Gene interaction network analysis system and method based on symbology dynamics

Abstract

The invention is suitable for the technical field of bioinformatics, and provides a gene interaction network analysis system and method based on symbology dynamics. The method comprises the steps of original gene data preprocessing, dynamic gene interaction network construction, network dynamic state symbol coding, symbol sequence feature analysis, analysis strategy dynamic adjustment, key information identification and result output. The system comprises corresponding function modules. According to the method, a symbolic dynamics theory and gene interaction network analysis are fused, a gene network dynamic process is converted into a symbol sequence through an exclusive symbol coding algorithm, network dynamic characteristics are extracted by means of methods such as symbol entropy calculation and symbol pattern recognition, and meanwhile, a dynamic adjustment analysis strategy is introduced; according to the method, a dynamic self-adaptive analysis system is constructed, dynamic changes of gene expression and regulation relationships are accurately described, the recognition accuracy of key interaction relationships, key nodes and key paths of a complex network is improved, and the method has important theoretical value and practical application prospects.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology, and in particular relates to a gene interaction network analysis system and method based on symbolic dynamics. Background Technology

[0002] Gene interaction networks are complex network structures formed by the regulation and expression associations between genes within an organism. They reflect the synergistic mechanisms of genes in life activities and are one of the core contents of bioinformatics research. By analyzing these networks, we can reveal gene functions, elucidate the molecular mechanisms of disease development, and provide important theoretical basis for disease diagnosis and treatment.

[0003] With the rapid development of high-throughput sequencing technology, the gene expression data and regulatory relationship data generated by biological experiments are growing exponentially, driving the continuous advancement of gene interaction network analysis technology. Existing gene interaction network analysis methods mainly include statistical correlation-based methods (such as Pearson correlation coefficient and mutual information method, where mutual information method calculates the mutual information value between gene expression data to determine the strength of correlation between genes, thereby constructing a gene interaction network and screening key interaction relationships), machine learning-based prediction methods (such as support vector machines and neural networks), and graph theory-based network topology analysis methods (such as degree centrality and betweenness centrality analysis). Graph theory-based network topology analysis methods construct undirected or directed graphs of gene interactions, abstracting genes as nodes and gene interactions as edges. By calculating topological indices such as degree centrality, betweenness centrality, and proximity centrality of nodes, key nodes in the network are identified based on the index values, and potential critical paths are identified using shortest path algorithms. These methods have achieved certain results in the structural analysis of simple gene interaction networks. However, as the complexity of gene interaction networks increases, the need to characterize the dynamic changes in these networks becomes increasingly urgent, and existing methods are gradually revealing their limitations, specifically:

[0004] (1) Insufficient dynamic characterization ability: Existing graph theory-based topology analysis methods only focus on the static structure of the network and cannot reflect the dynamic process of gene interaction changing with time and environment. Gene expression and regulation have significant dynamics, and static analysis leads to inaccurate characterization of the network's essential mechanism.

[0005] (2) Poor adaptability of complex networks: For complex gene interaction networks containing thousands or even tens of thousands of nodes and edges, existing statistical correlation-based analysis methods and machine learning-based prediction methods are easily affected by noisy data, making it difficult to distinguish between real interactions and false associations, resulting in low analysis accuracy.

[0006] (3) Low efficiency in identifying key information: Existing methods rely on a single topological index or correlation threshold to screen key nodes and paths, without considering the dynamic pattern characteristics of gene interactions, resulting in a high rate of missed detection and false detection of key interaction relationships, and making it impossible to accurately locate genes and paths that play a core role in network function.

[0007] (4) Lack of dynamic adjustment mechanism: The analysis strategy of existing methods is fixed and cannot adaptively adjust the analysis parameters and logic according to the dynamic changes of gene interaction networks, making it difficult to adapt to the analysis needs of gene interaction networks of different types and complexities.

[0008] Symbolic dynamics is a discipline that transforms the dynamic behavior of complex systems into symbolic sequences for analysis. Its core idea is to discretize continuous dynamic processes into analyzable symbolic sequences through quantification, hierarchical classification, and symbolic mapping of system states. The entropy, mode frequency, and other characteristics of these symbolic sequences are then used to reveal the system's intrinsic laws. This theory has been successfully applied to the analysis of complex systems in physics, chemistry, and engineering, providing new technical approaches for solving the dynamic analysis of gene-gene interaction networks.

[0009] To this end, this invention proposes a gene interaction network analysis system and method based on symbolic dynamics. Summary of the Invention

[0010] The purpose of this invention is to provide a gene interaction network analysis system and method based on symbolic dynamics, aiming to solve the problems mentioned in the background art.

[0011] The objective of this invention is achieved through the following technical solution:

[0012] A gene interaction network analysis method based on symbolic dynamics includes the following steps:

[0013] Step 1: Preprocessing raw genetic data;

[0014] Collect raw gene data and perform data cleaning, data standardization and noise removal, and output preprocessed dynamic gene data;

[0015] Step 2: Construction of dynamic gene interaction network;

[0016] Receive dynamic gene data, construct a dynamic gene interaction network based on time series, the dynamic gene interaction network uses genes as nodes, the regulatory relationships between genes as edges, and the weight of the edges as the regulatory strength, and output the dynamic gene interaction network.

[0017] Step 3: Symbolic encoding of network dynamic states;

[0018] Receive a dynamic gene interaction network, map the dynamic state of the dynamic gene interaction network into a symbol sequence using a preset symbol encoding algorithm, and output a symbol sequence set;

[0019] Step 4: Symbol sequence feature analysis;

[0020] Receive a set of symbol sequences, use symbol entropy calculation and symbol pattern recognition algorithms to extract feature parameters of the symbol sequences, and output a set of symbol sequence features.

[0021] Step 5: Analyze the dynamic adjustment of the strategy;

[0022] It receives a symbol sequence feature set, dynamically optimizes the symbol encoding parameters and analysis parameters according to preset adjustment rules, feeds back the optimized parameters iteratively, and finally outputs the optimized symbol sequence feature set;

[0023] Step 6: Identification of key information;

[0024] Receive the optimized symbol sequence feature set, combine it with network topology features, identify key interaction relationships, key nodes and key paths, and output key information identification results;

[0025] Step 7: Output the results;

[0026] Receive key information identification results and output them in the form of visual charts and data reports.

[0027] Furthermore, in step 1, the original gene data includes gene expression time series data and known regulatory relationship data between genes; data cleaning includes removing gene data with zero expression or missing values ​​exceeding a preset threshold, and filling the remaining missing values ​​using linear interpolation to obtain cleaned expression time series data; data standardization uses the Z-score standardization method to standardize the cleaned expression time series data to obtain a standardized data matrix; noise removal uses wavelet threshold denoising to process the standardized data matrix to obtain a denoised gene dynamic data matrix.

[0028] Furthermore, in step 2, the construction of the dynamic gene interaction network specifically includes: calculating the regulatory intensity between genes using time-delayed mutual information based on the denoised gene dynamic data matrix; constructing a directed weighted dynamic gene interaction network with genes as nodes and gene pairs with regulatory intensity greater than or equal to a preset threshold as edges.

[0029] Furthermore, in step 3, the dynamic state includes node state and edge state;

[0030] The node state symbol encoding is specifically as follows: calculate the rate of change of gene expression, map the rate of change to the corresponding symbol according to the preset interval to which it belongs, and obtain the node state symbol sequence;

[0031] The edge state symbol encoding is specifically as follows: calculate the rate of change of regulatory intensity between genes, map the rate of change to the corresponding symbol according to the preset interval to which it belongs, and obtain the edge state symbol sequence;

[0032] Integrate all node state symbol sequences and edge state symbol sequences to form a complete network symbol sequence set.

[0033] Furthermore, in step 4, the feature parameters include symbol entropy and pattern weight;

[0034] Symbol entropy is calculated through the following steps: dividing the symbol sequence into multiple subsequences, counting the frequency of occurrence of each subsequence, and then applying the formula... The symbolic entropy is calculated; where The symbol entropy represents the sequence of symbols; p represents the index of the symbol pattern; M represents the total number of distinct subsequences. Representing a subsequence The probability of occurrence;

[0035] Pattern weights are calculated through the following steps: identify key patterns in the symbol sequence and count the number of times the key patterns appear, then multiply the number of times the key patterns appear by the corresponding symbol entropy to obtain the pattern weight.

[0036] Furthermore, in step 5, dynamic optimization specifically includes:

[0037] If the average symbol entropy of the node state symbol sequence is lower than the node entropy threshold, or the average symbol entropy of the edge state symbol sequence is lower than the edge entropy threshold, then increase the number of intervals for the corresponding symbol mapping and return to step 3 for re-encoding.

[0038] If the variance of the pattern weights of the node state symbol sequence exceeds the node variance threshold, or the variance of the pattern weights of the edge state symbol sequence exceeds the edge variance threshold, then adjust the subsequence length for the corresponding symbol entropy calculation or the window size for key pattern recognition, and return to step 4 for reanalysis.

[0039] Furthermore, in step 5, the specific process of feedback iteration is as follows: repeat steps 3 to 5 until the average change of symbol entropy and the variance of mode weights satisfy the preset stability condition, at which point the optimized symbol sequence feature set is output.

[0040] Furthermore, in step 6:

[0041] Identifying key interactions specifically includes: calculating the comprehensive score of edges, which is the weighted sum of the pattern weights and symbol entropies corresponding to the edge state symbol sequences; and selecting edges whose comprehensive scores are greater than or equal to a preset score threshold, which are the key interactions between genes.

[0042] Identifying key nodes specifically includes: calculating the comprehensive score of a node, which is a weighted sum of the pattern weight, symbol entropy, and degree centrality of the node's state symbol sequence; and selecting nodes whose comprehensive score is greater than or equal to a preset score threshold, which are the key nodes in the network.

[0043] Identifying critical paths specifically includes: defining a path as a sequence of directed edges from the starting critical node to the ending critical node; calculating the path weight, which is the ratio of the sum of the comprehensive scores of all edges on the path to the path length; and selecting paths whose path weights are greater than or equal to a preset path weight threshold as critical paths.

[0044] The gene interaction network analysis system based on symbolic dynamics is used to implement the gene interaction network analysis method based on symbolic dynamics mentioned above. It includes a data preprocessing module, a network construction module, a symbol encoding module, a symbol sequence feature analysis module, a dynamic adjustment module, a key information identification module, and a result output module.

[0045] The data preprocessing module is used to collect raw gene data and perform data cleaning, data standardization and noise removal, and output preprocessed gene dynamic data;

[0046] The network construction module is used to receive gene dynamic data, construct a dynamic gene interaction network based on time series, and output the dynamic gene interaction network with genes as nodes, regulatory relationships between genes as edges, and edge weights as regulatory strengths.

[0047] The symbol encoding module is used to receive the dynamic gene interaction network, map the dynamic state of the dynamic gene interaction network into symbol sequences through a preset symbol encoding algorithm, and output a symbol sequence set;

[0048] The symbol sequence feature analysis module is used to receive a set of symbol sequences, extract feature parameters of the symbol sequences using symbol entropy calculation and symbol pattern recognition algorithms, and output a set of symbol sequence features.

[0049] The dynamic adjustment module receives the symbol sequence feature set, dynamically optimizes the symbol encoding parameters and analysis parameters according to preset adjustment rules, feeds back the optimized parameters to the symbol encoding module and the symbol sequence feature analysis module, and finally outputs the optimized symbol sequence feature set.

[0050] The key information identification module is used to receive the optimized symbol sequence feature set, combine it with network topology features, identify key interaction relationships, key nodes and key paths, and output key information identification results;

[0051] The results output module is used to receive the key information identification results and output them in the form of visual charts and data reports.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] This invention combines symbolic dynamics theory with gene interaction network analysis. By designing a proprietary symbolic encoding algorithm, it transforms the dynamic processes of gene networks into analyzable symbolic sequences. Then, using methods such as symbolic entropy calculation and symbolic pattern recognition, it accurately extracts dynamic network features. Simultaneously, it introduces a dynamically adjusted analysis strategy, constructing a dynamically adaptive analysis system. This effectively overcomes the limitations of static analysis in existing technologies, achieving precise characterization of the dynamic changes in gene expression and regulatory relationships. It effectively solves the problems of existing methods being unable to characterize network dynamic behavior, having low analytical accuracy, and inaccurate identification of key information. Through multi-dimensional comprehensive scoring to filter key information, combined with a dynamically optimized parameter adjustment mechanism, this invention not only effectively filters out noise data interference and enhances adaptability to networks of varying complexity, but also improves the accuracy of identifying key interactions, key nodes, and key paths in complex networks, reducing false negative and false positive rates. This invention is particularly suitable for gene function research, disease pathogenesis analysis, and targeted drug development, providing more reliable technical support for related fields and possessing significant theoretical value and practical application prospects. Attached Figure Description

[0054] Figure 1 This is a framework diagram of the system of the present invention.

[0055] Figure 2 This is a flowchart of the method of the present invention.

[0056] Figure 3 This is a flowchart of step 3 in the method of the present invention.

[0057] Figure 4 This is a flowchart of step 4 in the method of the present invention.

[0058] Figure 5 This is a flowchart of step 6 in the method of the present invention. Detailed Implementation

[0059] To provide a clearer understanding of the technical features, objectives, and beneficial effects of this invention, the technical solution of this invention is described in detail below, but this should not be construed as limiting the scope of implementation of this invention. Unless otherwise specified, the methods used in this invention are conventional methods in this technical field.

[0060] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0061] One embodiment of the present invention provides a gene interaction network analysis system based on symbolic dynamics, the framework of which is shown in the figure below. Figure 1As shown, the system includes a data preprocessing module, a network construction module, a symbol encoding module, a symbol sequence feature analysis module, a dynamic adjustment module, a key information identification module, and a result output module. The coordination relationship between these modules is as follows:

[0062] The data preprocessing module is used to collect raw gene data (including gene expression data and regulatory relationship data), perform data cleaning, data standardization and noise removal, and output preprocessed gene dynamic data to provide basic data for subsequent network construction and encoding.

[0063] The network construction module is used to receive preprocessed gene dynamic data, construct a dynamic gene interaction network based on time series (nodes are genes, edges are the regulatory relationships between genes that change over time, and the weight of the edges is the regulatory strength), and output the dynamic gene interaction network.

[0064] The symbol encoding module receives the dynamic gene interaction network and maps the dynamic state (changes in node expression levels and changes in edge weights) of the dynamic gene interaction network into symbol sequences using a preset symbol encoding algorithm, and outputs a set of symbol sequences.

[0065] The symbol sequence feature analysis module is used to receive a set of symbol sequences, extract feature parameters of the symbol sequences (such as symbol entropy value and high-frequency symbol patterns) using symbol entropy calculation and symbol pattern recognition algorithms, and output a set of symbol sequence features.

[0066] The dynamic adjustment module receives the symbol sequence feature set, dynamically optimizes the symbol encoding parameters and analysis parameters according to preset adjustment rules, feeds back the optimized parameters to the symbol encoding module and the symbol sequence feature analysis module, and finally outputs the optimized symbol sequence feature set.

[0067] The key information identification module is used to receive the optimized symbol sequence feature set, combine it with network topology features, identify key interaction relationships, key nodes and key paths, and output key information identification results.

[0068] The results output module receives the key information identification results and outputs them in the form of visual charts (such as network topology diagrams and critical path diagrams) and data reports for users to use for subsequent analysis.

[0069] Each module achieves real-time data interaction through a data transmission interface, forming a complete workflow of "data input - preprocessing - network construction - symbol encoding - feature analysis - dynamic optimization - key identification - result output".

[0070] Another embodiment of the present invention provides a gene interaction network analysis method based on symbolic dynamics, the flowchart of which is shown below. Figure 2 As shown, the method includes the following steps:

[0071] Step 1: Preprocessing raw genetic data;

[0072] 1.1 Data Acquisition: Collect time-series data of gene expression levels from the target biological samples (assuming the acquisition time point is...). The total number of data collection time points n≥50; the number of genes is m, m≥100), and the data format is matrix. Simultaneously, data on known regulatory relationships between genes (including activation and repression regulation) are collected to form an initial set of regulatory relationships. .

[0073] 1.2 Data Cleaning: Data Removal For gene data with expression levels of 0 or a missing value ratio exceeding 5%, the remaining missing values ​​were filled using linear interpolation to obtain cleaned expression level time series data. The formula is as follows: ;

[0074] in Indicates the first The gene in the first Expression levels at each time point; For the first The gene in the first Expression levels at each time point; For the first The gene in the first Expression levels at each time point; The index for the current time point; The index of the time point where the missing value occurred; and An index for non-missing time points adjacent to the missing value.

[0075] 1.3 Data Standardization: The Z-score standardization method was used to standardize the cleaned expression time series data to obtain a standardized data matrix. The formula is:

[0076] ;

[0077] in, For the first The gene in the first Standardized expression at each time point; Indicates the first The gene in the first Expression levels at each time point; Let be the mean expression level of the i-th gene. ; Let be the standard deviation of the expression level of the i-th gene. .

[0078] Standardized data matrix The expression levels of each gene follow a normal distribution with a mean of 0 and a variance of 1.

[0079] 1.4 Noise Removal: The standardized data matrix was processed using wavelet thresholding denoising. The db4 wavelet basis function was selected, the decomposition level was set to 3, and the threshold was set accordingly. (σ is the noise standard deviation, estimated from the high-frequency coefficients after wavelet decomposition). After thresholding the high-frequency coefficients, the data is reconstructed to obtain the denoised gene dynamic data matrix. .

[0080] Step 2: Construction of dynamic gene interaction network;

[0081] 2.1 Calculation of intergene regulatory strength: based on the denoised gene dynamic data matrix The regulatory strength between the i-th gene and the j-th gene was calculated using time-delayed mutual information (TDMI). The formula is:

[0082] ;

[0083] in: This represents the time delay mutual information value between the i-th gene and the j-th gene; i and j represent the i-th gene and the j-th gene, respectively, and are two nodes in the network used to describe the regulatory relationship between genes; The time delay (ranging from 1 to 3 time points, with the optimal τ determined by maximizing mutual information) represents the time interval between the expression level of the i-th gene at time point t and the expression level of the j-th gene at time point t+τ; a and b represent the expression level of the i-th gene at time point t, respectively. All possible values ​​of the j-th gene at time point expression level All possible values; This represents the expression level of the i-th gene at time point t and the expression level of the j-th gene at time point t+1. The joint probability density of expression levels; This represents the expression level of the i-th gene at time point t; This indicates that the j-th gene is at time point [time]. The amount of expression; This represents the probability density of the expression level of the i-th gene at time point t; This indicates that the j-th gene is at time point t+ The probability density of expression levels.

[0084] 2.2 Constructing a dynamic gene-gene interaction network: using genes as nodes, regulatory relationships between genes as edges, and edge weights... Indicates the intensity of regulation; when the intensity of regulation... When the value is ≥0.6 (a threshold determined based on experimental data statistics), network edges are formed, and a directed weighted dynamic gene interaction network is constructed. ,in: For a set of nodes ( (corresponding to the i-th gene); Let be the set of edges at time t. (This indicates that at time t, the i-th gene has a regulatory effect on the j-th gene). Let be the set of edge weights at time t.

[0085] Step 3: Symbolic encoding of network dynamic states (the flowchart for this step is shown below) Figure 3 (as shown)

[0086] 3.1 Determining the coding dimension: The coding dimension includes the node state dimension (changes in gene expression levels) and the edge state dimension (changes in regulatory strength), where the node state is represented by... The edge state is represented by (t) represents.

[0087] 3.2 Node State Symbol Encoding:

[0088] Calculate the rate of change of node states: (t)=( - ) / (t≥2), where: (t) represents the rate of change in the expression level of the i-th gene at time t relative to time t-1; This represents the expression level of the i-th gene at time point t; This represents the expression level of the i-th gene at time t-1; t represents the time point.

[0089] Node state hierarchy and symbol mapping: (t) is divided into 5 intervals, corresponding to 5 symbols {0,1,2,3,4}, according to the following rules:

[0090] (t) < -0.2 → Sign 0 (significant downregulation);

[0091] -0.2 ≤ (t) < -0.05 → Symbol 1 (slight downsizing);

[0092] -0.05 ≤ (t) ≤ 0.05 → Symbol 2 (stable);

[0093] 0.05 < (t) ≤ 0.2 → Sign 3 (slightly increased);

[0094] (t) > 0.2 → Symbol 4 (significantly upregulated);

[0095] Obtain the node state symbol sequence .

[0096] 3.3 Edge State Symbol Encoding:

[0097] Calculate the rate of change of edge states: ,in: This represents the rate of change in the regulatory intensity of the i-th gene on the j-th gene at time t; This represents the regulatory strength of the i-th gene on the j-th gene at time t; Indicates in The regulatory strength of the i-th gene on the j-th gene at time t; Indicates a point in time.

[0098] Edge state hierarchy and symbol mapping: The system is divided into three intervals, corresponding to three symbols {α, β, γ}, according to the following rules:

[0099] < -0.1 → Symbol α (regulation weakens);

[0100] -0.1 ≤ ≤ 0.1 → Symbol β (regulatory stability);

[0101] > 0.1 → Symbol γ (enhanced regulation);

[0102] Obtain the edge state symbol sequence = [ (2), (3),..., (n)].

[0103] 3.4 Construction of the overall network symbol sequence: Integrate all node state symbol sequences and edge state symbol sequences to form the overall network symbol sequence set. (The symbol sequence corresponding to the edge that does not have a regulatory relationship is empty).

[0104] Step 4: Symbol sequence feature analysis (the flowchart for this step is shown below) Figure 4 (as shown)

[0105] 4.1 Symbolic Entropy Calculation (Quantifying Network Dynamic Complexity):

[0106] An improved multi-scale symbol entropy (MSE) algorithm is used to calculate the entropy value of the symbol sequence, based on the node state symbol sequence. For example, the specific steps are as follows:

[0107] Step 4.1.1 Sequence Segmentation: [The sequence is then segmented...] Divide into L subsequences of length k, with subsequence length k=4 (the optimal value is determined by grid search), and overlap step size of 1, L = n-1-k+1 = nk;

[0108] Step 4.1.2 Pattern Counting: Count the frequency of occurrence of each subsequence. Let the subsequence be... The number of times it appears is Then frequency = / L;

[0109] Step 4.1.3 Symbol Entropy Calculation: ;in: Represents a sequence of node state symbols Symbolic entropy is a metric for measuring sequence complexity; `p` represents the summation symbol, indicating that the summation is performed over all possible symbol patterns; `p` represents the index of the symbol pattern, from 1 to M; `M` represents the total number of distinct subsequences (the symbol set size is 5). (where k is the length of the subsequence). Representing a subsequence The probability of occurrence.

[0110] Similarly, compute the edge state symbol sequence. symbolic entropy (The symbol set size is 3, ).

[0111] 4.2 Symbol Pattern Recognition (Extracting Key Dynamic Patterns):

[0112] Step 4.2.1 Define key patterns: The key pattern of a node is defined as three consecutive symbols containing at least two "0"s or "4"s (i.e., significant downregulation or significant upregulation occurs consecutively), and the key pattern of an edge is two consecutive symbols being "α" or "γ" (i.e., significant changes in regulation intensity (regulation weakens or regulation strengthens) occur consecutively).

[0113] Step 4.2.2 Pattern Matching: The symbol sequence is matched using a sliding window method (the window size corresponds to the key pattern length), and the occurrence frequency of the key pattern for each node and edge is counted. , ;

[0114] Step 4.2.3 Pattern Weight Calculation: The pattern weight is determined based on the product of the frequency of occurrence of the key pattern and the symbol entropy. The formula is as follows:

[0115] ;

[0116] ;

[0117] in: Represents a sequence of node state symbols Pattern weights; Represents a sequence of node state symbols The frequency of occurrence of key patterns; Represents a sequence of node state symbols The symbolic entropy; Representing the edge state symbol sequence Pattern weights; Representing the edge state symbol sequence The frequency of occurrence of key patterns; Representing the edge state symbol sequence The symbolic entropy.

[0118] The higher the weight value, the greater the impact of the dynamic behavior of the node or edge on the overall function of the network.

[0119] Step 5: Analyze the dynamic adjustment of the strategy;

[0120] 5.1 Setting the adjustment threshold: Based on the statistical analysis of experimental data, set the symbol entropy threshold. =1.8 (nodes) =1.2 (edges), pattern weight threshold =5.0 (node) =3.0 (edges).

[0121] 5.2 Encoding Parameter Adjustment:

[0122] If the average symbol entropy of the node state symbol sequence is < This indicates that the symbol division is too coarse. The number of intervals for the node state change rate is adjusted from 5 to 7, and the node state symbol encoding in step 3.2 is re-executed.

[0123] If the average symbol entropy of the edge state symbol sequence is < The number of intervals for the edge state change rate is adjusted from 3 to 5, and the edge state symbol encoding in step 3.3 is re-executed.

[0124] 5.3 Analytical Parameter Adjustment:

[0125] If the variance of the mode weights of the node state symbol sequence is >2.0, it indicates that the key modes are unevenly distributed. Adjust the length k of the subsequence from 4 to 5 and re-execute the symbol entropy calculation in step 4.1.

[0126] If the variance of the pattern weights of the edge state symbol sequence is greater than 1.5, adjust the sliding window size from 2 to 3 and re-execute the symbol pattern recognition in step 4.2.

[0127] 5.4 Iterative Optimization: Repeat steps 3.2-5.3 until the statistical characteristics of symbol entropy and mode weights meet the preset stability condition (the average change in symbol entropy over two consecutive iterations < 0.05), and output the optimized symbol sequence feature set (including the final set). , , W( )).

[0128] Step 6: Key Information Identification (Flowchart of this step is shown below) Figure 5 (as shown)

[0129] 6.1 Identification of key interactions:

[0130] Overall score for edge computation: = 0.6×W( )+0.4× (The pattern weight of the edge state symbol sequence accounts for 60%, and the symbol entropy of the edge state symbol sequence accounts for 40%).

[0131] Filtering key interactions: Setting scoring thresholds =4.0, filter ≥ The edges represent the key interactions between genes.

[0132] 6.2 Key Node Identification:

[0133] Overall score of computing nodes: = 0.5× +0.3× +0.2×Degree Degree For nodes Degree centrality (normalized to the [0,1] interval);

[0134] Filtering key nodes: Setting scoring thresholds =6.0, filter ≥ The nodes that are key nodes in the network are called critical nodes.

[0135] 6.3 Critical Path Identification:

[0136] Step 6.3.1 Path Definition: A path in a network is a sequence of directed edges from the starting critical node to the ending critical node. , where P represents a path in the network, which is a sequence of directed edges; k is the number of edges in the path, i.e. the path length; Let represent the set of all edges in the path; where and all nodes All nodes are critical nodes, and all edges are critical interaction edges.

[0137] Step 6.3.2 Calculate path weights: The weight of a path is the ratio of the sum of the overall scores of all edges to the path length, using the following formula:

[0138] Weight(P) = (∑(p=1 to k) Score( )) / k;

[0139] Where: Weight(P) is the weight of the path; p is the index of the edge in the path; Score( ) represents the sum of the overall scores for all edges.

[0140] This formula balances the influence of path length and edge score, avoiding excessive bias towards long or short paths.

[0141] Step 6.3.3 Filtering critical paths: Setting path weight thresholds =3.5, filter Weight(P)≥ The path; if there are multiple paths, they are sorted in descending order of weight value, and the top 20% (or at least 3 paths) are taken as critical paths to ensure coverage of the core regulatory pathways in the network.

[0142] Step 7: Output the results;

[0143] The results are output in the form of visualizations and data reports. The visualizations include drawing network topology diagrams, marking key nodes and critical paths, and displaying dynamic characteristics of key nodes; the data reports summarize information on key nodes, interactions, and paths, and record parameter settings during the analysis process.

[0144] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention.

Claims

1. A gene interaction network analysis method based on symbolic dynamics, characterized in that, Includes the following steps: Step 1: Preprocessing raw genetic data; Collect raw gene data and perform data cleaning, data standardization and noise removal, and output preprocessed dynamic gene data; Step 2: Construction of dynamic gene interaction network; Receive dynamic gene data, construct a dynamic gene interaction network based on time series, the dynamic gene interaction network uses genes as nodes, the regulatory relationships between genes as edges, and the weight of the edges as the regulatory strength, and output the dynamic gene interaction network. Step 3: Symbolic encoding of network dynamic states; Receive a dynamic gene interaction network, map the dynamic state of the dynamic gene interaction network into a symbol sequence using a preset symbol encoding algorithm, and output a symbol sequence set; Step 4: Symbol sequence feature analysis; Receive a set of symbol sequences, use symbol entropy calculation and symbol pattern recognition algorithms to extract feature parameters of the symbol sequences, and output a set of symbol sequence features. Step 5: Analyze the dynamic adjustment of the strategy; It receives a symbol sequence feature set, dynamically optimizes the symbol encoding parameters and analysis parameters according to preset adjustment rules, feeds back the optimized parameters iteratively, and finally outputs the optimized symbol sequence feature set; Step 6: Identification of key information; Receive the optimized symbol sequence feature set, combine it with network topology features, identify key interaction relationships, key nodes and key paths, and output key information identification results; Step 7: Output the results; Receive key information identification results and output them in the form of visual charts and data reports.

2. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 1, the original gene data includes gene expression time series data and known regulatory relationship data between genes; data cleaning includes removing gene data with zero expression or missing values ​​exceeding a preset threshold, and filling the remaining missing values ​​with linear interpolation to obtain cleaned expression time series data. Data standardization was performed using the Z-score standardization method to standardize the cleaned expression time series data, resulting in a standardized data matrix. Noise removal was performed using wavelet threshold denoising to process the standardized data matrix, resulting in a denoised gene dynamic data matrix.

3. The gene interaction network analysis method based on symbolic dynamics according to claim 2, characterized in that, In step 2, constructing a dynamic gene interaction network specifically includes: calculating the regulatory intensity between genes based on the denoised gene dynamic data matrix using time-delayed mutual information; constructing a directed weighted dynamic gene interaction network with genes as nodes and gene pairs with regulatory intensity greater than or equal to a preset threshold as edges.

4. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 3, the dynamic state includes node state and edge state; The node state symbol encoding is specifically as follows: calculate the rate of change of gene expression, map the rate of change to the corresponding symbol according to the preset interval to which it belongs, and obtain the node state symbol sequence; The edge state symbol encoding is specifically as follows: calculate the rate of change of regulatory intensity between genes, map the rate of change to the corresponding symbol according to the preset interval to which it belongs, and obtain the edge state symbol sequence; Integrate all node state symbol sequences and edge state symbol sequences to form a complete network symbol sequence set.

5. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 4, the feature parameters include symbol entropy and mode weight; Symbol entropy is calculated through the following steps: dividing the symbol sequence into multiple subsequences, counting the frequency of occurrence of each subsequence, and then applying the formula... The symbolic entropy is calculated; where The symbol entropy represents the sequence of symbols; p represents the index of the symbol pattern; M represents the total number of distinct subsequences. Representing a subsequence The probability of occurrence; Pattern weights are calculated through the following steps: identify key patterns in the symbol sequence and count the number of times the key patterns appear, then multiply the number of times the key patterns appear by the corresponding symbol entropy to obtain the pattern weight.

6. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 5, dynamic optimization specifically includes: If the average symbol entropy of the node state symbol sequence is lower than the node entropy threshold, or the average symbol entropy of the edge state symbol sequence is lower than the edge entropy threshold, then increase the number of intervals for the corresponding symbol mapping and return to step 3 for re-encoding. If the variance of the pattern weights of the node state symbol sequence exceeds the node variance threshold, or the variance of the pattern weights of the edge state symbol sequence exceeds the edge variance threshold, then adjust the subsequence length for the corresponding symbol entropy calculation or the window size for key pattern recognition, and return to step 4 for reanalysis.

7. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 5, the specific process of feedback iteration is as follows: repeat steps 3 to 5 until the average change of symbol entropy and the variance of mode weights meet the preset stability condition, and then output the optimized symbol sequence feature set.

8. The gene interaction network analysis method based on symbolic dynamics according to claim 1, characterized in that, In step 6: Identifying key interactions specifically includes: calculating the comprehensive score of edges, which is the weighted sum of the pattern weights and symbol entropies corresponding to the edge state symbol sequences; and selecting edges whose comprehensive scores are greater than or equal to a preset score threshold, which are the key interactions between genes. Identifying key nodes specifically includes: calculating the comprehensive score of a node, which is a weighted sum of the pattern weight, symbol entropy, and degree centrality of the node's state symbol sequence; and selecting nodes whose comprehensive score is greater than or equal to a preset score threshold, which are the key nodes in the network. Identifying critical paths specifically includes: defining a path as a sequence of directed edges from the starting critical node to the ending critical node; calculating the path weight, which is the ratio of the sum of the comprehensive scores of all edges on the path to the path length; and selecting paths whose path weights are greater than or equal to a preset path weight threshold as critical paths.

9. A gene interaction network analysis system based on symbolic dynamics, characterized in that, The gene interaction network analysis method based on symbol dynamics as described in any one of claims 1-8 includes a data preprocessing module, a network construction module, a symbol encoding module, a symbol sequence feature analysis module, a dynamic adjustment module, a key information identification module, and a result output module. The data preprocessing module is used to collect raw gene data and perform data cleaning, data standardization and noise removal, and output preprocessed dynamic gene data. The network construction module is used to receive gene dynamic data, construct a dynamic gene interaction network based on time series, and output the dynamic gene interaction network with genes as nodes, regulatory relationships between genes as edges, and edge weights as regulatory strengths. The symbol encoding module is used to receive the dynamic gene interaction network, map the dynamic state of the dynamic gene interaction network into a symbol sequence through a preset symbol encoding algorithm, and output a symbol sequence set. The symbol sequence feature analysis module is used to receive a set of symbol sequences, extract feature parameters of the symbol sequences using symbol entropy calculation and symbol pattern recognition algorithms, and output a set of symbol sequence features. The dynamic adjustment module is used to receive the symbol sequence feature set, dynamically optimize the symbol encoding parameters and analysis parameters according to the preset adjustment rules, feed back the optimized parameters to the symbol encoding module and the symbol sequence feature analysis module, and finally output the optimized symbol sequence feature set. The key information identification module is used to receive the optimized symbol sequence feature set, combine it with network topology features, identify key interaction relationships, key nodes and key paths, and output key information identification results. The result output module is used to receive the key information identification results and output them in the form of visual charts and data reports.

Images