Method and system for constructing sepsis miRNA marker model based on evolutionary algorithm

Abstract

The invention relates to a sepsis miRNA marker model construction method and system based on an evolutionary algorithm, and belongs to the field of bioinformatics. The method comprises the following steps: firstly, carrying out dimension reduction processing on a miRNA expression matrix; then, a search objective function is constructed, and optimization objectives of training model performance maximization and elimination of multiple collinearity among miRNA screening results are taken; then, an improved genetic algorithm search strategy is adopted, a search target is decomposed into a plurality of sub-problems to be solved, for example, a three-classification problem is decomposed into three two-classification problems, and the optimal miRNA combination of healthy patients, severe and non-sepsis patients and sepsis patients is searched and recognized respectively. And finally, through parameter tuning, an optimal miRNA marker is obtained, and a sepsis miRNA marker model is constructed, for example, a machine learning algorithm such as a random forest or a support vector machine is used. The invention provides an efficient and accurate sepsis miRNA (micro Ribonucleic Acid) marker mining method.

Description

Technical Field

[0001] The present invention belongs to the field of bioinformatics and relates to a method and system for constructing a sepsis miRNA marker model based on an evolutionary algorithm. Background Art

[0002] Sepsis is a systemic inflammatory response syndrome caused by infection, which develops rapidly and has a high mortality rate. In recent years, with the development of high-throughput sequencing technology, microRNA (miRNA), as a type of non-coding RNA, has been found to have a significant impact on the occurrence and development of a variety of diseases, including sepsis. miRNA has shown great potential in the mining of sepsis markers, and changes in its expression level can be used as an early marker of sepsis.

[0003] However, it is still a challenge to effectively mine useful markers from a large amount of miRNA expression data. Existing miRNA marker screening methods have the following main problems:

[0004] High data dimension: miRNA expression matrix usually contains tens of thousands of miRNAs, with high data dimension and high computational complexity.

[0005] Multicollinearity: There are complex interactions between miRNAs, which leads to multicollinearity problems and affects the generalization ability of the model.

[0006] Model performance evaluation: The selection of indicators and calculation methods for evaluating the performance of miRNA marker models need further study.

[0007] Therefore, it is necessary to develop an efficient and accurate miRNA marker screening method to address the shortcomings of existing methods and improve the accuracy and efficiency of sepsis miRNA marker mining. Summary of the invention

[0008] In view of this, the purpose of the present invention is to provide a method and system for constructing a sepsis miRNA marker model based on an evolutionary algorithm, which can mine effective markers from complex miRNA expression data and construct an accurate model.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] A method for constructing a sepsis miRNA marker model based on an evolutionary algorithm, the method comprising the following steps:

[0011] Input the miRNA expression matrix after TPM processing; TPM values ​​represent the number of specific genes per million transcripts, which are normalized to relative abundance;

[0012] The miRNA expression matrix was subjected to dimensionality reduction for screening of sepsis miRNA markers;

[0013] The objective function for searching for sepsis miRNA markers was constructed, with maximizing the training model performance and eliminating multicollinearity among miRNA screening results as the optimization goals;

[0014] An improved genetic algorithm search strategy is used to decompose the search target into multiple sub-problems for solution;

[0015] Through parameter tuning, the optimal miRNA markers were obtained and the sepsis miRNA marker model was constructed.

[0016] Furthermore, the miRNA expression matrix dimensionality reduction process is composed of DESeq differential expression analysis, differentially expressed gene screening, and screening result presentation. The specific working principle is as follows:

[0017] Use DESeq differential expression analysis to output genes with significant differential expression;

[0018] Screening of significantly differentially expressed genes based on confidence;

[0019] The corresponding differential expression volcano plot was drawn for analysis.

[0020] Furthermore, the sepsis miRNA marker search objective function is:

[0021]

[0022] Among them, ModelScore represents the model performance score, Penalty represents the penalty coefficient, and f now represents the existing feature set, |f now | indicates the number of existing features, and MaxFeature indicates the maximum number of features.

[0023] Furthermore, in the sepsis miRNA marker search objective function, the model performance score is calculated using the input feature set f now , the corresponding miRNA expression data is used to train and test the random forest model, and the F1 Score of the test is the model performance score.

[0024] Furthermore, in the decomposition of the search target, the multi-classification problem with large information requirements is decomposed into multiple binary classification problems with small information requirements. For the best sepsis miRNA marker combination, it is expressed as follows:

[0025]

[0026] Among them, R best represents the best combination of sepsis miRNA markers, It indicates that the optimal miRNA combination for identifying healthy patients is represents the best miRNA combination for identifying critically ill but non-septic patients, represents the best miRNA combination for identifying sepsis patients;

[0027] According to the above formula, R best The search for and The union of the search results.

[0028] Furthermore, the improved genetic algorithm search strategy is specifically: using three genetic iterative populations to search respectively and The two-classification problems of these three sub-search models are referred to as discriminant HC, discriminant PC and discriminant SP respectively. To allow parallel operation and speed up the search, the three sub-search models are iterated at the same time, and the first m miRNA combinations of the three sub-search models are taken as and The candidate combination set and Cartesian product of all candidate combination sets is performed to combine the "perfect combination" candidate set R for the three-classification problem group :

[0029]

[0030] Prove that three candidate sub-combinations of size m can form at most combinations of size m 3 The "perfect combination" candidate set;

[0031] set up and There are and Then in R group The miRNA combination with the highest score must be R best ; To speed up the search of the model, R group As the search set, perform a crossover, genetic and mutation step on the population in the set, and take the best result as R best .

[0032] Furthermore, the parameter tuning is specifically as follows: grid search is performed on the initial feature number, population number and crossover rate of the sub-genetic algorithm; the best scoring sepsis miRNA marker combination is obtained; and the data corresponding to the combination is put into the random forest model to search for the best parameters, thereby obtaining the best model parameters of the random forest.

[0033] A sepsis miRNA marker model construction system based on an evolutionary algorithm, the system comprising:

[0034] The data preprocessing module receives the input miRNA expression matrix, performs dimensionality reduction on it, and outputs potential sepsis miRNA markers;

[0035] A search objective function module receives potential sepsis miRNA markers, constructs a search objective function, and outputs the search objective function;

[0036] The evolutionary algorithm module receives the search objective function and potential sepsis miRNA markers, and uses an improved genetic algorithm search strategy to decompose the search objective into multiple sub-problems for solution, decomposes the three-classification problem into three two-classification problems, and searches for the best miRNA combination for identifying healthy patients, identifying critically ill but non-septic patients, and identifying sepsis patients, and outputs the best miRNA marker combination;

[0037] A model building module receives the best miRNA marker combination, and uses a machine learning algorithm of a random forest or support vector machine to build a sepsis miRNA marker model, and outputs the sepsis miRNA marker model;

[0038] The model evaluation module receives the sepsis miRNA marker model and uses the accuracy or recall rate to evaluate the performance of the constructed sepsis miRNA marker model.

[0039] A computer-readable storage medium stores computer-executable instructions for executing the method.

[0040] .A computer program product comprising computer executable code of the method.

[0041] The beneficial effects of the present invention are as follows: the present invention provides a method for mining miRNA markers for sepsis, which uses the DESeq method to reduce the dimension of the miRNA expression matrix to reduce the complexity of subsequent screening; takes maximizing the performance of the training model and eliminating the multicollinearity between the miRNA screening results as the two main optimization goals to ensure that the search results are accurate and meet the requirements; and by decomposing the search problem into multiple sub-problems for solution, the model classification task is converted into three sub-tasks, providing more information for the feature screening model.

[0042] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0044] Figure 1 Flow chart of the miRNA expression data preprocessing method;

[0045] Figure 2 Flowchart of the sepsis miRNA marker mining method based on genetic algorithm. DETAILED DESCRIPTION

[0046] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0047] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0048] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "back" and the like indicate directions or positional relationships, they are based on the directions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0049] Aiming at hospital clinical data, this embodiment provides a miRNA marker mining method based on genetic algorithm.

[0050] The sepsis miRNA marker mining method based on genetic algorithm includes the following steps:

[0051] S1. Build a dataset.

[0052] S101, using the DESeq method to reduce the dimension of the miRNA expression matrix and obtain the miRNAs with significant differences;

[0053] S102, constructing a dataset consisting of miRNA expression data and clinical outcomes;

[0054] S103, dividing the data set into a training set and a test set;

[0055] Furthermore, an improved evolutionary algorithm search strategy is constructed.

[0056] S2. Set the sepsis miRNA marker search objective function as:

[0057]

[0058] Among them, ModelScore represents the model performance score, Penalty represents the penalty coefficient, and f now represents the existing feature set, |f now | indicates the number of existing features, and MaxFeature indicates the maximum number of features.

[0059] S3, constructing an evolutionary algorithm search strategy based on specific problems to obtain the best sepsis search model;

[0060] S301, convert the three-classification problem of healthy people HC, sick but non-septic people PC, and sepsis patients SP into three two-classification problems, namely, judging whether the sample is HC, judging whether the sample is PC, and judging whether the sample is SP.

[0061] S302, constructing an evolutionary algorithm search strategy based on the problem division results, using three genetic iterative populations to search respectively and The binary classification problems of these three sub-search models are referred to as discriminating HC, discriminating PC and discriminating SP. In order to allow parallel computing and speed up the search, this paper iterates the three sub-search models at the same time, and takes the first m miRNA combinations of the three sub-search models as and The candidate combination set and Cartesian product of all candidate combination sets is performed to combine the "perfect combination" candidate set R for the three-classification problem group .

[0062]

[0063] It can be proved that three sub-candidate combinations of size m can at most form a combination of size m. 3 The "perfect combination" candidate set.

[0064] Assumptions and There are and Then in R group The miRNA combination with the highest score must be R best However, due to the limitations of the sub-search model itself, and The components may be randomly distributed in and This will slow down the search speed of the search model. Therefore, in order to speed up the search speed of the model, this paper uses R group As the search set, perform a crossover, genetic and mutation step on the population in the set, and take the best result as R best .

[0065] S303, setting a specific sepsis search parameter range, using the training set to search, and obtaining the best sepsis miRNA marker combination.

[0066] S4. Obtain the sepsis miRNA marker model through parameter tuning;

[0067] Specifically: (1) The grid search method and training set were used to search for the optimal parameters of random forest and obtain the sepsis miRNA marker model with the best performance on the training set; (2) The performance of the sepsis miRNA marker model was further tested using the test set.

[0068] Example 1: Construction of a sepsis miRNA marker model based on hospital clinical data

[0069] Scenario: A hospital wants to use miRNA expression data to build a sepsis model so that patients can be identified earlier and treatment measures can be taken.

[0070] step:

[0071] Data collection: miRNA expression data of septic and non-septic patients were collected from the hospital and normalized by TPM.

[0072] Data preprocessing: The DESeq method was used to perform differential expression analysis on the miRNA expression matrix after TPM treatment to screen out miRNAs with significant differences.

[0073] Model construction:

[0074] Search objective function: Set the model performance score (such as F1Score of random forest) to be maximized and the number of features to be limited as the optimization goal.

[0075] Genetic algorithm: Using an improved genetic algorithm, the three-classification problem is decomposed into three two-classification problems (discriminating HC, discriminating PC, and discriminating SP), and the best miRNA combination is searched separately.

[0076] Parameter tuning: The grid search method was used to optimize the parameters of the random forest model to obtain the best sepsis miRNA marker model.

[0077] Model evaluation: Use the test set to evaluate the performance of the model, such as accuracy, recall, and other indicators.

[0078] Example 2: Construction of a sepsis miRNA marker model based on a public dataset

[0079] Scenario: Researchers want to use publicly available miRNA expression datasets to explore the potential application of miRNA markers for sepsis.

[0080] step:

[0081] Data download: miRNA expression data of septic and non-septic patients were downloaded from public databases (such as GEO and TCGA) and normalized by TPM.

[0082] Data preprocessing: The DESeq method was used to perform differential expression analysis on the miRNA expression matrix after TPM treatment to screen out miRNAs with significant differences.

[0083] Model construction:

[0084] Search objective function: Set the model performance score (such as the accuracy of the support vector machine) to be maximized and the number of features to be limited as the optimization goal.

[0085] Genetic algorithm: Using an improved genetic algorithm, the three-class classification problem is decomposed into three two-class classification problems, and the best miRNA combination is searched separately.

[0086] Parameter tuning: The parameters of the support vector machine model were optimized using the cross-validation method to obtain the best sepsis miRNA marker model.

[0087] Model evaluation: Use the test set to evaluate the performance of the model, such as accuracy, recall, and other indicators.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a sepsis miRNA marker model based on an evolutionary algorithm, characterized in that: The method comprises the following steps: Input the miRNA expression matrix after transcript per million (TPM); The miRNA expression matrix was subjected to dimensionality reduction for screening of sepsis miRNA markers; The objective function for searching for sepsis miRNA markers was constructed, with maximizing the training model performance and eliminating multicollinearity among miRNA screening results as the optimization goals; An improved genetic algorithm search strategy is used to decompose the search target into multiple sub-problems for solution; Through parameter tuning, the optimal miRNA markers were obtained and the sepsis miRNA marker model was constructed.

2. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 1, characterized in that: The miRNA expression matrix dimensionality reduction process consists of DESeq differential expression analysis, differentially expressed gene screening, and screening result presentation. The specific working principle is as follows: Differential gene expression analysis based on the negative binomial distribution (DESeq) was used to output genes with significant differential expression. Screening of significantly differentially expressed genes based on confidence; The corresponding differential expression volcano plot was drawn for analysis.

3. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 1, characterized in that: The sepsis miRNA marker search objective function is: Among them, ModelScore represents the model performance score, Penalty represents the penalty coefficient, and f now represents the existing feature set, |f now | indicates the number of existing features, and MaxFeature indicates the maximum number of features.

4. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 1, characterized in that: In the sepsis miRNA marker search objective function, the model performance score is calculated using the input feature set f now , the corresponding miRNA expression data is used to train and test the random forest model, and the F1 Score of the test is the model performance score.

5. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 1, characterized in that: In the search target decomposition, a multi-classification problem with a large amount of information demand is decomposed into multiple binary classification problems with a small amount of information demand. For the best sepsis miRNA marker combination, it is expressed as follows: Among them, R best represents the best combination of sepsis miRNA markers, It indicates the identification of the best miRNA combination in healthy patients. represents the best miRNA combination for identifying critically ill but non-septic patients, represents the best miRNA combination for identifying sepsis patients; According to the above formula, R best The search for and The union of the search results.

6. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 5, characterized in that: The improved genetic algorithm search strategy is specifically: using three genetic iterative populations to search respectively and The two-classification problems of these three sub-search models are referred to as discriminant HC, discriminant PC and discriminant SP respectively. To allow parallel operation and speed up the search, the three sub-search models are iterated at the same time, and the first m miRNA combinations of the three sub-search models are taken as and The candidate combination set and Cartesian product of all candidate combination sets is performed to combine the "perfect combination" candidate set R for the three-classification problem group : Prove that three sub-candidate combinations of size m can be combined into at most m 3 The "perfect combination" candidate set; set up and There are and Then in R group The miRNA combination with the highest score must be R best ; To speed up the search of the model, R group As the search set, perform a crossover, genetic and mutation step on the population in the set, and take the best result as R best .

7. The method for constructing a sepsis miRNA marker model based on an evolutionary algorithm according to claim 4, characterized in that: The parameter tuning is specifically as follows: grid search is performed on the initial feature number, population number and crossover rate of the sub-genetic algorithm; the best scoring sepsis miRNA marker combination is obtained; and the data corresponding to the combination is put into the random forest model to search for the best parameters, and the best model parameters of the random forest are obtained.

8. A sepsis miRNA marker model construction system based on evolutionary algorithm, characterized by: The system includes: The data preprocessing module receives the input miRNA expression matrix, performs dimensionality reduction on it, and outputs potential sepsis miRNA markers; A search objective function module receives potential sepsis miRNA markers, constructs a search objective function, and outputs the search objective function; The evolutionary algorithm module receives the search objective function and potential sepsis miRNA markers, and uses an improved genetic algorithm search strategy to decompose the search objective into multiple sub-problems for solution, decomposes the three-classification problem into three two-classification problems, and searches for the best miRNA combination for identifying healthy patients, identifying critically ill but non-septic patients, and identifying sepsis patients, and outputs the best miRNA marker combination; A model building module receives the best miRNA marker combination, and uses a machine learning algorithm of a random forest or support vector machine to build a sepsis miRNA marker model, and outputs the sepsis miRNA marker model; The model evaluation module receives the sepsis miRNA marker model and uses the accuracy or recall rate indicator to evaluate the performance of the constructed sepsis miRNA marker model.

9. A computer-readable storage medium, characterized in that: Computer executable instructions are stored thereon for executing the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that: A computer executable code comprising the method according to any one of claims 1 to 7.

Images