attention mechanism model is trained by the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the sample transmembrane proteins.

2. The method according to claim 1, characterized in that The model training of the attention mechanism model includes: Obtaining the extracellular membrane amino acid fragment sequence of the first sample single-transmembrane protein and the extracellular membrane amino acid fragment sequence of the second sample single-transmembrane protein; Obtaining an interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein, wherein the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein is determined by one or more biological research methods or information mining methods; The extracellular membrane amino acid fragment sequence of the first sample single transmembrane protein, the extracellular membrane amino acid fragment sequence of the second sample single transmembrane protein, and the interaction score between the first sample single transmembrane protein and the second sample single transmembrane protein are input into the self-attention mechanism model to perform model training on the self-attention mechanism model.

3. The method according to claim 1, characterized in that When it is identified based on the output result of the transmembrane prediction neural network that the first protein and the second protein are both single-transmembrane proteins, the number of extracellular membrane amino acid fragment sequences of the first protein is one, and the number of extracellular membrane amino acid fragment sequences of the second protein is one; The interaction score between the first protein and the second protein is obtained by inputting an extracellular membrane amino acid fragment sequence of the first protein and an extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model.

4. The method according to claim 1, wherein When it is identified based on the output result of the transmembrane prediction neural network that both the first protein and the second protein are multi-membrane proteins, the first protein has a plurality of extracellular membrane amino acid fragment sequences, and the second protein has a plurality of extracellular membrane amino acid fragment sequences; the method further comprises: Combining each extracellular membrane amino acid fragment sequence of the first protein with each extracellular membrane amino acid fragment sequence of the second protein in pairs to obtain a combined data set, wherein the combined data set comprises a plurality of combined elements, each combined element comprising one extracellular membrane amino acid fragment sequence of the first protein and one extracellular membrane amino acid fragment sequence of the second protein; Inputting the extracellular membrane amino acid fragment sequence of the first protein and the extracellular membrane amino acid fragment sequence of the second protein into a trained attention mechanism model to obtain an interaction score output by the trained attention mechanism model includes: Inputting each combination element into a trained attention mechanism model to obtain an interaction score of each combination element output by the trained attention mechanism model; The highest score among the interaction scores of the combination elements is output, and the highest score is used to characterize the possibility of the interaction between the first protein and the second protein.

5. The method according to claim 4, characterized in that The method further comprises: determining a plurality of target combination elements having a score greater than a set threshold in the interaction score of each combination element; Obtaining the position of the extracellular membrane amino acid fragment sequence of the first protein in each combination element in the first protein, and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein; The interaction score of each target combination element and the position of the extracellular membrane amino acid fragment sequence of the first protein in the first protein and the position of the extracellular membrane amino acid fragment sequence of the second protein in the second protein in each target combination element are displayed.

6. The method according to claim 5, characterized in that The method further comprises: When the highest score is less than the score threshold, information indicating that the interaction between the first protein and the second protein is weak is output.

7. The method according to claim 1, characterized in that The attention mechanism model includes a word embedding layer, a position encoding layer, a multi-head attention layer and a multi-layer perceptron layer connected in sequence; Among them, the multi-head attention layer includes an encoding layer composed of multiple connected encoders, and the self-attention representation of the extracellular membrane amino acid fragment sequence generated by the encoding layer is input into the multi-layer perceptron layer after feature fusion.

8. The method according to claim 7, characterized in that The hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer, and multi-layer perceptron layer in the attention mechanism model are determined in the following manner: Configure the attention mechanism model to be trained through preset hyperparameters; Obtain the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the transmembrane proteins of each sample; The attention mechanism model to be trained is trained using the extracellular membrane amino acid fragment sequences of multiple sample transmembrane proteins and the interaction scores between the sample transmembrane proteins. During the model training process, the hyperparameters of the attention mechanism model and the length of the extracellular membrane amino acid fragment sequence of the input sample transmembrane protein are adjusted; When a video memory overflow of a device running model training is identified, the hyperparameters of the word embedding layer, position encoding layer, multi-head attention layer, and multi-layer perceptron layer in the attention mechanism model and the length threshold of the extracellular membrane amino acid fragment sequence input into the attention mechanism model are determined.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Citation Information

Cited By

  • A membrane protein function prediction method fusing spatial structure weighting and hierarchical constraint

    CN122392649A