A bioinformatics data processing platform for rodents

Abstract

The present invention discloses a biological information data processing platform for rodents, relating to the technical field of information data processing systems. The platform includes: a data acquisition and integration module, a gene analysis module; a protein analysis module; a metabolomics analysis module; an association mining module for mining the associations between genes, proteins, and metabolomics with the species of rodent based on the results of gene analysis, protein analysis, and metabolomics analysis, and visually displaying the associations; a user interaction module for providing a user-friendly user interface through which users can set permission management functions, upload data, select analysis functions, and view analysis results; and a database module for storing data received and generated by the platform. The platform is capable of integrating various types of rodent biological information, processing rodent biological information efficiently and analyzing it accurately.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biological information data processing systems, and particularly relates to a biological information data processing platform for rodents. BACKGROUND

[0002] In biomedical research, rodents (such as mice, rats, etc.) are widely used in various experiments due to their physiological characteristics similar to humans. By studying the biological information of rodents, such as genes and proteins, the pathogenesis of diseases and the action targets of drugs can be understood in depth. However, there are currently the following problems in the processing of rodent bioinformatics data: Data integration difficulty: different research institutions produce rodent biological data in various formats and store them in scattered places, making it difficult to effectively integrate and analyze. Single analysis tool: existing bioinformatics analysis tools often focus on the analysis of a specific type of data, such as gene sequence analysis or protein structure prediction, lacking a comprehensive analysis platform for all aspects of rodent biological information. Lack of specificity: there is no data processing process and algorithm optimized for the characteristics of rodents, resulting in low accuracy and efficiency of the analysis results. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a biological information data processing platform for rodents that can integrate multiple types of rodent biological information and is efficient and accurate in analysis.

[0004] To solve the above technical problems, the technical solution adopted by the present application is: a biological information data processing platform for rodents, comprising:

[0005] A data acquisition and integration module for acquiring rodent biological information data from multiple data sources, including gene sequence data, protein structure data, and metabolomics data; by formulating a unified data format standard, different formats of data are converted and integrated and stored in the database module;

[0006] A gene analysis module for implementing gene sequence alignment using a gene sequence alignment algorithm and analyzing the expression pattern of genes using a machine learning algorithm to predict the role of genes under different physiological conditions;

[0007] A protein analysis module for predicting the three-dimensional structure of a protein based on its amino acid sequence using a deep learning-based protein structure prediction model; and constructing a rodent biological protein-protein interaction network using a protein-protein interaction network analysis algorithm to obtain the functional relationship between proteins;

[0008] Metabolomics analysis module: using mass spectrometry data analysis algorithm, combined with database matching, identify the types of metabolites; through metabolic pathway enrichment analysis, determine the metabolic pathways related to specific physiological or pathological state;

[0009] Correlation mining module: according to the results of gene analysis, protein analysis and metabolomics analysis, the correlation between genes, proteins and metabolomics and the species of rodents is mined, and the correlation is visualized;

[0010] User interaction module: used to provide a friendly user interface, through which the user sets the permission management function, uploads data, selects analysis functions and views analysis results;

[0011] Database module, for storing data received and generated by the platform.

[0012] The beneficial effects produced by the above technical solutions are: the platform integrates various types of rodent bioinformatics data, avoids switching between different data sources and analysis tools, greatly improves the efficiency of data processing. The algorithm and model optimized for the characteristics of rodents can more accurately analyze and interpret rodent bioinformatics data, providing a reliable basis for biomedical research. BRIEF DESCRIPTION OF DRAWINGS

[0013] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0014] Figure 1 is the principle diagram of the platform described in the embodiments of the present application;

[0015] Figure 2 is the implementation flowchart of the gene analysis module in the platform described in the embodiments of the present application;

[0016] Figure 3 is the implementation flowchart of the protein analysis module in the platform described in the embodiments of the present application;

[0017] Figure 4 is the implementation flowchart of the metabolomics analysis module in the platform described in the embodiments of the present application;

[0018] Figure 5 is the implementation flowchart of the correlation mining module in the platform described in the embodiments of the present application. DETAILED DESCRIPTION

[0019] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0020] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other manners different from those described herein, and a person of ordinary skill in the art can make similar generalizations without departing from the spirit and scope of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0021] As shown in Figure 1 The embodiments of the present application disclose a rodent-oriented biological information data processing platform, comprising:

[0022] The data acquisition and integration module 101 is configured to acquire rodent biological information data from a plurality of data sources, including gene sequence data, protein structure data and metabolomics data; and convert and integrate data in different formats into a database module by formulating a unified data format standard.

[0023] It should be noted that the "unified data format standard" in the present application is prior art, which refers to a standardized file format, field definition, coding rule and metadata specification formulated for different types of data (such as gene sequence, protein structure, metabolomics data, etc.) in order to realize data compatibility and interoperability across platforms and data sources, and the specific implementation steps are not described herein.

[0024] In addition, the method of converting and integrating data in different formats in the present application is prior art, and the implementation method generally includes the following steps: data parsing and cleaning, format conversion and standardized mapping, data integration and deduplication, and quality control and storage, etc. The heterogeneous data is converted into a unified format that can be directly used by each module of the platform (such as gene analysis, association mining), and the specific implementation steps are not described herein.

[0025] The gene analysis module 102 is configured to realize gene sequence alignment by using a gene sequence alignment algorithm, and analyze the expression pattern of genes by using a machine learning algorithm to predict the role of genes in different physiological states.

[0026] It should be noted that the method of the step "and analyzing the expression pattern of the gene by using a machine learning algorithm" in the present application is a prior art, and its implementation method generally includes the following steps: data preparation and preprocessing, feature selection, machine learning model selection and training, model optimization, and analysis of the expression pattern of the gene by using the optimized model, and the specific implementation steps are not repeated here.

[0027] The protein analysis module 103: by using a deep learning-based protein structure prediction model, the three-dimensional structure of the protein is predicted according to the amino acid sequence of the protein; by using a protein interaction network analysis algorithm, a rodent protein interaction network is constructed, and the functional relationship between proteins is obtained;

[0028] The metabolomics analysis module 104: by using a mass spectrometry data analysis algorithm combined with database matching, the types of metabolites are identified; by using metabolic pathway enrichment analysis, the metabolic pathways related to specific physiological or pathological states are determined;

[0029] The correlation mining module 105: according to the results of gene analysis, protein analysis and metabolomics analysis, the correlation between genes, proteins and metabolomics and the rodent is mined, and the correlation is visualized and displayed;

[0030] The user interaction module 106: used for providing a friendly user interface, through which the user sets the permission management function, uploads data, selects analysis functions and views analysis results;

[0031] The database module 107 is used for storing the data received and generated by the platform.

[0032] Further, as shown in Figure 2 The processing method of the gene analysis module 102 includes the following steps:

[0033] 1) Gene sequence alignment:

[0034] The gene sequence data to be compared is obtained from the database module, including the query sequence (usually a new sequence input by the user) and the reference sequence (known rodent gene sequence stored in the platform database), and an index is established for the reference sequence to quickly find matching fragments. For large-scale gene sequence databases, efficient index structures such as Suffix Tree or Burrows-Wheeler Transform (BWT) can be used to speed up the search process; the query sequence is divided into seed fragments, and the length and position of the seed are adjusted according to the sequence characteristics of rodent genes, and the matching seed is found in the index, and the found seed matching is bidirectionally expanded and aligned, the scoring matrix for rodent genes is used to calculate the alignment score, and the threshold is set to determine whether it is an effective alignment, and the greedy algorithm or dynamic programming algorithm is used to merge and optimize the matching fragments to obtain the final sequence alignment result; the alignment result is usually presented in the form of a table or text, including the information of the query sequence and the reference sequence, the alignment position, the alignment score, the matching percentage, etc.

[0035] 2) Gene expression pattern analysis:

[0036] Feature selection and extraction: features are extracted from gene expression data for training and prediction of machine learning models, which can include gene expression levels, expression fold changes, and differentially expressed information between different samples, etc.

[0037] Construction and training of machine learning model: using support vector machine SVM as the machine learning model, the labeled feature data set (known gene expression characteristics in different physiological states and corresponding physiological state labels) is divided into training set and test set, and the training set and test set are used to train the machine learning model, and by optimizing the kernel function (such as linear kernel, radial basis kernel, etc.) and adjusting the hyperparameters (such as penalty parameter C and kernel parameter γ), the model can achieve optimal performance on the training set;

[0038] Prediction and evaluation: input the gene expression characteristics to be predicted into the trained machine learning model to get the prediction results of the gene in different physiological states, usually in the form of probability or category. The performance of the model is evaluated, and the accuracy, recall rate, F1 value, etc. are calculated using the test set to evaluate the reliability of the model.

[0039] Further, as shown in Figure 3 the processing method of the protein analysis module includes the following steps:

[0040] 1) Protein structure prediction

[0041] Protein structure model construction: Use AlphaFold architecture as the basic architecture of protein structure prediction, including input layer, middle layer and output layer; the input layer uses one-hot encoding to represent each amino acid sequence of the protein as a high-dimensional vector, for example, for 20 common amino acids, an amino acid can be represented as a vector of length 20, where the position corresponding to the amino acid is 1 and the rest is 0; the middle layer includes multiple Transformer modules to learn the long-range dependencies in the amino acid sequence, the Transformer module contains multi-head self-attention mechanism, which allows the model to focus on information at different positions simultaneously, enhancing the ability to capture sequence information; the output layer is used to output the predicted protein structure information, usually atomic coordinates or some feature representations of the structure such as distance matrix, angle, etc., which are finally converted into the three-dimensional structure of the protein through geometric calculations.

[0042] Training data collection and preprocessing: Collect rodent protein data with known structures, clean the amino acid sequences to remove incomplete or incorrect sequence information; convert the protein structure into the input representation required by the model, increase the diversity of the data by rotating, translating and adding noise operations, and improve the generalization ability of the model;

[0043] Model training: divide the preprocessed data into training set, validation set and test set, train the model using the training set, adjust the model hyperparameters according to the performance of the validation set, and evaluate the performance of the final model using the test set;

[0044] Structure prediction: for a new rodent protein amino acid sequence, encode it into an input vector and input it into the trained model, the model outputs the predicted protein structure features, and then converts these features into a three-dimensional protein structure through a structure reconstruction algorithm;

[0045] 2) Protein-protein interaction network analysis

[0046] Network construction: represent proteins as nodes and interactions as edges to construct an undirected graph where is the set of protein nodes, is the edge set, and different weights are assigned to the edges according to the strength of the interaction;

[0047] Network analysis algorithm: use Markov clustering algorithm MCL to divide the nodes in the undirected graph into different clusters, perform expansion and inflation operations on the adjacency matrix to enhance the connection patterns between nodes, and finally converge to a stable state through iterative updating of the matrix to obtain different clusters;

[0048] Function relationship mining: according to the clustering results, combining the functional annotation information of the proteins, the hypergeometric distribution is used for functional enrichment analysis of each cluster, and the functional relationship between the proteins is obtained according to the analysis results; through the above steps and algorithms, the protein analysis module can realize the prediction of the protein structure of rodents and the analysis of the functional relationship between the proteins, and provide important support for in-depth understanding of the function of the proteins and biological processes

[0049] Further, as shown in Figure 4 the implementation method of the metabolomics analysis module includes the following steps:

[0050] 1) Metabolite species identification

[0051] Mass spectrometry data acquisition:

[0052] Through liquid chromatography-mass spectrometry (LC-MS), the metabolomics mass spectrometry data of the rodent biological sample is obtained, and the metabolomics mass spectrometry data includes the mass-to-charge ratio and the corresponding signal intensity of the metabolite;

[0053] Data preprocessing:

[0054] The centroid method is used to identify the peaks of the metabolite ion peaks from the original mass spectrometry data, and the local maximum value of the signal intensity is found as the center of the peak, and the formula is:

[0055] ;

[0056] Wherein, is the mass-to-charge ratio of the i-th metabolite, is the signal intensity of the i-th metabolite, and is the scanning range where the peak is located;

[0057] The dynamic time warping (DTW) algorithm is used to align the obtained peaks, the Savitzky-Golay filtering algorithm is used to remove noise and improve data quality, and the total ion flow normalization algorithm is used to normalize the signal intensity, so that the data of different samples are comparable;

[0058] Database matching:

[0059] The HMDB (Human Metabolome Database) metabolite database is used as a comparison database, and the mass-to-charge ratio of the metabolite in the preprocessed mass spectrometry data is matched with the accurate mass of the known metabolite in the database, and a certain mass error range (usually ppm level) is set, and the formula is:

[0060] ;

[0061] Wherein, is the observed mass-to-charge ratio, is the expected mass-to-charge ratio of the metabolite in the database;

[0062] For metabolites that can generate fragment ions, the observed fragment ion pattern is matched with the fragment pattern in the database using a similarity measure, whose formula is:

[0063] ;

[0064] wherein, and is the signal intensity of the observed fragment ion of the i-th metabolite and the fragment ion of the metabolite in the database;

[0065] Combined with the accurate mass matching and the fragment ion matching, a comprehensive score is calculated for each metabolite to determine the final matching result;

[0066] 2) Metabolic pathway enrichment analysis:

[0067] Using the KEGG (Kyoto Encyclopedia of Genes and Genomes) metabolic pathway database as the comparative database, which contains the information of metabolic pathways and the role of metabolites in the pathway, the identified metabolite list is used as input, and the relative abundance information of each metabolite in different samples is obtained;

[0068] Let the total metabolite set be , wherein the metabolites belonging to a certain pathway are , and there are in the identified metabolite set, and belong to this pathway, and the enrichment significance is calculated using the hypergeometric distribution formula:

[0069] ;

[0070] represents that in a given total, randomly extracting metabolites, of which or more belong to the pathway metabolites, the smaller the value, the more significant the enrichment of the pathway;

[0071] The relative abundance change of metabolites under different physiological or pathological conditions is calculated, and the formula is:

[0072] ;

[0073] Combined with the value and the value, the significantly changed and enriched metabolic pathways are screened out;

[0074] According to the enriched metabolic pathways, combined with biological knowledge, the functional significance of these pathways in a specific physiological or pathological state is explained, for example, which metabolic pathways are up-regulated or down-regulated, and the physiological functions that may be affected.

[0075] Further, as shown in Figure 5 The implementation method of the association relationship mining module includes the following steps:

[0076] Data integration: integrate data from the gene analysis module, protein analysis module and metabolomics analysis module into a data set;

[0077] Correlation analysis: calculate the Pearson and Spearman correlation coefficients between different types of data on the above data set to determine their linear and monotonic relationships; perform partial least squares regression to find the relationship between the predictor variable and the response variable; according to the correlation and regression analysis results, use network analysis methods to construct a comprehensive network, and perform module analysis;

[0078] Visualization: select visualization tools and graph types, and draw networks, heat maps, scatter plots or bar charts according to the analysis results.

[0079] Further, the data integration includes the following steps:

[0080] Gene analysis data: obtain gene sequence alignment results, gene expression pattern analysis results and predicted effects of genes in different physiological states from the gene analysis module, which may include gene sequence similarity scores, gene expression levels, gene function prediction results, etc.

[0081] Protein analysis data: includes three-dimensional structure information of proteins, protein interaction network information, functional relationship data between proteins, such as protein structure features obtained from deep learning-based protein structure prediction models, and protein interaction strength and network topology information determined by protein interaction network analysis algorithms.

[0082] Metabolomics analysis data: obtain metabolite species identification results, metabolic pathway enrichment analysis results and metabolite abundance data, etc., which are used to understand the state of metabolites and changes in metabolic pathways.

[0083] Data integration: integrate data from the gene analysis module, protein analysis module and metabolomics analysis module into a data set.

[0084] Further, the correlation analysis includes the following steps:

[0085] Correlation analysis:

[0086] Pearson correlation coefficient is used to measure the linear correlation between two variables, the formula is:

[0087] ;

[0088] where, and are the observed values of two randomly selected variables (e.g., gene expression levels and metabolite abundance) in the dataset, and are the mean values of the two variables, is the number of observations;

[0089] Spearman rank correlation coefficient is used to measure the monotonic relationship between two variables, independent of the distribution of data, suitable for non-linear relationship, the formula is:

[0090] ;

[0091] where is the difference in rank of the two selected variables in the above data set, is the sample size;

[0092] Pearson correlation coefficient and Spearman rank correlation coefficient are used to analyze the relationship between gene expression levels and metabolite abundance, or the relationship between protein interaction strength and metabolic pathway activity, to determine whether there is a linear or monotonic association between the data;

[0093] Partial Least Squares Regression (PLS):

[0094] Through Partial Least Squares Regression (PLS) algorithm, the potential structural relationship between predictor variables and response variables is constructed:

[0095] ;

[0096] ;

[0097] where, is the predictor variable matrix (such as gene expression and protein data), is the response variable matrix (such as metabolomics data or physiological phenotype), and are latent variables, and are loading matrices, and are residual matrices; PLS is used to find the relationship between predictor variables and response variables, such as finding which genes and proteins are most relevant to the changes in metabolites.

[0098] Network analysis:

[0099] Genes, proteins, and metabolites are used as nodes, and a weighted network is constructed based on the correlations obtained from the above analysis. If the correlation exceeds a certain threshold, an edge is added between the corresponding nodes, and the weight of the edge is the absolute value of the correlation or other association strength indicators.

[0100] Using the Louvain detection algorithm, the network is divided into different modules. The formula is:

[0101] ;

[0102] in, The number of edges, is a node The degree, is a node Degrees within the module, and is a node and nodes The module tag, is an indicator function, when Module analysis is used to discover sets of closely related genes, proteins, and metabolites that may functionally collaborate with each other.

[0103] Furthermore, the visual presentation includes the following steps:

[0104] Network visualization: Use Cytoscape to visualize the constructed comprehensive network. Different shapes and colors are used for nodes in the network to represent different types of entities (genes, proteins, metabolites). The width and color of the edges represent the strength of the association or the size of the correlation.

[0105] Heatmap visualization: Use Python's Matplotlib library to draw heatmaps to show the correlations between genes, proteins, and metabolites. Display the correlation matrix or principal component analysis results in the form of a heatmap, with the color depth indicating the strength of the correlation or the contribution of the principal component. Rows and columns can represent different genes, proteins, and metabolites respectively.

[0106] Scatter plots and histograms: For pairwise associations, including the relationship between gene expression levels and metabolite abundance, scatter plots are used for display, where the X-axis represents gene expression levels, the Y-axis represents metabolite abundance, and each dot represents a sample; for principal component analysis results or enrichment analysis results, histograms are used to display the proportion of explained variance of different principal components or the enrichment significance of different metabolic pathways.

[0107] Through the above steps and analysis methods, the correlation relationship mining module can effectively mine the correlation relationships between genes, proteins and metabolomics data, and display these relationships in an intuitive manner, providing a comprehensive perspective for studying the physiological and pathological states of rodents.

[0108] In summary, the platform can integrate various types of rodent biological information, and the processing of rodent biological information is efficient and accurate.

Claims

1. A biological information data processing system for rodents, characterized by include: Data acquisition and integration module: used to collect rodent bioinformatics data from multiple data sources, including gene sequence data, protein structure data, and metabolomics data; By formulating a unified data format standard, data in different formats can be converted and integrated and stored in the database module; Gene analysis module: used to implement gene sequence alignment using a gene sequence alignment algorithm, and analyze gene expression patterns using a machine learning algorithm to predict the effects of genes under different physiological states; Protein Analysis Module: This module uses a deep learning-based protein structure prediction model to predict the three-dimensional structure of a protein based on its amino acid sequence. It also uses a protein interaction network analysis algorithm to construct a rodent protein interaction network and obtain functional relationships between proteins. Metabolomics analysis module: uses mass spectrometry data analysis algorithms combined with database matching to identify metabolite types; Identify metabolic pathways associated with specific physiological or pathological states through metabolic pathway enrichment analysis; Association mining module: Based on the results of gene analysis, protein analysis and metabolomics analysis, it mines the association between genes, proteins and metabolomics and the rodent species, and visualizes the association; User interaction module: used to provide a friendly user interface through which users can set permission management functions, upload data, select analysis functions and view analysis results; Database module, used to store data received and generated by the platform; The implementation method of the metabolomics analysis module comprises the following steps: 1) Identification of metabolite types: Mass spectrometry data acquisition: Liquid chromatography-mass spectrometry (LC-MS) is used to obtain metabolomics mass spectrometry data of rodent biological samples. Metabolomics mass spectrometry data include the mass-to-charge ratio of metabolites and the corresponding signal intensity. Data preprocessing: The centroid method is used to identify the peaks of metabolite ions from the raw mass spectrum data. The centroid method finds the local maximum of the signal intensity as the center of the peak. The formula is: ； in, is the mass-to-charge ratio of the ith metabolite, is the signal intensity of the ith metabolite, and is the scan range where the peak is located; The dynamic time warping (DTW) algorithm was used to align the peaks, the Savitzky-Golay filtering algorithm was used to remove noise, and the total ion current normalization algorithm was used to normalize the signal intensity to make the data of different samples comparable. Database Matching: Using the HMDB metabolite database as a comparison database, the metabolite mass-to-charge ratio in the preprocessed mass spectrometry data was matched with the accurate mass of the known metabolites in the database, and a certain mass error range was set. The formula is: ； in, is the observed mass-to-charge ratio, is the expected mass-to-charge ratio of the metabolite in the database; For metabolites that can produce fragment ions, the observed fragment ion patterns are matched with the fragment patterns in the database using a similarity metric, which is formulated as: ； in, and is the signal intensity of the observed fragment ion of the ith metabolite and the fragment ion of the metabolite in the database; Combining accurate mass matching and fragment ion matching, a comprehensive score is calculated for each metabolite to determine the final matching result; 2) Metabolic pathway enrichment analysis: The KEGG metabolic pathway database was used as a comparison database, and the identified metabolite list was used as input to obtain the relative abundance information of each metabolite in different samples; Assume that the total metabolite set is , among which the metabolites belonging to a certain pathway are There are a total of , belonging to this channel are The hypergeometric distribution formula was used to calculate the enrichment significance: ； In a given population, randomly select metabolites, including The probability of more than one metabolite belonging to the pathway. The smaller the value, the more significant the enrichment of the pathway. The relative abundance changes of metabolites under different physiological or pathological conditions were calculated using the following formula: ； Combine Value and Values ​​were used to screen out significantly changed and enriched metabolic pathways; Based on the enriched metabolic pathways and combined with biological knowledge, the functional significance of these metabolic pathways in specific physiological or pathological conditions can be obtained; The processing method of the gene analysis module includes the following steps: 1) Gene sequence alignment: Obtaining gene sequence data to be aligned from a database module, including a query sequence and a reference sequence, and establishing an index for the reference sequence; dividing the query sequence into seed segments, and adjusting the length and position of the seeds according to the sequence characteristics of rodent genes; searching for matching seeds in the index; performing a bidirectional extended alignment on the found seed matches; calculating an alignment score using a scoring matrix for rodent genes; determining whether the alignment is valid based on a set threshold; and merging and optimizing the matching segments using a greedy algorithm or a dynamic programming algorithm to obtain a final sequence alignment result; 2) Gene expression pattern analysis: Feature selection and extraction: Extracting features from gene expression data for training and prediction of machine learning models. These features include gene expression levels, expression fold changes, and differences between different samples. Machine learning model construction and training: Using support vector machines (SVMs) as the machine learning model, the labeled feature dataset is divided into training and test sets. The machine learning model is trained using the training and test sets. By optimizing the model's kernel function and adjusting the model's hyperparameters, the model achieves optimal performance on the training set. Prediction and evaluation: The gene expression features to be predicted are input into the trained machine learning model to obtain the predicted results of the gene's role in different physiological states.

2. A rodent-oriented biological information data processing system according to claim 1, characterized in that: The processing method of the protein analysis module comprises the following steps: 1) Protein structure prediction: Protein structure model construction: The AlphaFold architecture is used as the basic architecture for protein structure prediction, consisting of an input layer, an intermediate layer, and an output layer. The input layer uses one-hot encoding to represent each amino acid sequence of a protein as a high-dimensional vector. The intermediate layer includes multiple Transformer modules for learning long-range dependencies in amino acid sequences. The output layer is used to output predicted protein structure information. Training data collection and preprocessing: Collect rodent protein data with known structures, clean the amino acid sequences, and remove incomplete or erroneous sequence information. Convert the protein structures into the input representations required by the model, and increase data diversity through rotation, translation, and noise addition to improve the model's generalization capabilities. Model training: Divide the preprocessed data into training, validation, and test sets. Use the training set to train the model, adjust the model hyperparameters based on the performance of the validation set, and use the test set to evaluate the performance of the final model. Structure prediction: For a new rodent protein amino acid sequence, it is encoded as an input vector and fed into a trained model. The model outputs predicted protein structural features, which are then converted into a three-dimensional protein structure using a structure reconstruction algorithm. 2) Protein interaction network analysis: Network construction: proteins are represented as nodes and interactions as edges to construct an undirected graph ,in is the set of protein nodes, is a set of edges, with different weights assigned to the edges according to the strength of their interactions; Network analysis algorithm: Use the Markov clustering algorithm (MCL) to divide the nodes in the undirected graph into different clusters. Expand and inflate the adjacency matrix to enhance the connection pattern between nodes. By iteratively updating the matrix, it eventually converges to a stable state and obtains different clusters. Functional relationship mining: Based on the clustering results and combined with the functional annotation information of proteins, functional enrichment analysis is performed on each cluster using hypergeometric distribution, and the functional relationships between proteins are obtained based on the analysis results.

3. The rodent-oriented biological information data processing system according to claim 1, characterized in that: The implementation method of the association relationship mining module includes the following steps: Data integration: Integrate data from the gene analysis module, protein analysis module, and metabolomics analysis module into one dataset; Correlation analysis: Pearson and Spearman correlation coefficients were calculated between different types of data in the above datasets to determine the linear and monotonic relationships between them; partial least squares regression was performed to find the relationship between the predictor variables and the response variables; based on the correlation and regression analysis results, a comprehensive network was constructed using network analysis methods, and modular analysis was performed; Visualization: Select a visualization tool and graph type to draw a network, heat map, scatter plot, or bar chart based on the analysis results.

4. A rodent-oriented biological information data processing system according to claim 3, characterized in that: The data integration includes the following steps: Gene analysis data: Obtain gene sequence alignment results, gene expression pattern analysis results, and predicted effects of genes under different physiological states from the gene analysis module; Protein analysis data: including protein three-dimensional structure information, protein interaction network information, and functional relationship data between proteins; Metabolomics analysis data: obtain metabolite species identification results, metabolic pathway enrichment analysis results, and metabolite abundance data; Data integration: The data from the gene analysis module, protein analysis module, and metabolomics analysis module are integrated into one dataset.

5. The rodent-oriented biological information data processing system according to claim 3, characterized in that: The association analysis includes the following steps: Correlation analysis: The Pearson correlation coefficient is used to measure the linear correlation between two variables. The formula is: ； in, and are the random observations of two variables in the data set, and is the average of the above two variables, is the number of observations; The Spearman rank correlation coefficient is used to measure the monotonic relationship between two variables. It does not depend on the distribution of the data and is applicable to nonlinear relationships. The formula is: ； in is the rank difference between the two variables selected in the above dataset, is the sample size; Pearson correlation coefficient and Spearman rank correlation coefficient were used to analyze the relationship between gene expression level and metabolite abundance, or the relationship between protein interaction strength and metabolic pathway activity to determine whether there was a linear or monotonic association between the data; Partial Least Squares Regression PLS: The potential structural relationship between the predictor variables and the response variables is constructed through the partial least squares regression PLS algorithm: ； ； in, is the predictor variable matrix, is the response variable matrix, and is a latent variable, and is the loading matrix, and is the residual matrix, and the relationship between the predictor and response variables is found through the above formula; Network Analysis: Genes, proteins, and metabolites are used as nodes, and a weighted network is constructed based on the correlations obtained from the above analysis. If the correlation exceeds a certain threshold, an edge is added between the corresponding nodes, and the weight of the edge is the absolute value of the correlation or other association strength indicators. Using the Louvain detection algorithm, the network is divided into different modules. The formula is: ； in, The number of edges, is a node The degree, is a node Degrees within the module, and is a node and nodes The module tag, is an indicator function, when 1 if yes, 0 otherwise.

6. A rodent-oriented biological information data processing system according to claim 3, characterized in that: The visual presentation includes the following steps: Network visualization: Use Cytoscape to visualize the constructed comprehensive network. Different shapes and colors are used to represent different types of entities for nodes in the network, and the width and color of the edges represent the strength of association or the size of the correlation. Heatmap visualization: Use Python's Matplotlib library to draw heatmaps to show the correlations between genes, proteins, and metabolites. Display the correlation matrix or principal component analysis results in the form of a heatmap, with the color depth indicating the strength of the correlation or the contribution of the principal component. Rows and columns can represent different genes, proteins, and metabolites respectively. Scatter plots and histograms: For pairwise associations, including the relationship between gene expression levels and metabolite abundance, scatter plots are used for display, where the X-axis represents gene expression levels, the Y-axis represents metabolite abundance, and each dot represents a sample; for principal component analysis results or enrichment analysis results, histograms are used to display the proportion of explained variance of different principal components or the enrichment significance of different metabolic pathways.

Images