Biological reagent function prediction method and system

Abstract

The invention relates to the technical field of biological function prediction, in particular to a biological reagent function prediction method and system. The method comprises the following steps: carrying out biomolecule information acquisition on a biological reagent public database to obtain a biological reagent biomolecule information set; performing biological characteristic analysis on the biological reagent biomolecule information set to obtain biological reagent protein sequence comparison characteristics, biological reagent molecular structure domain characteristics and biological reagent specific molecular function domain characteristics; constructing a biological reagent function prediction network framework and performing function prediction model training construction to generate a biological reagent function prediction model; obtaining molecular characteristics of the biological reagent to be predicted and performing target function prediction to obtain a target biological reagent function prediction result; and performing function verification optimization according to the target biological reagent function prediction result to generate a target biological reagent function optimization result. According to the invention, effective pre-judgment and screening of biological reagent functions can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of biological function prediction, and in particular to a biological reagent function prediction method and system. Background Art

[0002] In recent years, with the rapid development of artificial intelligence and machine learning technologies, data-driven biological reagent function prediction methods have gradually become a research hotspot. These methods can extract complex rules and perform function predictions on a larger scale by analyzing large-scale biological data and using algorithms such as deep learning, support vector machines, and random forests. Machine learning methods can better integrate biological information from different sources, such as genomic data, proteomic data, and literature data, to improve the accuracy and efficiency of predictions. However, traditional function prediction methods have some limitations. For example, the functions of many biological reagents are diverse, and they exhibit different biological activities under different environments or conditions. It is difficult to fully consider the characteristics of molecular domains and molecular functional domains, resulting in low accuracy in the function prediction of biological reagents. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a method and system for predicting the function of a biological agent to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for predicting the function of a biological agent comprises the following steps:

[0005] Step S1: collecting biomolecule information from a public database of biological reagents to obtain a biomolecule information set of biological reagents, wherein the biomolecule information set of biological reagents includes biomolecule protein sequence data, biomolecule structural domain data, and biomolecule functional domain data of biological reagents;

[0006] Step S2: performing sequence alignment analysis on the biological reagent protein sequence data to obtain the biological reagent protein sequence alignment features; performing protein structure feature prediction analysis on the biological reagent molecular domain data to generate the biological reagent molecular domain features; performing specific molecular function feature extraction based on the biological reagent molecular function domain data to obtain the biological reagent specific molecular function domain features;

[0007] Step S3: constructing a biological reagent function prediction network framework, and inputting the biological reagent protein sequence alignment features, biological reagent molecular domain features, and biological reagent specific molecular functional domain features into the biological reagent function prediction network framework for function prediction model training and construction, so as to generate a biological reagent function prediction model; obtaining the molecular features of the biological reagent to be predicted, and inputting the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction, so as to obtain the target biological reagent function prediction result;

[0008] Step S4: According to the function prediction result of the target biological reagent, the corresponding biological reagent to be predicted is functionally verified and optimized to generate the function optimization result of the target biological reagent.

[0009] Further, step S1 includes the following steps:

[0010] Step S11: Preliminary screening of the original data of biological reagents in the biological reagent public database to obtain a set of original biological information of biological reagents;

[0011] Step S12: performing protein sequence extraction processing on the original biological information set of the biological reagent to obtain biological reagent protein sequence data, including the amino acid molecular sequence and gene molecular sequence corresponding to each biological reagent;

[0012] Step S13: performing molecular domain recognition, calibration and extraction processing on the amino acid molecular sequence and gene molecular sequence corresponding to the biological reagent protein sequence data to obtain biological reagent molecular domain data, including the secondary domain sequence fragment and the tertiary domain sequence fragment corresponding to each biological reagent;

[0013] Step S14: performing molecular functional domain annotation and parsing processing on the molecular structural domain data of the biological reagent to obtain the molecular functional domain data of the biological reagent, including the enzyme catalytic activity, binding affinity and substance transport and degradation functional domain fragments corresponding to each biological reagent;

[0014] Step S15: Integrate the biological reagent protein sequence data, the biological reagent molecular structure domain data and the biological reagent molecular function domain data to obtain a biological reagent biomolecule information set.

[0015] Furthermore, the biological reagent original biological information set in step S11 includes biological reagent original information data corresponding to enzymes, antibodies and microorganism types.

[0016] Further, step S13 includes the following steps:

[0017] Step S131: Pre-identify the molecular domain of the corresponding amino acid molecular sequence and gene molecular sequence in the biological reagent protein sequence data to obtain the biological reagent amino acid molecular domain region sequence and the biological reagent gene molecular domain region sequence;

[0018] Step S132: Obtaining corresponding amino acid molecular structure sites and amino acid molecular interaction forces through the amino acid molecular structure domain region sequence of the biological reagent, and performing sequence structure hydrophobicity evaluation and analysis on the corresponding amino acid molecular structure domain region sequence of the biological reagent based on the amino acid molecular structure sites and amino acid molecular interaction forces to obtain the hydrophobicity score of the amino acid molecular structure of the biological reagent;

[0019] Step S133: performing molecular structure functional charge analysis on the amino acid molecular structure domain region sequence of the biological reagent to obtain the amino acid molecular structure functional charge characteristics of the biological reagent, including the amino acid molecular structure charge distribution, the amino acid molecular structure functional charge coupling degree and the amino acid molecular structure charge density;

[0020] Step S134: Based on the hydrophobicity score of the amino acid molecular structure of the biological reagent and the functional charge characteristics of the amino acid molecular structure of the biological reagent, a structural domain metric calculation is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent using a protein secondary structure metric calculation formula to obtain the secondary structure degree of the amino acid molecular region sequence; based on the secondary structure degree of the amino acid molecular region sequence, a secondary structure calibration process is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent to obtain a secondary structure domain sequence fragment corresponding to the biological reagent;

[0021] Step S135: Perform molecular tertiary structure deduction analysis on the molecular domain region sequence of the biological reagent gene to obtain the tertiary domain sequence fragment corresponding to the biological reagent.

[0022] Furthermore, the protein secondary structure metric calculation formula in step S134 is specifically:

[0023]

[0024] In the formula, S s is the secondary structure degree of the amino acid molecule region sequence, L is the total length of the amino acid molecule domain sequence, r is the domain spatial position parameter, H(r) is the hydrophobicity score of the amino acid molecule structure of the biological reagent at the spatial position r of the amino acid molecule region sequence, α h is the hydrophobicity weight factor associated with the amino acid, C(r) is the charge distribution of the amino acid molecular structure function of the biological reagent at the spatial position r of the amino acid molecular region sequence, O(r) is the charge coupling degree of the amino acid molecular structure function of the amino acid molecular region sequence at the spatial position r, M(r) is the charge density of the amino acid molecular structure at the spatial position r of the amino acid molecular region sequence, β c is the charge characteristic weight factor related to the amino acid, F(r) is the characteristic parameter of the amino acid molecular structure functional group in the amino acid molecular region sequence at the spatial position r, γ f is the weight factor of the functional group characteristics related to amino acids, and η is the correction coefficient of the secondary structure degree of the amino acid molecule region sequence.

[0025] Further, step S135 includes the following steps:

[0026] Conducting gene molecular dynamics simulation analysis on the sequence of the molecular structure domain region of the biological reagent gene to generate the biological reagent gene molecular dynamics simulation process;

[0027] Perform gene molecule folding path deduction on the biological reagent gene molecular dynamics simulation process to generate a biological reagent gene molecular folding deduction path diagram;

[0028] Based on the biological reagent gene molecular folding deduction path diagram, the domain folding energy distribution analysis is performed on the corresponding biological reagent gene molecular domain region sequence to obtain the biological reagent gene molecular domain folding energy distribution diagram;

[0029] Obtain the corresponding gene molecule folding energy stability score and gene molecule folding energy distribution gradient through the folding energy distribution diagram of the biological reagent gene molecule domain, and perform matching calculation on the corresponding biological reagent gene molecule domain region sequence using the protein tertiary structure matching calculation formula based on the gene molecule folding energy stability score and the gene molecule folding energy distribution gradient, so as to obtain the tertiary structure matching degree of the gene molecule region sequence;

[0030] Among them, the protein tertiary structure matching calculation formula is as follows:

[0031]

[0032] In the formula, M is the tertiary structure matching degree of the gene molecule region sequence, Ω is the range of the gene molecule folding space, and x is the coordinate of the gene molecule folding space. is the folding energy Hamiltonian at the gene molecule folding space coordinate x, δ is the gene molecule folding energy adjustment coefficient, is the gradient symbol, U fold (x) is the gene molecule folding energy stability score at the gene molecule folding space coordinate x, ε is the gene molecule folding energy stability influencing parameter, is the gene molecule folding energy distribution gradient, θ is the folding energy distribution gradient influence coefficient, ξ is the correction coefficient of the tertiary structure matching degree of the gene molecule region sequence;

[0033] Based on the tertiary structure matching degree of the gene molecule region sequence, the corresponding biological reagent gene molecule domain region sequence is subjected to tertiary structure calibration processing to obtain the tertiary structure domain sequence fragment corresponding to the biological reagent.

[0034] Further, step S15 includes the following steps:

[0035] Step S151: performing functional domain division on the corresponding secondary structure domain sequence fragments in the biological reagent molecule structure domain data to obtain biological reagent molecule secondary structure functional domain type segments, including biological reagent molecule enzyme functional domain fragments and biological reagent molecule binding functional domain fragments;

[0036] Step S152: obtaining known catalytic enzyme reaction mechanisms and enzyme substrate characteristics, and performing catalytic potential prediction analysis on the enzyme functional domain fragments of the biological reagent molecule based on the known catalytic enzyme reaction mechanisms and enzyme substrate characteristics to generate a biological reagent molecule catalytic potential score; performing enzyme catalytic activity annotation processing on the corresponding biological reagent molecule enzyme functional domain fragments based on the biological reagent molecule catalytic potential score to obtain the enzyme catalytic activity functional domain fragments corresponding to each biological reagent;

[0037] Step S153: simulating different environmental factor constraints and dynamically annotating the binding affinity of the biological reagent molecule binding domain fragments based on the different environmental factor constraints to obtain the binding affinity domain fragments corresponding to each biological reagent;

[0038] Step S154: performing a transmembrane transport capacity prediction analysis on the tertiary domain sequence fragments corresponding to the biological reagent molecular structure domain data to obtain a predicted value of the transmembrane transport capacity of the biological reagent molecular structure;

[0039] Step S155: Based on the predicted value of the transmembrane transport capacity of the biological reagent molecular structure, the corresponding tertiary structural domain sequence fragments are annotated for substance transport degradation to obtain the substance transport degradation functional domain fragments corresponding to each biological reagent.

[0040] Further, step S2 includes the following steps:

[0041] Step S21: performing sequence alignment analysis on each protein sequence in the biological reagent protein sequence data to obtain a biological reagent protein sequence alignment matrix, which includes alignment scores, matching lengths, and gap regions between each protein sequence;

[0042] Step S22: performing sequence difference statistical analysis on the corresponding protein sequences in the biological reagent protein sequence data based on the biological reagent protein sequence comparison matrix to obtain the biological reagent protein sequence comparison features, including the corresponding amino acid residue substitution, insertion and deletion difference features between the protein sequences;

[0043] Step S23: performing protein structure feature prediction analysis on the biological agent molecular domain data to generate biological agent molecular domain features, including molecular domain stability, molecular domain hydrophobic interaction force, and molecular domain binding energy features;

[0044] Step S24: extract specific molecular function features based on the biological reagent molecular function domain data to obtain the biological reagent specific molecular function domain features, including the biological reagent enzyme activity level, biological reagent binding affinity and biological reagent material transport and degradation ability characteristics.

[0045] Further, step S3 includes the following steps:

[0046] Step S31: constructing a biological reagent function prediction network framework by connecting a convolutional neural network CNN, a recurrent neural network RNN, and a Transformer neural network;

[0047] Step S32: inputting the biological reagent protein sequence alignment features, the biological reagent molecular domain features and the biological reagent specific molecular functional domain features into the biological reagent function prediction network framework to train and construct a function prediction model, while considering the activity, structural stability and specific function prediction targets corresponding to the biological reagent to generate a biological reagent function prediction model;

[0048] Step S33: obtaining molecular features of the biological agent to be predicted, including protein sequence alignment features, molecular structural domain features, and molecular functional domain features corresponding to the biological agent to be predicted;

[0049] Step S34: inputting the molecular features of the biological agent to be predicted into the biological agent function prediction model to perform target function prediction to obtain a target biological agent function prediction result.

[0050] Furthermore, the present invention also provides a biological reagent function prediction system for executing the biological reagent function prediction method as described above, the biological reagent function prediction system comprising:

[0051] A biological reagent molecule information collection module is used to collect biological reagent molecule information from a biological reagent public database to obtain a biological reagent biological reagent molecule information set, wherein the biological reagent biological reagent molecule information set includes biological reagent protein sequence data, biological reagent molecule structure domain data, and biological reagent molecule function domain data;

[0052] The biological reagent molecular feature analysis module is used to perform sequence comparison analysis on the biological reagent protein sequence data to obtain the biological reagent protein sequence comparison features; perform protein structure feature prediction analysis on the biological reagent molecular domain data to generate the biological reagent molecular domain features; extract specific molecular function features based on the biological reagent molecular function domain data to obtain the specific molecular function domain features of the biological reagent;

[0053] The biological reagent target function prediction module is used to construct a biological reagent function prediction network framework, and input the biological reagent protein sequence alignment features, biological reagent molecular domain features and biological reagent specific molecular function domain features into the biological reagent function prediction network framework for function prediction model training and construction to generate a biological reagent function prediction model; obtain the molecular features of the biological reagent to be predicted, and input the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction to obtain the target biological reagent function prediction result;

[0054] The biological reagent function verification and optimization module is used to perform function verification and optimization on the corresponding biological reagent to be predicted according to the function prediction result of the target biological reagent, so as to generate the function optimization result of the target biological reagent.

[0055] Beneficial effects of the present invention:

[0056] 1. Compared with the prior art, the method for predicting the function of biological reagents proposed in the present invention has the beneficial effect that, by collecting information from the public database of biological reagents, a biological molecule information set including protein sequence data, molecular domain data and molecular functional domain data is obtained. The key to this step is that the public database brings together a large amount of verified and studied data, so that the collected biological molecule information has high quality and reliability. Protein sequence data is the basis for studying the structure and function of biological molecules. It can provide basic genetic information of biological reagents and help researchers understand their functions in different biological systems; molecular domain data helps to identify the functional regions of proteins, such as enzyme active centers, receptor binding sites, etc., and then reveal the connection between molecular function and structure; and the collection of functional domain data can help researchers identify the role of molecules in specific biological processes, thereby providing direction for further functional research and application development. This information set not only contains the basic molecular sequence data of each biological reagent, but also systematically summarizes the structural characteristics and functional characteristics of the molecule, providing rich materials for further biological research, avoiding the limitations of a single data source, and thus improving the efficiency of data utilization. Secondly, by performing sequence comparison analysis on the protein sequence data of biological reagents to obtain the alignment features of the protein sequences of biological reagents, we can deeply explore the variation features of protein sequences, help reveal the functional differences of biological reagents at the molecular level, identify the regularity of these differences and their functional significance in specific organisms, and make the research on the function of biological reagent proteins and their variability characteristics more in-depth and comprehensive. At the same time, by performing protein structural feature prediction analysis on the molecular domain data of biological reagents, we can deeply understand the structural regions and structural characteristics of proteins. The structural domains of proteins are the basis of their functions, and different structural domains have different functional characteristics. By predicting the stability, hydrophobic interaction force and binding energy of these structural domains, we can provide deep insights into the functional mechanism of proteins, help scientists more accurately understand the structural domain characteristics of biological reagents corresponding to protein molecules, and then promote the functional prediction process of biological reagents. By extracting specific molecular functional features based on the molecular functional domain data of biological reagents, its functional role in biological processes can be comprehensively evaluated. This process not only involves the extraction of features such as protein enzyme activity, binding affinity, and material transport and degradation ability, but also provides important data support for the functional prediction of biological reagents. This step can optimize the application performance of biological reagents through in-depth analysis of these functional domain characteristics, and provide strong data support and theoretical basis for the application and development of biological reagents, thereby fully considering the functional domain characteristics of protein molecules corresponding to biological reagents.Then, through the combination of convolutional neural network (CNN), recurrent neural network (RNN) and Transformer neural network, a biological reagent function prediction network framework was constructed, and the biological reagent protein sequence alignment features, biological reagent molecular domain features and biological reagent specific molecular functional domain features were input into the biological reagent function prediction network framework for function prediction model training and construction, and a multi-dimensional biological reagent function prediction model was established. Among them, the protein sequence alignment features help to identify amino acid sequence patterns related to known protein functions and promote functional speculation of similar protein families; the molecular domain features help to capture the tertiary structure and other important functional domains of proteins, and provide structural information closely related to molecular functions; and the specific molecular functional domain features can reveal regions related to certain specific biological functions (such as enzyme activity, receptor binding, signal transduction, etc.). The input of these features can greatly improve the performance of the model in complex biological data. By combining the activity, structural stability and specific function prediction goals of biological reagents for training, the model can not only stay at the learning of surface features, but also help the model to comprehensively analyze the functions of molecules from different levels, thereby improving the diversity and accuracy of biological reagent prediction results. The target function is predicted by inputting the molecular features of the biological reagent to be predicted into the biological reagent function prediction model, thereby obtaining the function prediction results of the target biological reagent. The key to this step is to provide biologically meaningful prediction results by combining the real biological molecular features with the trained prediction model. Through the efficient computing and pattern recognition capabilities of the neural network model, key information can be extracted from a large number of molecular features in a very short time, and the functional performance of biological reagents in specific biological or pharmacological environments can be predicted, such as their potential activities in enzyme catalysis, receptor binding, antigen recognition, etc., to help quickly screen potential biological reagents, thereby greatly improving the efficiency and success rate of function prediction. Finally, the function of the target biological reagent is verified and optimized through the function prediction results of the target biological reagent. The key to this stage is that the function prediction results based on the prediction model can systematically guide the experimental design, reduce the uncertainty and repeatability in the experiment, and improve the verification efficiency. If the experimental results are consistent with the prediction, the reliability of the function prediction model can be confirmed; if the experimental results are inconsistent with the prediction, the model parameters can be further adjusted or the data set can be expanded to continuously optimize the model. This process ultimately promotes the qualitative and quantitative analysis of the function of biological reagents, and also promotes the research and development and practical application of new biological reagents, thereby improving the accuracy of function prediction.

[0057] 2. The biological reagent function prediction system proposed in the present invention is generally composed of a biological reagent molecular information acquisition module, a biological reagent molecular feature analysis module, a biological reagent target function prediction module and a biological reagent function verification and optimization module. It can implement any biological reagent function prediction method described in the present invention, and is used to combine the operations between computer programs running on each module to implement the biological reagent function prediction method. The internal structures of the system cooperate with each other, which can greatly reduce duplication of work and manpower investment, and can quickly and effectively provide a more accurate and efficient biological reagent function prediction process, thereby simplifying the operation process of the biological reagent function prediction system. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0059] Figure 1 A schematic diagram of the steps of the method for predicting the function of a biological reagent of the present invention;

[0060] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0061] Figure 3 for Figure 2 Detailed step flow chart of step S15 in FIG. DETAILED DESCRIPTION

[0062] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0063] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for predicting the function of a biological agent. In the embodiments of the present invention, please refer to Figure 1 FIG. 1 is a schematic diagram of the steps of the method for predicting the function of a biological agent of the present invention. In this example, the method for predicting the function of a biological agent includes the following steps:

[0064] Step S1: collecting biomolecule information from a public database of biological reagents to obtain a biomolecule information set of biological reagents, wherein the biomolecule information set of biological reagents includes biomolecule protein sequence data, biomolecule structural domain data, and biomolecule functional domain data of biological reagents;

[0065] In an embodiment of the present invention, raw data related to biological reagents are obtained from public biological reagent databases (such as UniProt, PDB, GeneBank, etc.), which contain a wide range of protein and gene sequence information. The screening process includes filtering relevant records by keywords (such as biological reagent name, category, function, etc.) or according to specific screening conditions (such as species, experimental data type, protein expression system, etc.), and using a programming language (such as Python) in combination with a bioinformatics toolkit (such as Biopython) to capture and screen data. After screening, a raw data set containing all qualified biological reagent-related proteins and their corresponding gene sequences is obtained, including enzymes, antibodies, and biological reagent raw information data corresponding to microbial types, and protein sequences and gene sequences are extracted therefrom. First, the sequence files in FASTA format or GenBank format stored in the database are parsed using tools such as Biopython to extract the protein amino acid sequence and the corresponding gene sequence corresponding to each biological reagent, thereby obtaining the biological reagent protein sequence data. By using professional domain prediction software to identify the molecular domains of the extracted amino acid sequences and gene sequences, specifically, tools such as InterProScan, Pfam, SMART, etc. can be used to calibrate and identify the domains of protein amino acid sequences. These tools can automatically identify the domains in protein sequences based on known domain databases and protein sequence alignment algorithms. By calibrating the domains, the amino acid sequence of each biological reagent can be divided into several domain fragments with specific functions, such as secondary domain and tertiary domain sequence fragments, thereby obtaining the molecular domain data of the biological reagent. At the same time, based on the previously extracted molecular domain data of the biological reagent, the molecular functional domain of each biological reagent is further annotated and analyzed. In specific implementation, databases such as KEGG, Reactome, Gene Ontology (GO), etc. can be used to perform functional annotations on each domain. By comparing and annotating the domains, the functional information of each biological reagent is identified, including enzyme catalytic activity and molecular binding affinity (obtained from the secondary domain), material transport and degradation (obtained from the tertiary domain), and other functional characteristics, thereby obtaining the molecular functional domain data of the biological reagent. Then, by completing the comprehensive data integration of biological reagents, the specific operation includes integrating the protein sequence data, molecular domain data and molecular functional domain data obtained from the previous analysis. The integrated biological reagent biomolecule information set contains detailed information of each biological reagent, including amino acid and gene sequences, domain division, functional domain annotation, etc., and finally the biological reagent biomolecule information set is obtained.

[0066] Step S2: performing sequence alignment analysis on the biological reagent protein sequence data to obtain the biological reagent protein sequence alignment features; performing protein structure feature prediction analysis on the biological reagent molecular domain data to generate the biological reagent molecular domain features; performing specific molecular function feature extraction based on the biological reagent molecular function domain data to obtain the biological reagent specific molecular function domain features;

[0067] In the embodiment of the present invention, a multiple sequence alignment (MSA) is performed on each protein sequence in the biological reagent protein sequence data. Common alignment tools such as BLAST (Basic Local Alignment Search Tool) or Clustal Omega can be used to align the protein sequences. The input protein sequence data is first preprocessed to remove low-quality sequences and standardize the formats of all sequences (such as FASTA format). During the alignment process, the scores of amino acid substitutions are calculated by setting a suitable alignment score matrix (such as BLOSUM or PAM), and suitable gap penalties and extension penalties are set. Penalty) is used to process insertion and deletion regions. The alignment tool will output an alignment matrix according to the set parameters, which includes the alignment score, matching length, and insertion and deletion regions (gap) between each pair of protein sequences, and realizes sequence difference statistical analysis of the corresponding protein sequences. The goal of this step is to identify and count the differences between different protein sequences, mainly including amino acid residue substitutions, insertions and deletions. Using the alignment results, the residue differences of each paired sequence are first extracted, and the frequency of amino acid substitutions is calculated by calculating the changes in amino acids at each alignment position. At the same time, the insertion and deletion positions in the sequence are analyzed to identify which regions have insertions or deletions, thereby obtaining the alignment characteristics of the biological reagent protein sequence. At the same time, by predicting and analyzing the protein structural features of the previously extracted biological reagent molecular domain data, the domain of the protein sequence is analyzed. First, the three-dimensional structure of each protein sequence is predicted using protein structure prediction software (such as Phyre2, I-TASSER). This process will predict the spatial structure of the protein sequence based on the sequence information, and obtain the conformation and interaction information of the domain. On this basis, the stability of the domain can be further calculated, for example, by evaluating the free energy of the protein or by simulating the protein folding process to infer the stability. The hydrophobic interaction force refers to the non-polar interaction between the hydrophobic amino acid residues in the protein molecule. The use of molecular dynamics simulation (such as GROMACS software) can calculate the energy change of this interaction, thereby predicting the molecular domain characteristics of the biological reagent, including the molecular domain stability, the molecular domain hydrophobic interaction force and the molecular domain binding energy characteristics.Then, based on the molecular functional domain data of biological reagents, the specific functional characteristics of proteins are extracted. First, the protein sequence is mapped to the known functional domain database through bioinformatics tools (such as InterProScan, Pfam) to determine the functional domains it contains and their corresponding biological functions. For each determined functional domain, the enzyme activity level is predicted using existing experimental data or computational models (such as Enzyme Commission number, EC number), and the binding affinity of the protein is calculated. In addition, combined with the transport and degradation ability characteristic data of biological reagents, biological process networks (such as KEGG database) or molecular simulation tools can be further used to predict the activity of proteins in the process of material transport or degradation. These specific molecular functional characteristics include enzyme activity, binding affinity and transport and degradation ability, and finally the specific molecular functional domain characteristics of biological reagents are obtained.

[0068] Step S3: constructing a biological reagent function prediction network framework, and inputting the biological reagent protein sequence alignment features, biological reagent molecular domain features, and biological reagent specific molecular functional domain features into the biological reagent function prediction network framework for function prediction model training and construction, so as to generate a biological reagent function prediction model; obtaining the molecular features of the biological reagent to be predicted, and inputting the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction, so as to obtain the target biological reagent function prediction result;

[0069] In an embodiment of the present invention, a neural network framework for predicting the function of a biological agent is constructed by using a combination of a convolutional neural network (CNN), a recurrent neural network (RNN) and a Transformer neural network. In a specific implementation, CNN is used for local feature extraction, RNN processes the time dependency of feature sequence data, and Transformer captures the global dependency through its self-attention mechanism, and these features are fused and output through a fully connected layer, thereby constructing a biological agent function prediction network framework. By inputting the protein sequence alignment features, molecular domain features and molecular functional domain features of the biological agent into the constructed function prediction network framework, these features will be sent to an integrated learning model (such as a multi-layer perceptron, MLP) after being processed by the aforementioned CNN, RNN and Transformer networks, and the function prediction model is trained. During the training process, the network not only considers the basic molecular features of the biological agent, but also includes target information such as the activity, structural stability and specific function of the biological agent. These targets serve as supervisory signals to help the network perform multi-task learning, thereby improving the prediction accuracy of the model, thereby training and generating a biological agent function prediction model. By obtaining corresponding molecular features for the biological reagent to be predicted, these features include protein sequence alignment features, molecular domain features and molecular functional domain features of the biological reagent to be predicted, and the molecular features of the biological reagent to be predicted are obtained. Then, by inputting the molecular features of the biological reagent to be predicted into the previously trained biological reagent function prediction model, the specific operation includes inputting the protein sequence alignment features, molecular domain features and molecular functional domain features of the biological reagent to be predicted into the model in a predetermined format. The model uses the aforementioned deep neural network (a combination of CNN, RNN and Transformer) and a multi-task learning framework based on these input data to comprehensively predict the function of the biological reagent. The predicted targets usually include molecular activity, structural stability and specific functions, etc. After being processed by the model, this information will be output in the form of probability, and each function will be scored. The results output by the model can provide researchers with the function prediction results of the target biological reagent, and finally the function prediction results of the target biological reagent are obtained.

[0070] Step S4: According to the function prediction result of the target biological reagent, the corresponding biological reagent to be predicted is functionally verified and optimized to generate the function optimization result of the target biological reagent.

[0071] In an embodiment of the present invention, the function of the target biological reagent is verified and optimized by using the results of the biological reagent function prediction model. First, the molecular features of the biological reagent to be predicted are input to obtain the function prediction results output by the model. This prediction result can be a classification label (such as "enzyme activity" or "receptor binding") or a function score (such as the possibility score of a specific function). The prediction result provides a theoretical basis for subsequent experimental verification. Next, according to the prediction result, the biological reagent to be verified is experimentally verified. For example, for a biological reagent predicted to be "enzyme activity", it can be verified by enzyme activity experiments (such as substrate conversion rate determination, enzyme inhibition test, etc.) to verify whether it has the expected enzyme activity. For the "receptor binding" prediction result, it can be further verified by receptor binding experiments (such as ELISA, flow cytometry, etc.). After the experimental verification, if it is found that the function prediction is inconsistent with the experimental results, the function prediction model can be optimized and adjusted according to the actual situation, for example, re-extraction of features, model architecture adjustment or training data expansion can be performed to improve the prediction accuracy of the model. In addition, the experimental results can also be fed back to the database to optimize the relevant data sets of the biological reagents, further improve the accuracy and reliability of the model, and finally generate the function optimization results of the target biological reagents.

[0072] Further, as an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps:

[0073] Step S11: Preliminary screening of the original data of biological reagents in the biological reagent public database to obtain a set of original biological information of biological reagents;

[0074] In an embodiment of the present invention, raw data related to biological reagents are obtained from public biological reagent databases (such as UniProt, PDB, GeneBank, etc.), which contain a wide range of protein and gene sequence information. The screening process includes filtering relevant records by keywords (such as biological reagent name, category, function, etc.) or according to specific screening conditions (such as species, experimental data type, protein expression system, etc.), and using a programming language (such as Python) in combination with a bioinformatics toolkit (such as Biopython) to capture and screen data. After screening, a raw data set containing all qualified biological reagent-related proteins and their corresponding gene sequences is obtained, including enzymes, antibodies, and biological reagent raw information data corresponding to microbial types, and finally a biological reagent raw biological information set is obtained.

[0075] Step S12: performing protein sequence extraction processing on the original biological information set of the biological reagent to obtain biological reagent protein sequence data, including the amino acid molecular sequence and gene molecular sequence corresponding to each biological reagent;

[0076] In an embodiment of the present invention, by extracting protein sequences and gene sequences from the previously obtained original biological information set, firstly, using tools such as Biopython to parse the sequence files in the FASTA format or GenBank format stored in the database, the protein amino acid sequence and the corresponding gene sequence corresponding to each biological reagent are extracted. In this process, it is necessary to ensure that each protein sequence correctly matches its corresponding gene sequence to avoid data confusion. At the same time, a protein sequence database (such as UniProt) can be used for confirmation and correction to ensure that the extracted sequence is a complete protein coding sequence. The obtained protein sequence data includes the amino acid molecular sequence and gene molecular sequence corresponding to each biological reagent, and finally the biological reagent protein sequence data is obtained.

[0077] Step S13: performing molecular domain recognition, calibration and extraction processing on the amino acid molecular sequence and gene molecular sequence corresponding to the biological reagent protein sequence data to obtain biological reagent molecular domain data, including the secondary domain sequence fragment and the tertiary domain sequence fragment corresponding to each biological reagent;

[0078] In an embodiment of the present invention, molecular domain identification is performed on the extracted amino acid sequence and gene sequence by using professional domain prediction software. Specifically, tools such as InterProScan, Pfam, SMART, etc. can be used to calibrate and identify the domains of protein amino acid sequences. These tools can automatically identify the domains in protein sequences based on known domain databases and protein sequence alignment algorithms. By calibrating the domains, the amino acid sequence of each biological agent can be divided into several domain fragments with specific functions. At the same time, three-dimensional structural information related to the domain can also be identified, such as secondary domain and tertiary domain sequence fragments. The molecular domain data generated by this process will provide a basis for subsequent molecular function annotation, and ultimately the molecular domain data of the biological agent is obtained.

[0079] Step S14: performing molecular functional domain annotation and parsing processing on the molecular structural domain data of the biological reagent to obtain the molecular functional domain data of the biological reagent, including the enzyme catalytic activity, binding affinity and substance transport and degradation functional domain fragments corresponding to each biological reagent;

[0080] In an embodiment of the present invention, based on the previously extracted biological reagent molecular domain data, the molecular functional domain of each biological reagent is further annotated and analyzed. In specific implementation, databases such as KEGG, Reactome, Gene Ontology (GO), etc. can be used to perform functional annotations on each domain. By comparing and annotating the domains, the functional information of each biological reagent is identified, including functional characteristics such as enzyme catalytic activity and molecular binding affinity (obtained on the secondary domain), substance transport and degradation (obtained on the tertiary domain). For example, through GO annotation, the enzyme catalytic functional domain of a protein can be identified, and then its corresponding biochemical reaction type can be inferred; by combining with the KEGG pathway database, the role of the biological reagent in cell signaling can be further inferred. The output of this step is the functional domain data of each biological reagent, specifically including detailed information on functional domains such as enzyme catalytic activity, binding affinity, and substance transport and degradation, and finally the molecular functional domain data of the biological reagent is obtained.

[0081] Step S15: Integrate the biological reagent protein sequence data, the biological reagent molecular structure domain data and the biological reagent molecular function domain data to obtain a biological reagent biomolecule information set.

[0082] In an embodiment of the present invention, by completing the comprehensive data integration of biological reagents, the specific operation includes integrating the protein sequence data, molecular domain data and molecular functional domain data obtained by the previous analysis. This process can be achieved by writing a data processing script and using the Pandas library in Python to merge the data by ID. The data of each biological reagent is integrated into a unified database or data framework, which includes the amino acid sequence of the protein, the distribution of the domain, and the corresponding functional domain information. The integrated biological reagent biomolecule information set contains detailed information of each biological reagent, including amino acid and gene sequences, domain division, functional domain annotation, etc., and finally the biological reagent biomolecule information set is obtained.

[0083] Further, as an embodiment of the present invention, refer to Figure 3 As shown, Figure 2 Detailed step flow diagram of step S13 in the embodiment, step S13 includes the following steps:

[0084] Step S131: Pre-identify the molecular domain of the corresponding amino acid molecular sequence and gene molecular sequence in the biological reagent protein sequence data to obtain the biological reagent amino acid molecular domain region sequence and the biological reagent gene molecular domain region sequence;

[0085] In an embodiment of the present invention, the molecular domains are identified and predicted by the amino acid molecular sequence and the corresponding gene molecular sequence in the biological reagent protein sequence data. First, the protein sequence data is input into the domain identification tool (such as Pfam, InterPro, SMART, etc.), and the known domain model in its database is used to compare the characteristics of the amino acid sequence to identify the corresponding domain region. For the gene molecular sequence, the gene domain prediction is performed by the same method to identify the gene domain associated with the protein amino acid sequence. This process scores and screens the comparison results of the known domains by multiple comparison methods (such as BLAST, HMMER, etc.) to obtain the amino acid molecular domain region sequence of the biological reagent and the gene molecular domain region sequence. These identified domain regions can provide necessary spatial position information for subsequent analysis, and finally obtain the amino acid molecular domain region sequence of the biological reagent and the gene molecular domain region sequence of the biological reagent.

[0086] Step S132: Obtaining corresponding amino acid molecular structure sites and amino acid molecular interaction forces through the amino acid molecular structure domain region sequence of the biological reagent, and performing sequence structure hydrophobicity evaluation and analysis on the corresponding amino acid molecular structure domain region sequence of the biological reagent based on the amino acid molecular structure sites and amino acid molecular interaction forces to obtain the hydrophobicity score of the amino acid molecular structure of the biological reagent;

[0087] In the embodiment of the present invention, based on the sequence of the amino acid molecular domain region of the biological reagent, the amino acid molecular structure sites and the interaction forces between the amino acid molecules are further analyzed by bioinformatics methods. First, the amino acid sequence is subjected to molecular dynamics simulation by using structure prediction software (such as AutoDock, Dock, Rosetta, etc.), and the relative position of each amino acid in the spatial structure and the corresponding interaction sites are analyzed. On this basis, the hydrophobicity of the amino acid molecule is quantitatively evaluated by calculating the hydrophobicity score. This process can be achieved by calculating the hydrophobicity index of the amino acid (such as the Kyte-Doolittle hydrophobicity index) to evaluate the hydrophobicity trend within the amino acid molecular region, wherein the Kyte-Doolittle score (Kyte & Doolittle, 1982) is a method for evaluating the hydrophilicity or hydrophobicity of the amino acid side chain in water. By combining the amino acid molecular structure site and the amino acid interaction force, the hydrophobicity score of the amino acid molecular structure of the biological reagent is calculated, that is, Where n is the total number of amino acid molecular structure sites, x i is the spatial position corresponding to the ith amino acid molecular structure site, P iis the amino acid interaction force corresponding to the ith amino acid molecular structure site, reflecting the hydrophobic characteristics of the region, and finally the hydrophobicity score of the amino acid molecular structure of the biological reagent is obtained.

[0088] Step S133: performing molecular structure functional charge analysis on the amino acid molecular structure domain region sequence of the biological reagent to obtain the amino acid molecular structure functional charge characteristics of the biological reagent, including the amino acid molecular structure charge distribution, the amino acid molecular structure functional charge coupling degree and the amino acid molecular structure charge density;

[0089] In the embodiments of the present invention, by performing molecular structure functional charge analysis on the sequence of the amino acid molecular domain region of the biological reagent, the charge characteristics of the region are further explored, and the charge distribution of the amino acid molecular region is calculated using molecular simulation software (such as APBS, DelPhi, etc.) to obtain the charge distribution of each amino acid residue. Further, through charge coupling analysis, combined with molecular dynamics simulation data, the charge interaction strength between each amino acid in the amino acid molecular domain, that is, the amino acid molecular structure functional charge coupling degree, is calculated. This analysis also includes an assessment of charge density, that is, analyzing the charge density per unit volume in the region, and then inferring its influence on the overall charge characteristics of the biological reagent. These charge characteristic data are helpful to understand the charge behavior of the biological reagent in solution, and finally obtain the charge characteristics of the amino acid molecular structure function of the biological reagent.

[0090] Step S134: Based on the hydrophobicity score of the amino acid molecular structure of the biological reagent and the functional charge characteristics of the amino acid molecular structure of the biological reagent, a structural domain metric calculation is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent using a protein secondary structure metric calculation formula to obtain the secondary structure degree of the amino acid molecular region sequence; based on the secondary structure degree of the amino acid molecular region sequence, a secondary structure calibration process is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent to obtain a secondary structure domain sequence fragment corresponding to the biological reagent;

[0091] In the embodiment of the present invention, a suitable secondary structure metric calculation formula is formed by combining the total length of the amino acid molecular domain sequence, the hydrophobicity score of the amino acid molecular structure of the biological reagent, the hydrophobicity weight factor related to the amino acid, the functional charge distribution of the amino acid molecular structure of the biological reagent, the functional charge coupling degree of the amino acid molecular structure, the charge density of the amino acid molecular structure, the charge characteristic weight factor related to the amino acid, the characteristic parameter of the functional group of the amino acid molecular structure, the functional group characteristic weight factor related to the amino acid and the related parameters, and the corresponding biological reagent amino acid molecular domain region sequence is subjected to the structural domain metric calculation to quantitatively calculate the secondary structure degree of the region, thereby obtaining the secondary structure degree of the amino acid molecular region sequence. At the same time, by combining these secondary structure metrics, the algorithm is used to perform secondary structure calibration, and the final secondary structure domain sequence fragment is determined, that is, the secondary structure metric at the corresponding site is greater than or equal to the threshold value. The goal of this operation is to provide the secondary structure information of the amino acid molecule of the biological reagent in space, and finally obtain the secondary structure domain sequence fragment corresponding to the biological reagent.

[0092] Step S135: Perform molecular tertiary structure deduction analysis on the molecular domain region sequence of the biological reagent gene to obtain the tertiary domain sequence fragment corresponding to the biological reagent.

[0093] In an embodiment of the present invention, a tertiary structure deduction and analysis is performed on the gene molecular domain region sequence of the biological reagent, and a three-dimensional structure prediction is performed on the amino acid sequence obtained by translating the gene sequence using structure prediction software (such as I-TASSER, Phyre2, Rosetta, etc.). The process firstly performs rapid docking on the amino acid sequence to construct its preliminary tertiary structure model, and further optimizes its structure through molecular dynamics simulation, adjusts the spatial position of the amino acid residues, and obtains a stable tertiary structure domain sequence fragment. In the process, simulation technology is used to evaluate the spatial arrangement of amino acids in the three-dimensional structure, molecular interactions, and corresponding binding sites, to obtain a reliable tertiary structure sequence of the biological reagent, and finally obtains a tertiary structure domain sequence fragment corresponding to the biological reagent.

[0094] Furthermore, the protein secondary structure metric calculation formula in step S134 is specifically:

[0095]

[0096] In the formula, S s is the secondary structure degree of the amino acid molecule region sequence, L is the total length of the amino acid molecule domain sequence, r is the domain spatial position parameter, H(r) is the hydrophobicity score of the amino acid molecule structure of the biological reagent at the spatial position r of the amino acid molecule region sequence, α his the hydrophobicity weight factor associated with the amino acid, C(r) is the charge distribution of the amino acid molecular structure function of the biological reagent at the spatial position r of the amino acid molecular region sequence, O(r) is the charge coupling degree of the amino acid molecular structure function of the amino acid molecular region sequence at the spatial position r, M(r) is the charge density of the amino acid molecular structure at the spatial position r of the amino acid molecular region sequence, β c is the charge characteristic weight factor related to the amino acid, F(r) is the characteristic parameter of the amino acid molecular structure functional group in the amino acid molecular region sequence at the spatial position r, γ f is the weight factor of the functional group characteristics related to amino acids, and η is the correction coefficient of the secondary structure degree of the amino acid molecule region sequence.

[0097] The present invention obtains a protein secondary structure measurement calculation formula by using a specific mathematical model and verifying it, which is used to perform domain measurement calculation on the corresponding biological reagent amino acid molecular domain region sequence. The protein secondary structure measurement calculation formula can evaluate the secondary structure characteristics of the amino acid molecular region from multiple dimensions by introducing multiple parameters (such as hydrophobicity, functional charge, functional group characteristics, etc.). Specifically: hydrophobicity plays an important role in protein folding and structure formation, especially affecting the aggregation of internal non-polar amino acid residues. Hydrophobicity has a significant effect on protein stability, folding dynamics and function; the charge distribution of proteins is crucial to their stability, interactions and conformational changes, especially in the process of protein interaction with other molecules (such as ligands, proteins or DNA). The formula helps to more accurately predict the spatial conformation of proteins by considering factors such as charge distribution, charge coupling and charge density; the properties of functional groups in proteins (such as amino groups, carboxyl groups, amides, etc.) also have an important influence on their structural stability and function. This parameter helps capture the changes in the chemical environment in proteins and further enhances the ability to predict the structure. Secondly, in the formula, α h , β c , γ fAs the weight factors of hydrophobicity, charge characteristics and functional group characteristics, they can be adjusted according to different protein types or research needs. This flexibility ensures that the method can be customized and optimized according to different research objectives and data backgrounds, providing more accurate and detailed structural predictions. By adjusting these weight factors, the calculation formula can be optimized to adapt to the properties of different proteins, especially some proteins with special structures or functions. In the formula, r is used as the domain spatial position parameter, which represents the position of the protein molecule domain in space. Through this position dependence, the folding state of the protein in three-dimensional space and the influence of its local environment on the secondary structure can be accurately evaluated. Amino acid residues at different positions have different physicochemical properties. Therefore, the formula takes into account the specific spatial position of each residue, making the evaluation of the secondary structure more detailed. In addition, the adaptability and robustness of the calculation formula are enhanced by the introduction of the correction coefficient. This coefficient allows the calculation formula to be modified as necessary to adapt to the errors in actual data or experiments, further improving the accuracy and practicality of the prediction. By combining multiple factors such as hydrophobicity, charge characteristics, and functional groups for calculation, the formula can give a comprehensive secondary structure metric. This comprehensive evaluation method is not limited to a single characteristic, but fully considers how the protein folds, stabilizes and performs functions in its natural environment, thereby providing a more realistic secondary structure metric. In summary, this formula fully considers the secondary structure degree S of the amino acid molecule region sequence s , the total length of the amino acid molecular domain sequence L, the spatial position parameter of the domain r, the hydrophobicity score H(r) of the amino acid molecular domain sequence of the biological reagent amino acid molecular structure at the spatial position r, and the hydrophobicity weight factor α related to the amino acid h , the charge distribution C(r) of the amino acid molecular structure function of the biological reagent of the amino acid molecular region sequence at the spatial position r, the charge coupling degree O(r) of the amino acid molecular structure function of the amino acid molecular region sequence at the spatial position r, the charge density M(r) of the amino acid molecular structure of the amino acid molecular region sequence at the spatial position r, and the charge characteristic weight factor β related to the amino acid c , the amino acid molecular structure functional group characteristic parameter F(r) of the amino acid molecular region sequence at spatial position r, and the functional group characteristic weight factor γ related to the amino acid f , the correction coefficient η of the secondary structure degree of the amino acid molecule region sequence, according to the secondary structure degree S of the amino acid molecule region sequence s The correlation between the above parameters constitutes a functional relationship The formula can realize the domain measurement calculation process of the corresponding biological reagent amino acid molecular domain region sequence. At the same time, by introducing the correction coefficient η of the secondary structure degree of the amino acid molecular region sequence, it can be adjusted according to the error situation occurring in the calculation process, thereby improving the accuracy and applicability of the protein secondary structure measurement calculation formula.

[0098] Further, step S135 includes the following steps:

[0099] Conducting gene molecular dynamics simulation analysis on the sequence of the molecular structure domain region of the biological reagent gene to generate the biological reagent gene molecular dynamics simulation process;

[0100] In an embodiment of the present invention, the domain region sequence in the biological reagent gene molecule is extracted to obtain the sequence information of the target gene, especially the relevant domain region sequence, through high-throughput sequencing technology or gene chip technology, and then the gene molecule is dynamically simulated using molecular dynamics simulation software (such as GROMACS, AMBER or CHARMM). Specifically, an appropriate force field (for example, CHARMM27 or OPLS-AA) is selected, and the gene molecule domain region sequence is mapped to the simulation space, and a system model is constructed by using these software, and an appropriate temperature (such as 300K), pressure (such as 1atm) and time step (such as 2fs) are set. The simulation process needs to be carried out on a time scale of at least several hundred nanoseconds to ensure that the dynamic process can fully converge. During the simulation process, the energy minimization technology is used to remove the errors caused by the structure, and the molecular behavior changes are analyzed through molecular dynamics trajectory data to generate a gene molecular dynamics simulation process file, which records the structural changes and energy fluctuations of the molecule in the time evolution, and finally generates the biological reagent gene molecular dynamics simulation process.

[0101] Preferably, the gene molecule folding path is deduced for the biological agent gene molecular dynamics simulation process to generate a biological agent gene molecular folding deduction path diagram;

[0102] In an embodiment of the present invention, the folding process of the gene molecule from the initial random conformation to the final stable conformation is analyzed according to the trajectory data in the dynamics simulation process of the biological reagent gene molecule. To this end, a structural biology tool (such as Rosetta or Folding@Home) is used to perform a detailed analysis of the dynamics simulation trajectory, and the conformational changes of the molecule at each moment in the simulation process, especially the changes in the key structural domain area, are gradually analyzed to deduce the folding path of the gene molecule. Generally, the analysis step includes the dynamic process of identifying the secondary structure (such as α-helix and β-fold) formed in the folding process, and inferring the energy change in the folding process based on the intermolecular interaction force (such as hydrogen bond, van der Waals force, electrostatic force, etc.). By analyzing the deduced path, a folding path diagram is drawn to reflect the various possible pathways of molecular folding and the energy state of each pathway. This graph represents the process of the biological reagent gene molecule from the unfolded state to the final folding, and finally generates a biological reagent gene molecule folding deduction path diagram.

[0103] Preferably, based on the biological agent gene molecule folding deduction pathway diagram, a domain folding energy distribution analysis is performed on the corresponding biological agent gene molecule domain region sequence to obtain a biological agent gene molecule domain folding energy distribution diagram;

[0104] In an embodiment of the present invention, according to the folding path diagram, for each key domain region, the domain folding energy analysis method is used to further analyze the energy changes experienced by each region during the folding process, so as to use energy calculation tools (such as FoldX, Rosetta Energy Minimization, etc.) to quantitatively analyze the folding energy of the biological reagent gene molecule, and by calculating the total energy of the gene molecule in different conformations, including factors such as van der Waals energy, electrostatic energy, and solvent accessibility, the energy difference of different folding states is determined, and these data are visualized into an energy distribution diagram, reflecting the concentrated distribution areas of energy and the folding areas with lower energy under different folding states. This diagram can help further understand which domain areas are more stable in energy and which areas are more unstable or bottleneck areas of folding during the gene molecule folding process, and finally obtain the domain folding energy distribution diagram of the biological reagent gene molecule.

[0105] Preferably, the corresponding gene molecule folding energy stability score and gene molecule folding energy distribution gradient are obtained through the folding energy distribution diagram of the biological reagent gene molecule domain, and the corresponding biological reagent gene molecule domain region sequence is matched and calculated using the protein tertiary structure matching calculation formula based on the gene molecule folding energy stability score and the gene molecule folding energy distribution gradient, so as to obtain the tertiary structure matching degree of the gene molecule region sequence;

[0106] In an embodiment of the present invention, the folding stability score of each domain region is calculated according to the folding energy distribution diagram, and the scoring basis includes the position of the energy trough, the energy distribution, and the stability of the domain region during the folding process, and the gradient value corresponding to the energy distribution is statistically analyzed to obtain the gene molecule folding energy stability score and the gene molecule folding energy distribution gradient. At the same time, by combining the gene molecule folding space range, the gene molecule folding space coordinates, the folding energy Hamiltonian, the gene molecule folding energy adjustment coefficient, the gene molecule folding energy stability score, the gene molecule folding energy stability influencing parameter, the gene molecule folding energy distribution gradient, the folding energy distribution gradient influencing coefficient and related parameters, a suitable tertiary structure matching calculation formula is formed to match the corresponding biological reagent gene molecule domain region sequence, so as to match the target gene molecule by comparing the secondary structure information of the target gene molecule with the known protein three-dimensional structure database (such as PDB), and evaluate the similarity of the target domain region with the known protein three-dimensional structure in the database. This value represents the similarity of the tertiary structure of the gene molecule region sequence, and finally obtains the tertiary structure matching degree of the gene molecule region sequence.

[0107] Among them, the protein tertiary structure matching calculation formula is as follows:

[0108]

[0109] In the formula, M is the tertiary structure matching degree of the gene molecule region sequence, Ω is the range of the gene molecule folding space, and x is the coordinate of the gene molecule folding space. is the folding energy Hamiltonian at the gene molecule folding space coordinate x, δ is the gene molecule folding energy adjustment coefficient, is the gradient symbol, U fold (x) is the gene molecule folding energy stability score at the gene molecule folding space coordinate x, ε is the gene molecule folding energy stability influencing parameter, is the gene molecule folding energy distribution gradient, θ is the folding energy distribution gradient influence coefficient, ξ is the correction coefficient of the tertiary structure matching degree of the gene molecule region sequence;

[0110] The present invention obtains a protein tertiary structure matching calculation formula by using a specific mathematical model and verifying that the corresponding biological reagent gene molecular structure domain region sequence is matched and calculated. The protein tertiary structure matching calculation formula can calculate and predict the tertiary structure of the biological reagent gene molecular structure domain region sequence through an accurate mathematical model. This process not only improves the understanding of the gene molecule folding behavior, but also provides valuable references for downstream biological experiments. Multiple parameters in the formula (such as folding energy Hamiltonian, folding energy stability score, folding energy distribution gradient, etc.) can help adjust and optimize the folding path and energy distribution in real time during the folding process. Through the systematic deduction of the folding path and energy stability, the folding path with high stability and reasonable structure can be effectively identified and selected, thereby providing a scientific basis for the precise design of the gene molecular structure. The folding energy adjustment coefficient, stability score, influencing parameter, and energy distribution gradient influence coefficient in the formula can all be used to quantify the folding stability and efficiency of the protein. Through the adjustment of these coefficients, the trend, stability, and efficiency of protein folding can be predicted under different folding conditions, helping to select appropriate conditions for experiments and reducing the complexity of the experiment. The matching value in the protein tertiary structure matching calculation formula provides a quantitative indicator for the final structure prediction, indicating the similarity between the predicted tertiary structure and the real structure, which helps to evaluate the relative structural adaptability of the gene molecule region sequence of the biological reagent under different folding paths, so as to determine the sequence closest to its natural folding state. The use of this formula is not only applicable to conventional protein folding prediction, but also can be used for personalized design and optimization of specific biological reagent gene molecules. For example, when designing protein drugs or biomaterials with specific functions, the formula can accurately predict and optimize its tertiary structure according to its domain folding path, energy distribution and other characteristics, and improve its functional performance and stability. By calibrating the tertiary structure matching degree of the gene molecule region sequence, the corresponding folding results can be deduced before the actual experiment, and the optimal solution can be selected for the experiment, which greatly improves the success rate of biological experiments and avoids a large number of repeated experiments and inefficient attempts. In summary, the formula fully considers the tertiary structure matching degree M of the gene molecule region sequence, the gene molecule folding space range Ω, the gene molecule folding space coordinate x, and the folding energy Hamiltonian at the gene molecule folding space coordinate x. , gene molecule folding energy adjustment coefficient δ, gradient sign , the gene molecule folding energy stability score U at the gene molecule folding space coordinate x fold (x), gene molecule folding energy stability influencing parameter ε, gene molecule folding energy distribution gradient , the folding energy distribution gradient influence coefficient θ, the correction coefficient ξ of the tertiary structure matching degree of the gene molecular region sequence, and the mutual correlation between the tertiary structure matching degree M of the gene molecular region sequence and the above parameters constitute a functional relationship The formula can realize the matching calculation process of the corresponding biological reagent gene molecular domain region sequence. At the same time, by introducing the correction coefficient ξ of the tertiary structure matching degree of the gene molecular region sequence, it can be adjusted according to the error situation occurring in the calculation process, thereby improving the accuracy and applicability of the protein tertiary structure matching calculation formula.

[0111] Preferably, the tertiary structure calibration process is performed on the corresponding biological reagent gene molecular domain region sequence based on the tertiary structure matching degree of the gene molecular region sequence to obtain the tertiary structure domain sequence fragment corresponding to the biological reagent.

[0112] In an embodiment of the present invention, according to the tertiary structure matching degree of the gene molecular region sequence obtained in the previous step, the sequence fragments with a matching degree exceeding a threshold are selected as the target region for tertiary structure calibration processing, so as to visualize these regions by using structural biology software (such as PyMOL, Chimera, etc.), and further adjust and optimize their three-dimensional structures. Specifically, the target sequence region is compared with the existing high-resolution protein three-dimensional structure model, and the spatial structure of the target sequence is adjusted according to the spatial arrangement and interaction force of the domain, and the most likely tertiary structure model is gradually generated. This process involves backtracking optimization of the model (such as energy minimization and constraint adjustment) until a stable structural state is reached, and finally the tertiary structure domain sequence fragment corresponding to the biological reagent is obtained.

[0113] Further, step S15 includes the following steps:

[0114] Step S151: performing functional domain division on the corresponding secondary structure domain sequence fragments in the biological reagent molecule structure domain data to obtain biological reagent molecule secondary structure functional domain type segments, including biological reagent molecule enzyme functional domain fragments and biological reagent molecule binding functional domain fragments;

[0115] In the embodiment of the present invention, by analyzing the corresponding secondary domain sequence fragments in the biological reagent molecule domain data, the secondary domain fragments therein are identified and divided into functional domains. This process is based on the amino acid sequence of the biological reagent molecule, and the secondary structure prediction software (such as PSIPRED, JPred or I-TASSER) is used to predict the secondary structure. The amino acid sequence is predicted by the tool, and the functional domain is annotated and confirmed in combination with the known biological functional domain database (such as Pfam, SMART, InterPro). Each secondary domain is based on its amino acid sequence and The prediction results are assigned to specific functional domain types, including enzyme functional domains and binding functional domains. Enzyme functional domains include fragments with catalytic activity, such as classification based on certain enzyme catalytic mechanisms (such as hydrolysis, oxidation-reduction, transfer, etc.); binding functional domains are related to binding to specific molecules or ions, such as receptor binding domains or DNA binding domains. Through the above tools and methods, the secondary structure functional domains of biological reagents are divided into multiple fragments and classified into different functional domain types, and finally the secondary structure functional domain type segmentation of biological reagent molecules is obtained, including biological reagent molecule enzyme functional domain fragments and biological reagent molecule binding functional domain fragments.

[0116] Step S152: obtaining known catalytic enzyme reaction mechanisms and enzyme substrate characteristics, and performing catalytic potential prediction analysis on the enzyme functional domain fragments of the biological reagent molecule based on the known catalytic enzyme reaction mechanisms and enzyme substrate characteristics to generate a biological reagent molecule catalytic potential score; performing enzyme catalytic activity annotation processing on the corresponding biological reagent molecule enzyme functional domain fragments based on the biological reagent molecule catalytic potential score to obtain the enzyme catalytic activity functional domain fragments corresponding to each biological reagent;

[0117] In an embodiment of the present invention, a predictive analysis of the catalytic potential of the enzyme is performed based on the sequence fragments of the enzyme functional domain of the biological reagent, combined with known catalytic enzyme reaction mechanisms and enzyme substrate characteristic data. This process obtains the catalytic mechanisms and enzyme substrate characteristics of different enzyme reactions by consulting known enzyme reaction databases (such as BRENDA, KEGG, EC numbers, etc.), and uses the laws of the enzyme catalytic reaction mechanism. Based on the similarity of enzyme substrate characteristics and catalytic mechanisms, a machine learning algorithm (such as support vector machine, random forest, etc.) is used for model training and prediction to generate a catalytic potential score for each enzyme functional domain fragment. This score reflects the potential size of the functional domain in the catalytic reaction based on the prediction of the catalytic activity. The higher the score, the stronger the catalytic potential of the enzyme functional domain, thereby predicting and generating a catalytic potential score for the biological reagent molecule. At the same time, the corresponding enzyme functional domain fragments of the biological reagent molecules are annotated with the enzyme catalytic activity by combining the catalytic potential scores of the biological reagent molecules obtained in the previous analysis. Based on the catalytic potential scores obtained previously, each enzyme functional domain fragment will be assigned a catalytic activity score, which reflects the actual enzyme catalytic ability of the functional domain. On this basis, by analyzing the substrate characteristics, reaction mechanism and experimental data of the enzyme (such as enzymatic parameters: Km, Vmax, etc.), the enzyme catalytic activity is further annotated in detail. This process can be assisted by the enzymatic experimental data model (such as the Michaelis-Menten equation) to verify and correct the prediction results to ensure the accuracy of the catalytic activity annotation. According to the catalytic potential score and activity annotation, the activity evaluation of each enzyme functional domain fragment is completed, and then the catalytic activity level of each enzyme functional domain is annotated, and finally the enzyme catalytic activity functional domain fragment corresponding to each biological reagent is obtained.

[0118] Step S153: simulating different environmental factor constraints and dynamically annotating the binding affinity of the biological reagent molecule binding domain fragments based on the different environmental factor constraints to obtain the binding affinity domain fragments corresponding to each biological reagent;

[0119] In an embodiment of the present invention, the binding affinity of the binding domain fragment of the biological reagent molecule is dynamically annotated. First, based on the known binding domain type (such as receptor binding domain, protein-protein interaction domain, etc.) and the molecular structure data of the binding ligand, the molecular docking software (such as AutoDock, DOCK, GOLD, etc.) is used to simulate the influence of different environmental factors (such as pH value, temperature, ion concentration, etc.) on the binding affinity. By setting these environmental factors and dynamically simulating and calculating the energy of the binding process, the binding affinity prediction value of each binding domain fragment under different environmental conditions is obtained. The predicted value of the binding affinity is based on the score function of the molecular docking, the binding constant (Kd), the binding free energy (ΔG) and other parameters, thereby providing the annotation data of the binding affinity for each biological reagent molecule, and finally obtaining the binding affinity domain fragment corresponding to each biological reagent.

[0120] Step S154: performing a transmembrane transport capacity prediction analysis on the tertiary domain sequence fragments corresponding to the biological reagent molecular structure domain data to obtain a predicted value of the transmembrane transport capacity of the biological reagent molecular structure;

[0121] In an embodiment of the present invention, the transmembrane transport ability of the corresponding tertiary domain sequence fragments in the biological reagent molecular domain data is predicted and analyzed, and the transport potential is predicted by using a known transmembrane transport protein database (such as TCDB) and transport mechanism, combined with the amino acid sequence of the biological reagent molecule and its tertiary structure information, using a transmembrane protein prediction tool (such as TMHMM, Phobius, TOPCONS, etc.). The tool predicts whether the biological reagent molecule has the ability of transmembrane transport by analyzing the transmembrane helical structure and its hydrophilicity and hydrophobicity in the amino acid sequence. On this basis, according to the characteristics of the membrane protein and experimental data, a score of transmembrane transport ability is further provided for each tertiary domain fragment, indicating the possibility of the functional domain fragment as a transmembrane transport protein, and finally obtaining a predicted value of the transmembrane transport ability of the biological reagent molecular structure.

[0122] Step S155: Based on the predicted value of the transmembrane transport capacity of the biological reagent molecular structure, the corresponding tertiary structural domain sequence fragments are annotated for substance transport degradation to obtain the substance transport degradation functional domain fragments corresponding to each biological reagent.

[0123] In an embodiment of the present invention, by combining the previously obtained predicted value of transmembrane transport capacity, the tertiary domain fragments of the biological reagent molecule are further annotated for material transport degradation. By simulating the influence of different environmental conditions (such as temperature, pH, ion concentration, etc.) on the material transport process, dynamic simulation tools (such as GROMACS, AMBER, etc.) are used to predict whether the biological reagent molecule performs the degradation function of material transport on the cell membrane under a specific environment. In combination with the existing transport reaction kinetic model (such as Michaelis-Menten kinetics or Hill equation), the transport degradation characteristics of each tertiary domain fragment are annotated, and by predicting the transport capacity and degradation activity of the functional domain under different environmental conditions, an annotation label of the material transport degradation functional domain is given to each functional domain fragment, and finally the material transport degradation functional domain fragment corresponding to each biological reagent is obtained.

[0124] Further, step S2 includes the following steps:

[0125] Step S21: performing sequence alignment analysis on each protein sequence in the biological reagent protein sequence data to obtain a biological reagent protein sequence alignment matrix, which includes alignment scores, matching lengths, and gap regions between each protein sequence;

[0126] In the embodiment of the present invention, a multiple sequence alignment (MSA) is performed on each protein sequence in the biological reagent protein sequence data. Common alignment tools such as BLAST (Basic Local Alignment Search Tool) or Clustal Omega can be used to align the protein sequences. The input protein sequence data is first preprocessed to remove low-quality sequences and standardize the formats of all sequences (such as FASTA format). During the alignment process, the scores of amino acid substitutions are calculated by setting a suitable alignment score matrix (such as BLOSUM or PAM), and suitable gap penalties and extension penalties are set. Penalty) is used to process insertion and deletion regions. The alignment tool will output an alignment matrix according to the set parameters, which includes the alignment score, matching length, and insertion and deletion regions (gaps) between each pair of protein sequences. The alignment score reflects the similarity of the protein sequences, the matching length indicates the number of consecutively matched amino acids between the two sequences in the alignment, and insertions and deletions refer to the unmatched regions in the sequence. By analyzing this information, a protein sequence alignment matrix can be obtained, and finally a biological reagent protein sequence alignment matrix can be obtained.

[0127] Step S22: performing sequence difference statistical analysis on the corresponding protein sequences in the biological reagent protein sequence data based on the biological reagent protein sequence comparison matrix to obtain the biological reagent protein sequence comparison features, including the corresponding amino acid residue substitution, insertion and deletion difference features between the protein sequences;

[0128] In an embodiment of the present invention, after completing the protein sequence alignment matrix, the corresponding protein sequence is subjected to sequence difference statistical analysis. The goal of this step is to identify and count the differences between different protein sequences, mainly including amino acid residue substitutions, insertions and deletions. Using the alignment results, the residue differences of each aligned sequence are first extracted, and the frequency of amino acid substitutions is counted by calculating the changes in amino acids at each aligned position. At the same time, the insertion and deletion positions in the sequence are analyzed to identify which regions have insertions or deletions. Specifically, the Biopython library in the Python programming language can be used to parse the alignment data and perform difference statistics, and the amino acid changes of each pair of protein sequences are checked one by one through the alignment matrix, and the type of substitution (such as polar-polar substitution, acid-alkaline substitution, etc.) is recorded. Based on these difference information, a sequence difference statistical analysis table is established, which includes the substitution type of each amino acid position, the position of the insertion and deletion region, and the statistical data of the substitution, and finally the biological reagent protein sequence alignment feature is obtained.

[0129] Step S23: performing protein structure feature prediction analysis on the biological agent molecular domain data to generate biological agent molecular domain features, including molecular domain stability, molecular domain hydrophobic interaction force, and molecular domain binding energy features;

[0130] In an embodiment of the present invention, the protein structural features are predicted and analyzed by analyzing the structural domains of the protein sequences through the previously extracted biological reagent molecular domain data. First, the three-dimensional structure of each protein sequence is predicted using protein structure prediction software (such as Phyre2, I-TASSER). This process predicts the spatial structure of the protein sequence based on the sequence information, and obtains the conformation and interaction information of the structural domain. On this basis, the stability of the structural domain can be further calculated, for example, by evaluating the free energy of the protein or by simulating the protein folding process to infer the stability. The hydrophobic interaction force refers to the non-polar interaction between the hydrophobic amino acid residues in the protein molecule. The energy change of this interaction can be calculated using molecular dynamics simulation (such as GROMACS software), and the binding energy characteristics of the molecular domain are further calculated by combining the molecular dynamics simulation results to predict the binding mode and binding strength of the domain with other molecules. The results of these analyses will help understand the functionality and stability of the structural domain, and ultimately generate the molecular structural domain characteristics of the biological reagent, including the molecular structural domain stability, the molecular structural domain hydrophobic interaction force, and the molecular structural domain binding energy characteristics.

[0131] Step S24: extract specific molecular function features based on the biological reagent molecular function domain data to obtain the biological reagent specific molecular function domain features, including the biological reagent enzyme activity level, biological reagent binding affinity and biological reagent material transport and degradation ability characteristics.

[0132] In an embodiment of the present invention, specific functional characteristics of proteins are extracted based on molecular functional domain data of biological reagents. First, the protein sequence is mapped to a known functional domain database through bioinformatics tools (such as InterProScan, Pfam) to determine the functional domains contained therein and their corresponding biological functions. For each determined functional domain, the enzyme activity level is predicted using existing experimental data or computational models (such as Enzyme Commission number, EC number), and the binding affinity of the protein is calculated. The binding affinity can be used to predict the binding strength of the protein with its ligand through molecular docking analysis (using tools such as AutoDock and Dock). In addition, combined with the transport and degradation ability characteristic data of the biological reagent, the activity of the protein in the material transport or degradation process can be further predicted using a biological process network (such as the KEGG database) or a molecular simulation tool. These specific molecular functional characteristics include enzyme activity, binding affinity and transport and degradation ability, and finally the specific molecular functional domain characteristics of the biological reagent are obtained.

[0133] Further, step S3 includes the following steps:

[0134] Step S31: constructing a biological reagent function prediction network framework by connecting a convolutional neural network CNN, a recurrent neural network RNN, and a Transformer neural network;

[0135] In an embodiment of the present invention, a neural network framework for predicting the function of a biological agent is constructed by using a combination of a convolutional neural network (CNN), a recurrent neural network (RNN) and a Transformer neural network. In a specific implementation, CNN is used to extract local features, RNN processes the time dependency of feature sequence data, and Transformer captures global dependencies through its self-attention mechanism. First, the protein sequence features of the biological agent are input into the CNN model, and CNN extracts local features in the protein sequence, such as the arrangement of amino acids and local structural information, through convolution operations. Then, the output of CNN is used as the input of RNN to capture long-distance dependencies in the protein sequence, such as the interaction between structural domains and functional domains, and the output of RNN is input into the Transformer network. The multi-head self-attention mechanism in the Transformer is used to further extract complex global features in the sequence, and these features are fused and output through a fully connected layer, and finally a biological agent function prediction network framework is constructed.

[0136] Step S32: inputting the biological reagent protein sequence alignment features, the biological reagent molecular domain features and the biological reagent specific molecular functional domain features into the biological reagent function prediction network framework to train and construct a function prediction model, while considering the activity, structural stability and specific function prediction targets corresponding to the biological reagent to generate a biological reagent function prediction model;

[0137] In an embodiment of the present invention, the protein sequence alignment features, molecular domain features and molecular functional domain features of the biological reagent are input into the constructed function prediction network framework. After being processed by the aforementioned CNN, RNN and Transformer networks, these features will be sent to an integrated learning model (such as a multi-layer perceptron, MLP) to train the function prediction model. During the training process, the network not only considers the basic molecular characteristics of the biological reagent, but also includes target information such as the activity, structural stability and specific function of the biological reagent. These targets serve as supervisory signals to help the network perform multi-task learning, thereby improving the prediction accuracy of the model. Specifically, the loss function in the network will contain multiple sub-items, each of which corresponds to a different target prediction task, such as activity prediction, stability evaluation, etc., and the losses of these targets are optimized through back propagation, so that the model can learn multiple aspects of prediction information at the same time, thereby generating an accurate function prediction model, and finally generating a biological reagent function prediction model.

[0138] Step S33: obtaining molecular features of the biological agent to be predicted, including protein sequence alignment features, molecular structural domain features, and molecular functional domain features corresponding to the biological agent to be predicted;

[0139] In an embodiment of the present invention, corresponding molecular features are obtained for the biological agent to be predicted, and these features include protein sequence alignment features, molecular domain features, and molecular functional domain features of the biological agent to be predicted. First, the protein sequence alignment features can be compared with the sequences in the known protein database (such as UniProt) to extract the conserved regions, sequence similarity, and variation information. Secondly, the molecular domain features are obtained by predicting the functional domains of the protein sequence (such as the Pfam database). These domains reflect the functional regions of the protein, and the molecular functional domain features can be predicted by functional annotation tools (such as InterPro) to provide features directly related to the molecular function. These features are combined to form a complete feature set of the biological agent to be predicted, and finally the molecular features of the biological agent to be predicted are obtained.

[0140] Step S34: inputting the molecular features of the biological agent to be predicted into the biological agent function prediction model to perform target function prediction to obtain a target biological agent function prediction result.

[0141] In an embodiment of the present invention, by inputting the molecular features of the biological reagent to be predicted into the previously trained biological reagent function prediction model, the specific operation includes inputting the protein sequence alignment features, molecular domain features and molecular functional domain features of the biological reagent to be predicted into the model in a predetermined format. The model uses the aforementioned deep neural network (a combination of CNN, RNN and Transformer) and a multi-task learning framework based on these input data to comprehensively predict the function of the biological reagent. The prediction targets usually include molecular activity, structural stability and specific functions, etc. After being processed by the model, this information will be output in the form of probability, and each function will be scored. The results output by the model can provide researchers with the function prediction results of the target biological reagent, and finally obtain the function prediction results of the target biological reagent.

[0142] Furthermore, the present invention also provides a biological reagent function prediction system for executing the biological reagent function prediction method as described above, the biological reagent function prediction system comprising:

[0143] A biological reagent molecule information collection module is used to collect biological reagent molecule information from a biological reagent public database to obtain a biological reagent biological reagent molecule information set, wherein the biological reagent biological reagent molecule information set includes biological reagent protein sequence data, biological reagent molecule structure domain data, and biological reagent molecule function domain data;

[0144] The biological reagent molecular feature analysis module is used to perform sequence comparison analysis on the biological reagent protein sequence data to obtain the biological reagent protein sequence comparison features; perform protein structure feature prediction analysis on the biological reagent molecular domain data to generate the biological reagent molecular domain features; extract specific molecular function features based on the biological reagent molecular function domain data to obtain the specific molecular function domain features of the biological reagent;

[0145] The biological reagent target function prediction module is used to construct a biological reagent function prediction network framework, and input the biological reagent protein sequence alignment features, biological reagent molecular domain features and biological reagent specific molecular function domain features into the biological reagent function prediction network framework for function prediction model training and construction to generate a biological reagent function prediction model; obtain the molecular features of the biological reagent to be predicted, and input the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction to obtain the target biological reagent function prediction result;

[0146] The biological reagent function verification and optimization module is used to perform function verification and optimization on the corresponding biological reagent to be predicted according to the function prediction result of the target biological reagent, so as to generate the function optimization result of the target biological reagent.

[0147] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for predicting the function of a biological agent, characterized in that: The following steps are involved: Step S1: collecting biomolecule information from a public database of biological reagents to obtain a biomolecule information set of biological reagents, wherein the biomolecule information set of biological reagents includes biomolecule protein sequence data, biomolecule structural domain data, and biomolecule functional domain data of biological reagents; Step S2: performing sequence alignment analysis on the biological reagent protein sequence data to obtain biological reagent protein sequence alignment features; Perform protein structure feature prediction analysis on biological reagent molecular domain data to generate biological reagent molecular domain features; Extract specific molecular functional features based on the molecular functional domain data of biological reagents to obtain specific molecular functional domain features of biological reagents; Step S3: constructing a biological reagent function prediction network framework, and inputting the biological reagent protein sequence alignment features, biological reagent molecular domain features, and biological reagent specific molecular functional domain features into the biological reagent function prediction network framework for function prediction model training and construction, so as to generate a biological reagent function prediction model; obtaining the molecular features of the biological reagent to be predicted, and inputting the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction, so as to obtain the target biological reagent function prediction result; Step S4: According to the function prediction result of the target biological reagent, the corresponding biological reagent to be predicted is functionally verified and optimized to generate the function optimization result of the target biological reagent.

2. The method for predicting the function of a biological agent according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Preliminary screening of the original data of biological reagents in the biological reagent public database to obtain a set of original biological information of biological reagents; Step S12: performing protein sequence extraction processing on the original biological information set of the biological reagent to obtain biological reagent protein sequence data, including the amino acid molecular sequence and gene molecular sequence corresponding to each biological reagent; Step S13: performing molecular domain recognition, calibration and extraction processing on the amino acid molecular sequence and gene molecular sequence corresponding to the biological reagent protein sequence data to obtain biological reagent molecular domain data, including the secondary domain sequence fragment and the tertiary domain sequence fragment corresponding to each biological reagent; Step S14: performing molecular functional domain annotation and parsing processing on the molecular structural domain data of the biological reagent to obtain the molecular functional domain data of the biological reagent, including the enzyme catalytic activity, binding affinity and substance transport and degradation functional domain fragments corresponding to each biological reagent; Step S15: Integrate the biological reagent protein sequence data, the biological reagent molecular structure domain data and the biological reagent molecular function domain data to obtain a biological reagent biomolecule information set.

3. The method for predicting the function of a biological agent according to claim 2, characterized in that: The biological reagent original biological information set in step S11 includes biological reagent original information data corresponding to enzymes, antibodies and microorganism types.

4. The method for predicting the function of a biological agent according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: Pre-identify the molecular domain of the corresponding amino acid molecular sequence and gene molecular sequence in the biological reagent protein sequence data to obtain the biological reagent amino acid molecular domain region sequence and the biological reagent gene molecular domain region sequence; Step S132: Obtaining corresponding amino acid molecular structure sites and amino acid molecular interaction forces through the amino acid molecular structure domain region sequence of the biological reagent, and performing sequence structure hydrophobicity evaluation and analysis on the corresponding amino acid molecular structure domain region sequence of the biological reagent based on the amino acid molecular structure sites and amino acid molecular interaction forces to obtain the hydrophobicity score of the amino acid molecular structure of the biological reagent; Step S133: performing molecular structure functional charge analysis on the amino acid molecular structure domain region sequence of the biological reagent to obtain the amino acid molecular structure functional charge characteristics of the biological reagent, including the amino acid molecular structure charge distribution, the amino acid molecular structure functional charge coupling degree and the amino acid molecular structure charge density; Step S134: Based on the hydrophobicity score of the amino acid molecular structure of the biological reagent and the functional charge characteristics of the amino acid molecular structure of the biological reagent, a structural domain metric calculation is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent using a protein secondary structure metric calculation formula to obtain the secondary structure degree of the amino acid molecular region sequence; based on the secondary structure degree of the amino acid molecular region sequence, a secondary structure calibration process is performed on the corresponding amino acid molecular structure domain region sequence of the biological reagent to obtain a secondary structure domain sequence fragment corresponding to the biological reagent; Step S135: Perform molecular tertiary structure deduction analysis on the molecular domain region sequence of the biological reagent gene to obtain the tertiary domain sequence fragment corresponding to the biological reagent.

5. The method for predicting the function of a biological agent according to claim 4, characterized in that: The protein secondary structure metric calculation formula in step S134 is specifically: In the formula, S s is the secondary structure degree of the amino acid molecule region sequence, L is the total length of the amino acid molecule domain sequence, r is the domain spatial position parameter, H(r) is the hydrophobicity score of the amino acid molecule structure of the biological reagent at the spatial position r of the amino acid molecule region sequence, α h is the hydrophobicity weight factor associated with the amino acid, C(r) is the charge distribution of the amino acid molecular structure function of the biological reagent at the spatial position r of the amino acid molecular region sequence, O(r) is the charge coupling degree of the amino acid molecular structure function of the amino acid molecular region sequence at the spatial position r, M(r) is the charge density of the amino acid molecular structure at the spatial position r of the amino acid molecular region sequence, β c is the charge characteristic weight factor related to the amino acid, F(r) is the characteristic parameter of the amino acid molecular structure functional group in the amino acid molecular region sequence at spatial position r, γ f is the weight factor of the functional group characteristics related to amino acids, and η is the correction coefficient of the secondary structure degree of the amino acid molecule region sequence.

6. The method for predicting the function of a biological agent according to claim 4, characterized in that: Step S135 includes the following steps: Conducting gene molecular dynamics simulation analysis on the sequence of the molecular structure domain region of the biological reagent gene to generate the biological reagent gene molecular dynamics simulation process; Perform gene molecule folding path deduction on the biological reagent gene molecular dynamics simulation process to generate a biological reagent gene molecular folding deduction path diagram; Based on the biological reagent gene molecular folding deduction path diagram, the domain folding energy distribution analysis is performed on the corresponding biological reagent gene molecular domain region sequence to obtain the biological reagent gene molecular domain folding energy distribution diagram; Obtain the corresponding gene molecule folding energy stability score and gene molecule folding energy distribution gradient through the folding energy distribution diagram of the biological reagent gene molecule domain, and perform matching calculation on the corresponding biological reagent gene molecule domain region sequence using the protein tertiary structure matching calculation formula based on the gene molecule folding energy stability score and the gene molecule folding energy distribution gradient, so as to obtain the tertiary structure matching degree of the gene molecule region sequence; Among them, the protein tertiary structure matching calculation formula is as follows: In the formula, M is the tertiary structure matching degree of the gene molecule region sequence, Ω is the range of the gene molecule folding space, and x is the coordinate of the gene molecule folding space. is the folding energy Hamiltonian at the gene molecule folding space coordinate x, δ is the gene molecule folding energy adjustment coefficient, is the gradient symbol, U fold (x) is the gene molecule folding energy stability score at the gene molecule folding space coordinate x, ε is the gene molecule folding energy stability influencing parameter, is the gene molecule folding energy distribution gradient, θ is the folding energy distribution gradient influence coefficient, ξ is the correction coefficient of the tertiary structure matching degree of the gene molecule region sequence; Based on the tertiary structure matching degree of the gene molecule region sequence, the corresponding biological reagent gene molecule domain region sequence is subjected to tertiary structure calibration processing to obtain the tertiary structure domain sequence fragment corresponding to the biological reagent.

7. The method for predicting the function of a biological agent according to claim 2, characterized in that: Step S15 includes the following steps: Step S151: performing functional domain division on the corresponding secondary structure domain sequence fragments in the biological reagent molecule structure domain data to obtain biological reagent molecule secondary structure functional domain type segments, including biological reagent molecule enzyme functional domain fragments and biological reagent molecule binding functional domain fragments; Step S152: obtaining known catalytic enzyme reaction mechanisms and enzyme substrate characteristics, and performing catalytic potential prediction analysis on the enzyme functional domain fragments of the biological reagent molecule based on the known catalytic enzyme reaction mechanisms and enzyme substrate characteristics to generate a biological reagent molecule catalytic potential score; performing enzyme catalytic activity annotation processing on the corresponding biological reagent molecule enzyme functional domain fragments based on the biological reagent molecule catalytic potential score to obtain the enzyme catalytic activity functional domain fragments corresponding to each biological reagent; Step S153: simulating different environmental factor constraints and dynamically annotating the binding affinity of the biological reagent molecule binding domain fragments based on the different environmental factor constraints to obtain the binding affinity domain fragments corresponding to each biological reagent; Step S154: performing a transmembrane transport capacity prediction analysis on the tertiary domain sequence fragments corresponding to the biological reagent molecular structure domain data to obtain a predicted value of the transmembrane transport capacity of the biological reagent molecular structure; Step S155: Based on the predicted value of the transmembrane transport capacity of the biological reagent molecular structure, the corresponding tertiary structural domain sequence fragments are annotated for substance transport degradation to obtain the substance transport degradation functional domain fragments corresponding to each biological reagent.

8. The method for predicting the function of a biological agent according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing sequence alignment analysis on each protein sequence in the biological reagent protein sequence data to obtain a biological reagent protein sequence alignment matrix, which includes alignment scores, matching lengths, and gap regions between each protein sequence; Step S22: performing sequence difference statistical analysis on the corresponding protein sequences in the biological reagent protein sequence data based on the biological reagent protein sequence comparison matrix to obtain the biological reagent protein sequence comparison features, including the corresponding amino acid residue substitution, insertion and deletion difference features between the protein sequences; Step S23: performing protein structure feature prediction analysis on the biological agent molecular domain data to generate biological agent molecular domain features, including molecular domain stability, molecular domain hydrophobic interaction force, and molecular domain binding energy features; Step S24: extract specific molecular function features based on the biological reagent molecular function domain data to obtain the biological reagent specific molecular function domain features, including the biological reagent enzyme activity level, biological reagent binding affinity and biological reagent material transport and degradation ability characteristics.

9. The method for predicting the function of a biological agent according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: constructing a biological reagent function prediction network framework by connecting a convolutional neural network CNN, a recurrent neural network RNN, and a Transformer neural network; Step S32: inputting the biological reagent protein sequence alignment features, the biological reagent molecular domain features and the biological reagent specific molecular functional domain features into the biological reagent function prediction network framework to train and construct a function prediction model, while considering the activity, structural stability and specific function prediction targets corresponding to the biological reagent to generate a biological reagent function prediction model; Step S33: obtaining molecular features of the biological agent to be predicted, including protein sequence alignment features, molecular structural domain features, and molecular functional domain features corresponding to the biological agent to be predicted; Step S34: inputting the molecular features of the biological agent to be predicted into the biological agent function prediction model to perform target function prediction to obtain a target biological agent function prediction result.

10. A biological reagent function prediction system, characterized in that: Used to perform the biological agent function prediction method as claimed in claim 1, the biological agent function prediction system comprises: A biological reagent molecule information collection module is used to collect biological reagent molecule information from a biological reagent public database to obtain a biological reagent biological reagent molecule information set, wherein the biological reagent biological reagent molecule information set includes biological reagent protein sequence data, biological reagent molecule structure domain data, and biological reagent molecule function domain data; The biological reagent molecular feature analysis module is used to perform sequence comparison analysis on the biological reagent protein sequence data to obtain the biological reagent protein sequence comparison features; perform protein structure feature prediction analysis on the biological reagent molecular domain data to generate the biological reagent molecular domain features; extract specific molecular function features based on the biological reagent molecular function domain data to obtain the specific molecular function domain features of the biological reagent; The biological reagent target function prediction module is used to construct a biological reagent function prediction network framework, and input the biological reagent protein sequence alignment features, biological reagent molecular domain features and biological reagent specific molecular function domain features into the biological reagent function prediction network framework for function prediction model training and construction to generate a biological reagent function prediction model; obtain the molecular features of the biological reagent to be predicted, and input the molecular features of the biological reagent to be predicted into the biological reagent function prediction model for target function prediction to obtain the target biological reagent function prediction result; The biological reagent function verification and optimization module is used to perform function verification and optimization on the corresponding biological reagent to be predicted according to the function prediction result of the target biological reagent, so as to generate the function optimization result of the target biological reagent.

Images