A consensus inference method for gene regulatory networks based on adaptive feature analysis

Abstract

This invention belongs to the field of bioinformatics and relates to a consensus inference method for gene regulatory networks based on adaptive feature analysis. First, by inputting the inference results of multiple algorithms and performing preprocessing, the system automatically extracts five key features: gene quantity, edge quantity, weight mean, weight variance, and network density, and calculates the consistency score among each algorithm. Second, based on the features and consistency scores, the ensemble weights of each algorithm are dynamically allocated, and a weighted Borda counting method is used to fuse them to generate a consensus ranking. The optimal screening threshold is adaptively determined based on the gene quantity and network density. Finally, the obtained consensus network is quantitatively quality-assessed, and an evaluation report including coverage, stability, and consensus strength, along with the final network, is output. This invention achieves fully automated, data-driven consensus inference, significantly reducing reliance on external parameters and prior knowledge, and providing a reliable and interpretable computational tool for accurately identifying gene regulatory relationships.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics, specifically relating to a consensus inference method for gene regulatory networks based on adaptive feature analysis. This method can automatically analyze input data features, dynamically adjust ensemble parameters, and output a high-quality consensus network. Background Technology

[0002] Gene regulatory network inference is crucial for understanding cellular function and disease mechanisms. Single inference algorithms, such as GENIE3 and GRNBOOST2, often produce unstable and biased results. To address this, the consensus method COFFEE uses a fixed-weight Borda counting strategy to fuse the edge-ranking results of multiple algorithms, improving robustness. However, this method still suffers from the following problems: ① Lack of adaptability: Using fixed weights and thresholds, it cannot dynamically adjust based on input data characteristics such as network size and density; ② Result distortion: Insufficient utilization of key features such as network density and weight distribution in the algorithm output leads to ensemble results deviating from the true biological context; ③ Lack of evaluation: The lack of a system for quantitatively evaluating the quality of the output consensus network casts doubt on the reliability of the results. Therefore, there is an urgent need in this field for an adaptive consensus inference method that can perceive data characteristics, make dynamic decisions, and provide objective evaluation. Summary of the Invention

[0003] This invention proposes a consensus inference method for gene regulatory networks based on adaptive feature analysis, aiming to solve the problems of strong parameter dependence, unstable results, and lack of objective evaluation in existing technologies, and to achieve data-driven, fully automated consensus network inference. The specific technical solution includes the following five steps:

[0004] Step 1: Preprocessing and Feature Extraction: Read the sorted edge list generated by multiple gene regulatory network inference algorithms and perform data cleaning. Then, automatically extract the multidimensional feature vector of the input data, including the number of genes, the number of edges, the total mean, the weight variance, and the core network density calculated based on high-confidence edges.

[0005] Step 2, Dynamic Weight Allocation: Calculate the consistency score between each input algorithm and the results of other algorithms. Based on the preset baseline weights of each algorithm, and combining their consistency scores and weight distribution characteristics, dynamically calculate the final weight of each algorithm in the consensus integration.

[0006] Step 3: Weighted Consensus Integration. Using the dynamic weights obtained in Step 2, the traditional Borda counting method is improved with weighted weights. The score of each edge in each algorithm is multiplied by its weight, and finally the weighted scores of all algorithms are summed to obtain the total consensus score of each edge, and a consensus ranking is generated based on this.

[0007] Step 4: Intelligent Threshold Selection and Quality Assessment. Based on the gene quantity and network density characteristics extracted in Step 1, the optimal threshold for screening the final consensus network is automatically calculated using a non-linear mapping function. Simultaneously, the generated consensus network undergoes a quantitative quality assessment, outputting three metrics: coverage, stability, and consensus strength.

[0008] Step 5: Output Results. When the output meets the threshold conditions, the final consensus network file will be output, along with a detailed quality assessment report.

[0009] A consensus inference method for gene regulatory networks based on adaptive feature analysis is proposed. The implementation process of step 1 is as follows: The system reads all algorithm result files from the user-specified directory. These files are tab-delimited text files and must contain three columns: Gene1, Gene2, and Edge_weight. Data cleaning is performed: First, all self-loop edges that are the same for Gene1 and Gene2 are removed. Second, the Edge_weight column is checked to ensure that it is a numeric type. Finally, the edge list of each algorithm is sorted in descending order according to the Edge_weight column to form an initial sorted list. Subsequently, the multidimensional feature vector of the input data is automatically extracted, including the number of genes, the number of edges, the mean weight, the variance of the weight, and the core network density.

[0010] A consensus inference method for gene regulatory networks based on adaptive feature analysis is proposed, and the implementation process of step 2 is as follows:

[0011] Calculate the consistency score vector among the algorithms; for each algorithm, take the top 100 edges from its sorted list to form a set; for any two algorithms, calculate their Jaccard similarity; the consistency score of each algorithm is the arithmetic mean of its similarity scores with all other algorithms; subsequently, perform dynamic weight allocation, and the final adaptive weights are given by the formula:

[0012] The calculation yielded the result.

[0013] A consensus inference method for gene regulatory networks based on adaptive feature analysis is proposed, and the implementation process of step 3 is as follows:

[0014] Perform a weighted Borda consensus ensemble; for each edge, its score in the algorithm is: The total consensus score of an edge is the sum of its scores in all the algorithms that have appeared. The system sorts all edges in descending order based on the total consensus score to generate the final consensus ranking.

[0015] A consensus inference method for gene regulatory networks based on adaptive feature analysis is proposed. Step 4 is implemented as follows: The intelligent threshold selection function is:

[0016] Simultaneously, a quantitative quality assessment of the consensus network is conducted, calculating three indicators: coverage, stability, and consensus strength.

[0017] A consensus inference method for gene regulatory networks based on adaptive feature analysis is proposed, and the implementation process of step 5 is as follows:

[0018] The system writes the edges filtered by the intelligent threshold in step 4, along with their corresponding total consensus score, standardized score, and calculated confidence level, into the final consensus network file. This consensus network file is a tab-delimited text file containing five columns: Gene1, Gene2, BordaScore, NormalizedScore, and Confidence. Simultaneously, the system generates a quality assessment report, also a tab-delimited text file, containing, but not limited to: the dynamic threshold used in this running example, the adaptive weights assigned to each algorithm, the feature vectors extracted from the input data, and the specific values ​​of the three quality indicators—coverage, stability, and consensus strength—calculated in step 4. Finally, the system generates the adaptive_consensus_network.tsv consensus network file and the quality_report.tsv quality assessment report in the specified output directory. Attached Figure Description

[0019] Figure 1 This is a diagram illustrating the overall architecture of a consensus inference method for gene regulatory networks based on adaptive feature analysis.

[0020] Figure 2 This is a flowchart of the adaptive feature extraction module.

[0021] Figure 3 The flowchart shows the dynamic weight allocation and weighted consensus integration module.

[0022] Figure 4 This is a flowchart for the intelligent threshold selection and quality assessment module. Detailed Implementation

[0023] The present invention will now be described in detail with reference to the accompanying drawings and examples.

[0024] Preprocessing and feature extraction. For example... Figure 1 and Figure 2 As shown, this step includes two stages: data input and feature calculation. In the data input stage, the system reads all algorithm result files from the user-specified directory. These files are tab-delimited text files and must contain three columns: Gene1: regulating gene, Gene2: regulated gene, ... Edge_weightThe intensity of the control is a non-negative real number. The system uses the pandas library to read all files in parallel and performs data cleaning operations: first, it removes all self-loop edges that are identical between Gene1 and Gene2; second, it checks... Edge_weight Columns, ensuring they are of numeric type; finally, for each algorithm's separate edge list, sort by... Edge_weight Sort the columns in descending order to form an initial sorted list. Through the above processing, the system will obtain four sorted lists of edges. In the feature extraction stage, the system automatically extracts multidimensional feature vectors from the input data. Among them, the number of genes Merge the Gene1 and Gene2 columns from all edge lists, and use the pandas.unique() function to get the set of unique genes, the number of which is 1. In this example, =1000; Number of edges Calculate the length of the edge list for each algorithm and take its arithmetic mean. In this example, =45000; Weighted mean With weighted variance : Merge all edge lists Edge_weight The column uses the NumPy library to calculate the mean and variance. In this example, =0.004, =0.0001; Core network density The system first extracts the top 100 edges by weight from the sorted list of each algorithm, forming a high-confidence edge set. E_top100 Here, 100 is a fixed parameter designed to capture the most precise regulatory relationship for each algorithm, and then calculate... = | E_top100 | / [ * ( - 1) / 2]. Among them, | E_ top 100 |for sets E_top100 The actual number of edges in the middle may vary depending on the algorithm, as the first 100 edges may be repeated. The denominator is the theoretically maximum number of edges. In this example, | E_top100 | = 5970, therefore = 5970 / (1000*999 / 2) = 0.012. Finally, the system generates a feature vector. F = [1000, 45000, 0.004, 0.0001, 0.012].

[0025] Dynamic weight allocation and consensus integration. For example... Figure 3As shown, the algorithm consistency is first calculated, followed by dynamic weight allocation, and finally weighted consensus integration is performed. First, the consistency score vector among the algorithms is calculated. For each algorithm i Take the first 100 edges from its sorted list to form a set. For any two algorithms i and j Calculate their Jaccard similarity = | ∩ | / | ∪ |. Each algorithm i Consistency score It is the arithmetic mean of its similarity to all other algorithms, i.e. = mean( ) for all j ≠ i In this embodiment, the calculation is as follows: C_ GENIE3 = 0.65, C_GRNBOOST2 = 0.58, C_PPCOR = 0.45, C_PIDC = 0.28. Subsequently, dynamic weight allocation is performed. The system pre-stores the baseline weights for each algorithm. : B_GENIE3 = 1.2, B_GRNBOOST2 = 1.15, B_PPCOR = 1.1, B_PIDC = 1.0; Final adaptive weights Calculated using the following formula: , where mean( C ) for all The mean, For the algorithm i The standard deviation of the weighted column, The mean of the standard deviations of the weights of all algorithms, and the adjustment coefficient. = 0.4, = 0.3. Taking GENIE3 as an example, substituting the numerical values, its final weight W_GENIE3 = 1.27 is obtained. Similarly, the weights of other algorithms can be obtained. Finally, weighted Borda consensus integration is performed. Let the length of the sorting list for a single algorithm be... N For each edge e Its algorithm i The score is: ( e ) = ( N - r),in r For this edge in the algorithm i In the ranking, 0 is the highest. (Side) e The total consensus score is the sum of the scores of all the algorithms that appeared in it: TotalScore( e ) = Σ ( e The system uses TotalScore () e Sort all edges in descending order to generate the final consensus ranking.

[0026] Intelligent threshold selection and quality assessment. For example... Figure 4 As shown, this step automatically determines the screening threshold and performs a quantitative evaluation of the consensus network. The intelligent threshold selection function is: ,in Function x Limited to the range Inside, the feature values ​​of this instance will be... =1000, Substituting 0.012 into the calculation: =0.7645, the system will retain the standardized consensus score. The edges are used to form the final consensus network. Simultaneously, the system performs a quantitative quality assessment, evaluating three indicators: coverage, stability, and consensus strength. Coverage is one of the key metrics. , It is an algorithm i The former K =100 edges These are the edges in the final consensus network; this metric measures the extent to which the consensus network covers the top predictions of each algorithm. Stability: Multiple consensus networks are generated by resampling 100 times using Bootstrap, and the calculation of each edge is performed. e The ranking set of these networks (ranks) e Then, its stability is calculated; consensus strength: The average score of all edges in the final consensus network is calculated, and the metric is finally written to the quality_report.tsv file.

[0027] The baseline weights described in this invention are pre-determined through web search and regression analysis on multiple gold-standard datasets such as DREAM4 and DREAM5. To verify the effectiveness of this invention, tests were conducted on the DREAM5 benchmark dataset. Compared with the traditional COFFE method, this invention improved the F-score from 0.721 to 0.744, an improvement of 2%, demonstrating the effectiveness and superiority of this invention.

[0028] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for inferring consensus of gene regulatory network based on adaptive feature analysis, characterized in that, Comprising the following steps: Step 1, preprocessing and feature extraction: read the sorted edge list generated by multiple gene regulatory network inference algorithms, and perform data cleaning; then, automatically extract the multi-dimensional feature vector of the input data, including the number of genes, the number of edges, the mean weight, the variance of the weight, and the core network density calculated based on high-confidence edges; Step 2, dynamic weight assignment: calculate the consistency score between each input algorithm and the results of other algorithms; Based on the preset baseline weight of each algorithm, combined with the consistency score and weight distribution characteristics, the final adaptive weight is calculated by the following formula: , wherein, is the baseline weight, is the algorithm consistency score, mean(C) is the mean of all , and is the weight column of the algorithm i , mean(σ) is the mean of all algorithm weight standard deviations, and are adjustment coefficients; Step 3, weighted consensus integration: use the dynamic weights obtained in step 2 to improve the traditional Borda counting method; the score of each edge in each algorithm will be multiplied by its weight, and the weighted scores of all algorithms will be summed up to obtain the total consensus score of each edge, and a consensus ranking will be generated accordingly; Step 4, intelligent threshold selection and quality evaluation: according to the number of genes and network density features extracted in step 1, a nonlinear mapping function is used to automatically calculate the best threshold for filtering the final consensus network; at the same time, the generated consensus network is quantitatively evaluated, and three indicators of coverage, stability and consensus strength are output; Step 5, result output: output the final consensus network file that meets the threshold condition, and attach a detailed quality evaluation report.

2. The method of claim 1, wherein, In step 1, the data cleaning operation includes: firstly, eliminating all self-loop edges of Gene1 and Gene2; secondly, checking Edge_weight the column to ensure it is a numerical type; finally, arranging the descending order of each algorithm's individual edge list according to Edge_weight the column to form an initial sorted list .

3. The method of claim 1, wherein, In step 4, the nonlinear mapping function is: where is the number of genes, is the network density, and the clamp(x, a, b) function limits x to the interval [a, b].

Images