Biomarker screening method and system, electronic equipment and medium

Abstract

The invention provides a biomarker screening method and system, electronic equipment and a medium, and the method comprises the following steps: carrying out preliminary screening on multi-modal biological sample data to obtain preliminary screening data, and preprocessing the preliminary screening data; training a plurality of machine learning models based on the pre-processed preliminary screening data, and selecting an optimal model from the plurality of machine learning models as a biomarker screening model; performing feature screening on the basis of features of substances in primary screening data output by the biomarker screening model and corresponding feature weight values, and performing secondary screening on the features subjected to feature screening by adopting a preset screening strategy to obtain a target biomarker panel; wherein the preset screening strategy comprises one or more of an accumulated weighted value screening method, a cross recursive feature elimination method, a correlation coefficient screening method and an accumulated feature AUC screening method. The problems existing in the existing biomarker screening process are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of high-throughput biological omics data analysis, and in particular to a biomarker screening method and system, an electronic device and a medium. BACKGROUND

[0002] In high-throughput biological omics data analysis, biomarker screening is the most important step. Traditional machine learning model-based biomarker screening mainly adopts a single machine learning model screening strategy, such as Lasso regression, random forest (RF), Naive Bayes (NB), support vector machine (SVM), etc. However, the single model screening strategy is prone to missing weak effect biomarkers, leading to unstable biomarker panels and affecting the clinical application of biomarkers. Therefore, multi-model screening strategies have emerged, but most of them are single model selection or have defects in multi-modal feature data preprocessing, which risks missing weak effect biomarkers and fails to integrate multi-modal data modeling due to data heterogeneity. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a biomarker screening method, system, electronic device and medium to improve the problems existing in the current biomarker screening process.

[0004] To achieve the above purpose, the technical scheme adopted by the present application is as follows: In a first aspect, the present application provides a biomarker screening method, comprising: performing preliminary screening on multi-modal biological sample data to obtain preliminary screening data, and preprocessing the preliminary screening data; training a plurality of machine learning models based on the preprocessed preliminary screening data, and selecting the best model from the plurality of machine learning models as a biomarker screening model; performing feature screening on the features of the substances in the preliminary screening data output by the biomarker screening model and the corresponding feature weight values, and performing secondary screening on the features after feature screening using a preset screening strategy to obtain a target biomarker panel; wherein the preset screening strategy includes one or more of the following: cumulative weight value screening method, cross-recursive feature elimination method, correlation coefficient screening method and cumulative feature AUC screening method.

[0005] Optionally, the preliminary screening of the multi-modal biological sample data to obtain the preliminary screening data, and the preprocessing of the preliminary screening data, comprise: performing difference analysis on the multi-modal biological sample data using a statistical test method to obtain the preliminary screening data; based on the data type of the preliminary screening data, filling the missing values of the preliminary screening data using a corresponding filling method, and standardizing the preliminary screening data.

[0006] Optionally, the plurality of machine learning models are trained based on the preprocessed screening data, and the best model is selected from the plurality of machine learning models as the biomarker screening model, including: training a plurality of machine learning models based on the preprocessed screening data, and calculating the evaluation index of each machine learning model by using the K-fold cross-validation strategy; wherein the evaluation index includes: accuracy, AUC value, recall rate, precision, F1 value and Kappa coefficient; and the best model is selected from the plurality of machine learning models as the biomarker screening model based on the evaluation index.

[0007] Optionally, the characteristics of the substances in the screening data output by the biomarker screening model and the corresponding feature weight values are subjected to feature screening, including: filtering out the features with feature weight values less than a preset weight threshold based on the feature weight values output by the biomarker screening model.

[0008] Optionally, after the secondary screening of the features screened by the preset screening strategy is performed to obtain the target biomarker panel, the method further includes: drawing the ROC curve of the biological sample data in the test set based on the biomarker screening model and the target biomarker panel; calculating the area under the ROC curve, and evaluating the target biomarker panel based on the area.

[0009] Optionally, the method further includes: performing marker evaluation and marker verification on each marker in the target biomarker panel; wherein the marker evaluation includes: performing marker evaluation by using a box plot or correlation analysis; and the marker verification includes: verifying the correctness of the target biomarker panel by using one or more of decision tree, support vector machine, random forest, K-nearest neighbor method and Naive Bayes method.

[0010] Optionally, after the marker evaluation and the marker verification are performed on each marker in the target biomarker panel, the method further includes: calculating the regression coefficient and the intercept of each marker in the target biomarker panel by using a logistic regression algorithm; generating a regression equation based on the regression coefficient and the intercept, constructing a diagnostic model, and performing performance evaluation on the diagnostic model.

[0011] In a second aspect, the present application provides a biomarker screening system, comprising: a preliminary screening module configured to perform preliminary screening on multi-modal biological sample data to obtain preliminary screening data, and to perform preprocessing on the preliminary screening data; a model training module configured to train a plurality of machine learning models based on the preprocessed preliminary screening data, and to select an optimal model from the plurality of machine learning models as a biomarker screening model; and a marker screening module configured to perform feature screening on a substance in the preliminary screening data output by the biomarker screening model based on a feature of the substance and a corresponding feature weight value, and to perform secondary screening on the features after the feature screening using a preset screening strategy to obtain a target biomarker panel; wherein the preset screening strategy comprises one or more of the following: an accumulated weight value screening method, a cross-recursive feature elimination method, a correlation coefficient screening method, and an accumulated feature AUC screening method.

[0012] In a third aspect, the present application provides an electronic device, comprising a processor and a memory, the memory storing computer executable instructions capable of being executed by the processor, and the processor executes the computer executable instructions to implement the steps of the method of any one of the first aspect.

[0013] In a fourth aspect, the present application provides a computer readable storage medium, the computer readable storage medium storing a computer program, and the computer program being executed by a processor to perform the steps of the method of any one of the first aspect.

[0014] The present application brings the following beneficial effects: The application provides the above-mentioned biomarker screening method, system, electronic equipment and medium, first, the multi-modal biological sample data is subjected to preliminary screening to obtain screening data, and the screening data is preprocessed; then, a plurality of machine learning models are trained based on the preprocessed screening data, and the best model is selected from the plurality of machine learning models as a biomarker screening model; finally, the characteristics of the substances in the screening data output by the biomarker screening model and the corresponding characteristic weight values are subjected to feature screening, and the features screened by the features are subjected to secondary screening by using a preset screening strategy, to obtain a target biomarker panel; wherein the preset screening strategy comprises one or more of the following: cumulative weight value screening method, cross-recursive feature elimination method, correlation coefficient screening method and cumulative feature AUC screening method. In the above-mentioned method, the multi-modal biological sample data is subjected to preliminary screening and preprocessing, so that the data heterogeneity problem of the multi-modal biological sample data can be improved; the best model is selected from the plurality of models as the biomarker screening model through multi-model training, and the output results of the biomarker screening model are used for feature screening, and the cumulative weight value screening method, the cross-recursive feature elimination method, the correlation coefficient screening method and the cumulative feature AUC screening method are used for secondary screening, so that the risk of missing weak effect markers can be effectively reduced, and biomarkers with high sensitivity (multiple detected markers) and high specificity (low redundancy) can be screened out.

[0015] Other features and advantages of the present application will be set forth in the descriptions below, and in part will become apparent to those skilled in the art upon examination of the following or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and attained by the structure particularly pointed out in the description, claims and drawings.

[0016] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0018] Figure 1 A flowchart of a biomarker screening method provided by an embodiment of the present application; Figure 2 A schematic diagram of a target biomarker panel provided by an embodiment of the present application; Figure 3A feature weight value schematic diagram provided for an embodiment of the present application; Figure 4 A ROC curve schematic diagram provided for an embodiment of the present application; Figure 5 A flowchart of another biomarker screening method provided for an embodiment of the present application; Figure 6 A structural schematic diagram of a biomarker screening system provided for an embodiment of the present application; Figure 7 A structural schematic diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described below in detail with reference to the accompanying drawings. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0020] At present, the process of screening biomarkers based on a machine learning model is generally based on large sample size omics data. 70%-80% of the data are used as a training set to screen and evaluate a marker panel with good classification effect, and the remaining 20%-30% of the data are used as a test set to verify the classification stability of the marker panel. However, this method has the following problems: (1) Model singularity problem: only a single model is used to model and screen markers, without considering the complementary features that other models may capture, and there is a risk of missing weak effect markers.

[0021] (2) Multi-model integration blank: the current multi-model screening strategy adopts a “voting method” to directly select or discard a single model result, or directly selects top features according to feature weights / importance, without performing feature importance fusion and feature redundancy evaluation and removal, which may cause the complexity of the marker panel.

[0022] (3) Defects in multi-modal data processing: the current process needs to separately process measurement / continuous data, and omics research (such as protein+metabolome) often fails to integrate due to data heterogeneity.

[0023] Therefore, the biomarker screening method, system, electronic device and medium provided by the embodiments of the present application can improve the problems existing in the current biomarker screening process.

[0024] To facilitate understanding of this embodiment, a biomarker screening method disclosed in this invention will first be described in detail. This method can be executed by electronic devices, such as smartphones, computers, and tablets. See also Figure 1 The flowchart shown illustrates a biomarker screening method, which mainly includes the following steps S101 to S103: Step S101: Perform preliminary screening on the multimodal biological sample data to obtain initial screening data, and preprocess the initial screening data.

[0025] In one embodiment, multimodal biological sample data is acquired, including proteomic data and metabolomic data, wherein the proteomic data includes at least: label-free proteome, TMT / iTRAQ proteome and DIA proteome; and the metabolomic data includes at least: non-targeted metabolome, target-like metabolome and lipidome.

[0026] After obtaining multimodal biological sample data, the first step is to preliminarily screen biomarkers through differential analysis to obtain initial screening data. Specifically, differential analysis can be performed on the multimodal biological sample data. For metabolomics data, differential analysis mainly uses the PLS-DA model, referring to three parameters: VIP, FC, and P-value, to screen for differentially expressed metabolites. VIP refers to the variable importance in the projection of the first principal component of the PLS-DA model, representing the contribution of metabolites to grouping; FC refers to the fold change, which is the ratio of the mean of the repeated quantitative values ​​of each metabolite in the comparison group; P-value is calculated using the T-test and represents the significance level of the difference. For proteomics data, differential analysis mainly uses the T-test, referring to two parameters: FC and P-value, to screen for differentially expressed proteins.

[0027] The initial screening data then undergoes preprocessing, employing appropriate methods to impute missing values ​​and standardize the data for different omics datasets. Specifically, for proteomics data, multiple imputation, KNN imputation, and mean imputation are selected based on user preferences to impute missing values, followed by logarithmic transformation for standardization. Metabolomics data is typically a pre-imputed data matrix, requiring only logarithmic transformation. These processes mitigate the issue of data heterogeneity during preprocessing.

[0028] Step S102: Train multiple machine learning models based on the preprocessed initial screening data, and select the best model from the multiple machine learning models as the biomarker screening model.

[0029] In one implementation, the preprocessed initial screening data is used to train four machine learning models: Ada BoostClassifier (Ada), Gradient Boosting Classifier (Gbc), Extreme Gradient Boosting (Xgboost), and Light Gradient Boosting Machine (Lightgbm). The best model from these four models is then selected as the biomarker screening model. In practice, multiple machine learning models are first trained based on the preprocessed initial screening data, and a K-fold cross-validation strategy is used to calculate the evaluation metrics for each model. These metrics include accuracy, AUC, recall, precision, F1 score, and Kappa coefficient. Then, the best model is selected from these models based on these evaluation metrics. Specifically, the model with the highest AUC can be chosen as the best model. Weights can also be assigned to each evaluation metric based on the specific task requirements and importance. Weight allocation can be based on expert experience, business needs, or data analysis. If the task prioritizes recall, higher weight can be given to recall; if the task prioritizes precision, higher weight can be given to precision. Then, a weighted score is calculated for each model, and finally, the model with the highest weighted score is selected as the best model. The weighted score formula is as follows: Weighted score = Accuracy × w 1 + AUC value × w 2 + Recall Rate × w 3+accuracy× w 4 + F1 value × w 5 + Kappa coefficient × w 6. Among them, w 1. w 2. w 3. w 4. w 5. w 6 represents the weight of each evaluation indicator.

[0030] Step S103: Based on the characteristics of substances and corresponding feature weight values ​​in the initial screening data output by the biomarker screening model, feature screening is performed, and a preset screening strategy is used to perform secondary screening on the features after feature screening to obtain the target biomarker panel.

[0031] In an embodiment, the preset screening strategy includes one or more of the following: Cumsum Importance, Recursive Feature Elimination Cross-Validation (REFCV), Correlation Coefficient, and Cumsum AUC. In the embodiment, the biomarker screening model is first screened based on the feature weight value of the substance, and the greater the feature weight value, the greater the contribution of the feature in the group. The features with smaller correlation (i.e., smaller feature weight values) are filtered out.

[0032] Further, after screening using the combination strategy of Cumsum Importance, Recursive Feature Elimination Cross-Validation, Correlation Coefficient (eliminating highly correlated markers), and Cumsum AUC, a set of target biomarker panels with high information and the fewest number of markers (i.e., the best biomarker panel) is obtained, as shown in FIG. 2. Figure 2

[0033] Cumsum Importance includes: first, sorting the features after feature screening according to the feature weight (or according to the original order), and calculating the cumulative weight sequence, for example: feature A weight 0.4, feature B weight 0.3, feature C weight 0.2, then feature A cumulative weight 0.4, feature B cumulative weight 0.7, feature C cumulative weight 0.9; then according to the pre-set cumulative weight threshold, selecting all features whose cumulative weight first exceeds the threshold, or using random sampling method to generate a random number between 0 and 1 r , finding the first cumulative weight greater than or equal to the random number r , and repeating multiple times to count the frequency of each feature being selected, and determining the final screened features according to the frequency.

[0034] REFCV includes: first determining the recursive feature elimination cross-validation model and the cross-validation strategy, and then inputting the feature data screened by Cumsum Importance into the model to automatically calculate the optimal feature subset.

[0035] Correlation Coefficient includes: first calculating the correlation coefficient matrix between features or between features and target variables, and then retaining features with absolute correlation coefficients higher than the threshold.

[0036] Cumsum AUC includes: first calculating the AUC value of each feature, and arranging the features in descending order of AUC value, and then selecting the top N features or features meeting the threshold according to actual needs.

[0037] ​The biomarker screening method provided by the application can improve the data heterogeneity problem of the multi-modal biological sample data by performing preliminary screening and preprocessing on the multi-modal biological sample data, select the best model as the biomarker screening model from the multiple models, and perform secondary screening by using the output result of the biomarker screening model and using the cumulative weight value screening method, the cross recursive feature elimination method, the correlation coefficient screening method and the cumulative feature AUC screening method, thereby effectively reducing the risk of missing weak effect markers and screening out biomarkers with high sensitivity (multiple detected markers) and high specificity (low redundancy).

[0038] In an embodiment, for the foregoing step S101, that is, when the preliminary screening is performed on the multi-modal biological sample data to obtain preliminary screening data, and the preliminary screening data is preprocessed, the following methods can be used, but are not limited to the following methods: First, the statistical test method is used to perform difference analysis on the multi-modal biological sample data to obtain the preliminary screening data. Then, based on the data type of the preliminary screening data, the corresponding filling method is used to fill in the missing values of the preliminary screening data, and the preliminary screening data is standardized.

[0039] In specific implementation, the data type can be protein data or metabolic data. For protein data, the corresponding method is selected according to the data condition to fill in the missing values and then perform standardization processing. For metabolic data, if there is no missing value, the standardization processing is directly performed, and if there is a missing value, the method is automatically selected according to the data condition to fill in the missing values and then perform standardization processing. The filling method can be minimum value filling or KNN filling.

[0040] In an embodiment, for the foregoing step S103, that is, when the features and the corresponding feature weight values of the substances in the preliminary screening data output by the biomarker screening model are screened, the following methods can be used, but are not limited to the following methods: based on the feature weight values output by the biomarker screening model, the features with feature weight values less than a preset weight threshold value are filtered out.

[0041] Referring to FIG. 2, Figure 3 As shown in FIG. 2, the feature weight value indicates the contribution degree of the feature in this group to a certain extent, and in the embodiment of the application, the weight threshold value can be preset, and then the features with feature weight values less than the weight threshold value are filtered out according to the weight threshold value.

[0042] In an embodiment, the acquired multi-modal biological sample data can be divided into a training set and a test set according to a certain proportion, after the best model is obtained by training the training set, and the target biomarker panel is obtained by using the best model, the test set can also be used to analyze the target biomarker panel. Specifically, the above method further comprises: first, based on the biomarker screening model and the target biomarker panel, the ROC curve of the biological sample data in the test set is drawn; then the area under the ROC curve is calculated, and the target biomarker panel is evaluated based on the area.

[0043] In specific implementation, the ROC curve of the test set is drawn based on the screened biomarker screening model and the target biomarker panel. The area under the ROC curve is called Area Under Curve (AUC), and AUC can be used to evaluate the sensitivity and specificity of biomarkers for predicting event occurrence. The sensitivity and specificity of each metabolite are determined by the optimal threshold of the ROC curve. The closer the AUC value is to 1, the higher the prediction accuracy. Therefore, the prediction accuracy of the target biomarker panel can be evaluated according to the area under the ROC curve in the embodiment of the application.

[0044] In an embodiment, the above method further comprises: performing marker evaluation and marker verification on each marker in the target biomarker panel; wherein the marker evaluation comprises: performing marker evaluation by box plot or correlation analysis; and the marker verification comprises: verifying the correctness of the target biomarker panel by using one or more of decision tree, support vector machine, random forest, K- nearest neighbor method, and naive Bayes method.

[0045] In specific implementation, when performing marker evaluation, the expression amount of the characteristic substance of the target biomarker panel in different groups can be displayed by box plot. The correlation between each marker included in the diagnostic panel can also be evaluated by correlation analysis. Generally, the lower the correlation, the less common information between the selected markers, and the simpler the diagnostic panel, and the target biomarker panel is the combination of this panel.

[0046] When validating biomarkers, decision trees (DT), support vector machines (SVM), random forests (RF), K-nearest neighbors (KNN), and naive bayes (NB) can be used to measure the ability of candidate biomarkers to distinguish different groups. Accuracy, AUC, recall, precision, F1 score, and Kappa coefficient are used to evaluate the classification results of the model. The closer the index is to 1, the better the selected model is, and the more accurate the results obtained by the model are.

[0047] In one implementation, after evaluating and validating each biomarker in the target biomarker panel, the method further includes: using a logistic regression algorithm to calculate the regression coefficients and intercepts of each biomarker in the target biomarker panel; generating a regression equation based on the regression coefficients and intercepts, constructing a diagnostic model, and evaluating the performance of the diagnostic model.

[0048] In practical implementation, Logistic Regression (LR) is a simple and efficient classification algorithm, mainly used to solve binary classification problems to reveal the probability of an event occurring. Regression equations are generated using the regression coefficients and intercepts of the markers to obtain a diagnostic model for subsequent diagnostic analysis. To better apply the diagnostic model to classification, determining the optimal diagnostic threshold is crucial. In this embodiment, the Youden's Index can be used to determine the optimal threshold; generally, a higher value indicates a better threshold.

[0049] Furthermore, the effectiveness of the diagnostic model can be assessed by plotting ROC curves; see [link to relevant documentation]. Figure 4 The ROC curve diagram shown indicates that the AUC of both the training and test sets is above 0.90, which means that the selected target biomarker panel diagnostic model has excellent classification performance and can be used as a diagnostic tool for practical applications. The optimal diagnostic threshold for the training set is the threshold when the Youden exponent is at its maximum by default.

[0050] For ease of understanding, this invention also provides a specific flowchart of the biomarker selection process, see [link to flowchart]. Figure 5 As shown, biological samples are first obtained, and the samples are standardized and preliminarily screened. Then, biomarkers are screened through machine learning models, and the best model is selected from four major machine learning models to obtain the target biomarker panel. After that, biomarker evaluation and validation are carried out, and a diagnostic model is constructed and validated.

[0051] The method provided by the embodiment of the present application shortens the analysis period, improves the problem of missing markers caused by single model or optimal selection in the traditional strategy, improves the number of markers detected, and realizes automatic modeling analysis and screening of markers.

[0052] For the biomarker screening method provided by the foregoing embodiment, the embodiment of the present application further provides a biomarker screening system, which refers to a structure schematic diagram of a biomarker screening system shown in Figure 6 The system mainly includes the following parts: The preliminary screening module 601 is configured to perform preliminary screening on the multi-modal biological sample data to obtain preliminary screening data, and perform preprocessing on the preliminary screening data. The model training module 602 is configured to train a plurality of machine learning models based on the preprocessed preliminary screening data, and select an optimal model from the plurality of machine learning models as a biomarker screening model. The marker screening module 603 is configured to perform feature screening on the features of the substances in the preliminary screening data output by the biomarker screening model and the corresponding feature weight values, and perform secondary screening on the features after the feature screening by using a preset screening strategy to obtain a target biomarker panel. The preset screening strategy includes one or more of the following: cumulative weight value screening method, cross-recursive feature elimination method, correlation coefficient screening method, and cumulative feature AUC screening method.

[0053] The biomarker screening system provided by the present application can improve the problem of data heterogeneity existing in the multi-modal biological sample data by performing preliminary screening and preprocessing on the multi-modal biological sample data. By training a plurality of models and selecting an optimal model as a biomarker screening model, and performing feature screening by using the output result of the biomarker screening model and performing secondary screening by using the cumulative weight value screening method, the cross-recursive feature elimination method, the correlation coefficient screening method, and the cumulative feature AUC screening method, the risk of missing weak effect markers can be effectively reduced, and biomarkers with high sensitivity (multiple detected markers) and high specificity (low redundancy) can be screened out.

[0054] In one embodiment, the preliminary screening module 601 is further configured to perform difference analysis on the multi-modal biological sample data by using a statistical test method to obtain preliminary screening data, perform missing value filling on the preliminary screening data by using a corresponding filling method based on the data type of the preliminary screening data, and perform standardization processing on the preliminary screening data.

[0055] In an implementation, the model training module 602 is further configured to train a plurality of machine learning models based on the preprocessed screening data, and calculate evaluation indexes of each machine learning model by using a K-fold cross-validation strategy; wherein the evaluation indexes include accuracy, AUC value, recall rate, precision, F1 value and Kappa coefficient; and select an optimal model from the plurality of machine learning models as the biomarker screening model based on the evaluation indexes.

[0056] In an implementation, the biomarker screening module 603 is further configured to filter out features with feature weight values less than a preset weight threshold based on the feature weight values output by the biomarker screening model.

[0057] In an implementation, the biomarker screening module 603 is further configured to draw a ROC curve of the biological sample data in the test set based on the biomarker screening model and the target biomarker panel; calculate an area under the ROC curve, and evaluate the target biomarker panel based on the area.

[0058] In an implementation, the device further includes a biomarker evaluation module configured to perform biomarker evaluation and biomarker verification on each biomarker in the target biomarker panel; wherein the biomarker evaluation includes biomarker evaluation by box plot or correlation analysis; and the biomarker verification includes verifying the correctness of the target biomarker panel by using one or more of decision tree, support vector machine, random forest, K-nearest neighbor method and Naive Bayes method.

[0059] In an implementation, the device further includes a diagnostic model construction module configured to calculate regression coefficients and intercepts of each biomarker in the target biomarker panel by using a logistic regression algorithm; generate a regression equation based on the regression coefficients and the intercepts, construct a diagnostic model, and perform performance evaluation on the diagnostic model.

[0060] It should be noted that the system provided by the embodiments of the present application has the same implementation principle and technical effects as the foregoing method embodiments, and for brevity, the parts not mentioned in the system embodiment part can be referred to the corresponding contents in the foregoing method embodiments. The specific numerical values provided in the embodiments of the present application are only exemplary and are not limited herein.

[0061] The embodiments of the present application further provide an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program performs the method described in any one of the above embodiments when executed by the processor.

[0062] Figure 7A structural schematic diagram of an electronic device provided by the embodiment of the present application is shown in the figure, which comprises a processor 70, a memory 71, a bus 72 and a communication interface 73, and the processor 70, the communication interface 73 and the memory 71 are connected through the bus 72; the processor 70 is used for executing the executable modules stored in the memory 71, such as computer programs.

[0063] The memory 71 can contain a high-speed random access memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 73 (which can be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. can be used.

[0064] The bus 72 can be an ISA bus, a PCI bus or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 7 Only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0065] The memory 71 is used for storing programs, and the processor 70 executes the programs after receiving execution instructions. The method executed by the device defined by the flow process disclosed in any of the foregoing embodiments of the present application can be applied to the processor 70 or realized by the processor 70.

[0066] The processor 70 can be an integrated circuit chip with signal processing capability. In implementation, each step of the above method can be completed by integrated logic circuit of hardware in the processor 70 or by instructions in the form of software. The processor 70 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. Each method, step and logic block diagram disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium in the art. The storage medium is located in the memory 71, and the processor 70 reads the information in the memory 71, and combines the hardware to complete the steps of the above method.

[0067] The computer program product of the readable storage medium provided by the embodiments of the present application comprises a computer readable storage medium storing program codes, and the program codes comprise instructions for executing the method described in the foregoing method embodiments. For specific implementation, reference can be made to the foregoing method embodiments, which will not be described here.

[0068] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the present application that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of software products. The computer software product is stored in a storage medium and includes instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0069] Finally, it should be noted that: the above-described embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit them. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily think of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed by the present application, or make equivalent replacements to some of the technical features. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for screening biomarkers, characterized in that, include: Preliminary screening data is obtained by initially screening multimodal biological sample data, and the preliminary screening data is then preprocessed. Multiple machine learning models are trained based on the preprocessed initial screening data, and the best model is selected from the multiple machine learning models as the biomarker screening model. Based on the characteristics of substances and corresponding feature weight values ​​in the initial screening data output by the biomarker screening model, feature screening is performed, and a preset screening strategy is used to perform secondary screening on the features after feature screening to obtain the target biomarker panel; wherein, the preset screening strategy includes one or more of the following: cumulative weight value screening method, cross-recursive feature elimination method, correlation coefficient screening method, and cumulative feature AUC screening method.

2. The method according to claim 1, characterized in that, Preliminary screening of multimodal biological sample data yields initial screening data, which is then preprocessed, including: Statistical tests were used to analyze the differences in the multimodal biological sample data to obtain preliminary screening data. Based on the data type of the initial screening data, the missing values ​​of the initial screening data are filled using the corresponding filling method, and the initial screening data is standardized.

3. The method according to claim 1, characterized in that, Multiple machine learning models are trained based on the preprocessed initial screening data, and the best model is selected from these models as the biomarker screening model, including: Multiple machine learning models are trained based on the preprocessed initial screening data, and the evaluation metrics for each machine learning model are calculated using the K-fold cross-validation strategy; wherein, the evaluation metrics include: accuracy, AUC value, recall, precision, F1 score and Kappa coefficient; The best model is selected from multiple machine learning models based on the evaluation metrics as the biomarker screening model.

4. The method according to claim 1, characterized in that, Feature screening is performed based on the characteristics of substances and their corresponding feature weights in the initial screening data output by the biomarker screening model, including: Based on the feature weight values ​​output by the biomarker screening model, features whose feature weight values ​​are less than a preset weight threshold are filtered out.

5. The method according to claim 1, characterized in that, After performing secondary screening on the features selected using a preset screening strategy to obtain the target biomarker panel, the process also includes: Based on the biomarker screening model and the target biomarker panel, ROC curves of biological sample data in the test set were plotted. Calculate the area under the ROC curve and evaluate the target biomarker panel based on the area.

6. The method according to claim 1, characterized in that, Also includes: Each biomarker in the target biomarker panel is evaluated and validated; wherein, the biomarker evaluation includes evaluating the biomarkers through box plots or correlation analysis; the biomarker validation includes validating the correctness of the target biomarker panel using one or more of the following methods: decision tree, support vector machine, random forest, K-nearest neighbor method, and Naive Bayes method.

7. The method according to claim 6, characterized in that, After evaluating and validating each biomarker in the target biomarker panel, the process further includes: The regression coefficients and intercepts of each biomarker in the target biomarker panel were calculated using a logistic regression algorithm. A diagnostic model is constructed based on the regression coefficients and the intercept, and the performance of the diagnostic model is evaluated.

8. A biomarker screening system, characterized in that, include: The preliminary screening module is used to perform preliminary screening on multimodal biological sample data to obtain initial screening data, and to preprocess the initial screening data; The model training module is used to train multiple machine learning models based on the preprocessed initial screening data, and select the best model from the multiple machine learning models as the biomarker screening model. The biomarker screening module is used to perform feature screening based on the characteristics of substances and corresponding feature weight values ​​in the initial screening data output by the biomarker screening model, and to perform secondary screening on the features after feature screening using a preset screening strategy to obtain the target biomarker panel; wherein, the preset screening strategy includes one or more of the following: cumulative weight value screening method, cross-recursive feature elimination method, correlation coefficient screening method, and cumulative feature AUC screening method.

9. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the steps of the method described in any one of claims 1 to 7.

Images