Construction method and application of influenza susceptibility markers and influenza high-risk population prediction model based on the markers

By combining differential expression and co-expression network analysis with a random forest model, a prediction model for high-risk influenza populations was constructed. This solved the problem of insufficient identification of influenza susceptibility in healthy adults, enabling early and accurate screening and intervention, and reducing the risk of influenza transmission.

CN116344067BActive Publication Date: 2024-11-19SUN YAT SEN UNIVERSITY SHENZHEN +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310108655.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-01-17
Publication Date
2024-11-19
Estimated Expiration
2043-01-17

AI Technical Summary

Technical Problem

Existing technologies lack effective identification and prediction of influenza susceptibility in healthy adults, especially due to insufficient transcriptional studies at the baseline level. This makes it difficult to accurately identify high-risk groups, increasing the spread and harm of influenza.

Method used

By screening baseline genetic data of influenza patients, differential expression analysis and co-expression network analysis were performed. Combined with a random forest model, a predictive model for high-risk influenza populations was constructed. Using a combination of biomarkers 'AMFR', 'HBQ1', 'DHRS9', 'SLC35E2A', 'BANK1', 'CD79A', 'TXNDC5', 'H2BC5', and 'PRKY', the predictive model for high-risk influenza populations achieved an AUC of 0.94, demonstrating high accuracy.

Benefits of technology

It enables the early detection of high-risk groups for influenza, allowing for effective intervention measures to reduce the risk of influenza transmission and health hazards. It has high accuracy and flexibility and is suitable for influenza susceptibility screening in healthy adult populations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116344067B_ABST
    Figure CN116344067B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for constructing and applying an influenza susceptibility marker and an influenza high-risk population prediction model based on the marker. The influenza susceptibility marker is a combination of "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY". The marker and the influenza risk prediction model established based on the marker are obtained based on the data of the baseline transcriptome of healthy adults, and can be used to screen influenza A H3N2 susceptible individuals in healthy adults, with an AUC of 0.94, an accuracy of 0.91, a precision of 1, a recall rate of 0.75, and an F1 score of 0.86, which have good accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of molecular biology, and specifically relates to a method for constructing and applying an influenza susceptibility marker and an influenza high-risk population prediction model based on the marker. Background Art

[0002] Influenza viruses regularly cause seasonal epidemics in humans and animals, and occasionally cause severe epidemics. Globally, there are approximately 3 million to 5 million severe cases each year, and approximately 290,000 to 650,000 people die from respiratory diseases caused by seasonal influenza, causing serious public health problems worldwide. There is an urgent need to develop more effective drugs, vaccines, and related protective measures to curb the spread, prevalence, and outbreaks of influenza viruses. Therefore, the development and design of more effective and targeted drugs, vaccines, and related protective measures is a top priority in current influenza virus research.

[0003] Influenza viruses have a wide range of infections and can infect people of any age. However, people exposed to the same influenza virus environment often have different outcomes. For example, some people have mild symptoms, some have severe symptoms, but some are asymptomatic. This shows that individuals have different susceptibility to influenza. Differences in influenza susceptibility lead to different risks of illness among people. Accurate identification of high-risk groups is crucial for accurate prevention and treatment of influenza and the development of new drugs and vaccines, which will help reduce the disease burden of influenza in humans.

[0004] In the related art, it is found that the reasons for the difference in host susceptibility to influenza are very complex, including factors from viruses, hosts and the environment, such as contact history, genetic factors, age, nutrition, etc. However, there are relatively few studies on influenza susceptibility at the transcriptional level. In the related art, the main focus is on the differences in transcriptional characteristics between susceptible and non-susceptible people after infection, and the classification models mostly rely on the characteristics after infection, but lack the identification of high-risk people before infection. The identification of high-risk people is more meaningful and has practical application value in capturing the characteristics of the body before viral infection, that is, focusing on the differences and similarities between susceptible and non-susceptible people at the baseline level. In addition, most of the current research focuses on the susceptibility research of special populations such as the elderly, children, and pregnant women, and there are very few susceptibility studies on healthy adults. However, healthy adults account for a considerable proportion of the total population, and they have high mobility and a wide range of activities. In addition, healthy adults play an important role as pillars in family, work and society, and are the main labor force for social production. These reasons make healthy adults more likely to spread and cause greater harm once they get influenza.

[0005] Therefore, the present invention will focus on the influenza susceptibility of healthy adults, and strive to find genes related to susceptibility at the baseline transcription level, and use these genes to construct a predictive model that can perform early risk screening for healthy people, so as to overcome the research gap in this aspect in the prior art. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the above-mentioned prior art. To this end, the present invention proposes a group of influenza susceptibility markers and a method and application of constructing a prediction model for influenza high-risk population based on the markers. The present invention screens out symptomatic and asymptomatic subject data from the genetic baseline data of influenza patients, performs differential expression analysis and co-expression network analysis respectively, and successfully constructs a prediction model that can accurately analyze influenza high-risk populations in combination with a random forest model. The AUC can reach 0.94, with a high accuracy rate, so that influenza high-risk populations can be discovered at a very early stage, intervention measures or treatment methods can be effectively implemented, and the occurrence and spread of influenza can be accurately controlled, reducing its harm to human health.

[0007] The first aspect of the present invention provides a group of influenza susceptibility markers, wherein the influenza susceptibility markers are "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY".

[0008] In some embodiments of the invention, the influenza comprises H3N2.

[0009] In some embodiments of the invention, the influenza is H3N2.

[0010] The second aspect of the present invention provides a method for constructing a prediction model for high-risk populations of influenza, comprising the following steps:

[0011] (1) Collect genetic data from subject samples and screen for influenza-related baseline data for use;

[0012] (2) Screening out the data of symptomatic and asymptomatic subjects from the baseline data, performing differential expression analysis and co-expression network analysis respectively, and selecting the intersection genes after differential expression analysis and co-expression network analysis as candidate genes;

[0013] (3) Based on candidate genes, a random forest was used to construct a prediction model for high-risk populations for influenza.

[0014] Among them, the candidate genes include the influenza susceptibility markers described in the first aspect of the present invention, specifically a combination of "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY".

[0015] In some embodiments of the present invention, the parameter pfp≤0.05 is set in the differential expression analysis.

[0016] In some embodiments of the present invention, the step of co-expression network analysis comprises:

[0017] (1) The samples were clustered using the Hclust function hierarchical clustering method in R;

[0018] (2) Calculate the soft threshold power β by selecting the soft threshold using the function in the WGCNAR package, calculate the Pearson correlation, and use the Pearson correlation results to construct a weighted adjacency matrix with the soft threshold power β;

[0019] (3) The weighted adjacency matrix is ​​converted into a topological overlap measure, and average linkage hierarchical clustering is performed based on the difference metric of the topological overlap measure, so that genes with similar expression patterns can be grouped into the same module.

[0020] In some embodiments of the present invention, the threshold β is set to 6-8.

[0021] In some embodiments of the present invention, the threshold β is set to 6, R 2 It can reach 0.9.

[0022] In the present invention, after testing, it is found that when the threshold β is set to 6, the average connectivity is relatively high.

[0023] In some embodiments of the invention, the topological overlap measure (TOM) is used to estimate the network connectivity of genes therein.

[0024] In some embodiments of the present invention, the relevant parameters of the average linkage hierarchical clustering are set as: TOMType=“unsigned”, min Module Size=15, reassign Threshold=0, merge Cut Height=0.25, deepsplit=2.

[0025] In the present invention, after testing, a total of 18 modules were obtained, of which 5 modules were closely related to susceptibility and non-susceptibility, specifically: "MECyan", "MEMidnightblue", "MEBlue", "MEGreenyellow", "MEGrey60", totaling 466 genes.

[0026] In some embodiments of the present invention, the differential expression analysis (DEGs) is based on the method of RankProd (version 3.18.0) to determine the differentially expressed genes (DEGs) between asymptomatic and symptomatic hosts. In the present invention, a total of 223 differentially expressed genes were screened out from 8478 genes.

[0027] In the present invention, the term "differentially expressed genes (DEGs)" refers to genes that have significant differences in expression under different states or conditions at the RNA level.

[0028] In some embodiments of the present invention, the co-expression network analysis is performed using the R package WGCNA to construct a co-expression network using the top 20% variant genes (1696 genes) among 8478 genes.

[0029] In some embodiments of the present invention, the specific operation of selecting the intersection genes after differential expression analysis and co-expression network analysis as candidate genes is: using module screening as the basis for screening representative genes (candidate genes), and module screening is specifically to randomly select two genes in a module, and use the expression profiles of the selected 2 genes and all genes in the other 2 modules to construct a random forest model, and use five-fold cross validation to evaluate the accuracy of the model. Test the random forest model composed of all 2 genes in the module and genes in the other 2 modules, compare their accuracy, and select the model with the highest accuracy in the module. The 2 genes constituting the model are the 2 genes selected from the module with the highest correlation with influenza susceptibility.

[0030] In some embodiments of the present invention, candidate genes can also be selected using machine learning methods such as variance threshold, SelectKBesT, recursive feature elimination, SelectFromModel feature selection, and sequential feature selection.

[0031] In some embodiments of the present invention, the method for constructing the influenza high-risk population prediction model also includes debugging the parameters of the model by further manual parameter adjustment, grid parameter adjustment, etc. to optimize the prediction model.

[0032] In some embodiments of the invention, the influenza comprises H3N2.

[0033] In some embodiments of the invention, the influenza is H3N2.

[0034] In some embodiments of the present invention, the method for constructing the influenza high-risk population prediction model is specifically:

[0035] (1) Microarray data related to influenza A / H3N2 based on influenza challenge experiments were collected from the Gene Expression Omnibus (GEO) database, and the microarray data of the baseline (day 0) data set was extracted, which included a total of 30 symptomatic hosts and 19 asymptomatic hosts. After data integration and preprocessing, the expression profiles of 8478 genes in 49 samples (30 symptomatic hosts and 19 asymptomatic hosts) were finally obtained;

[0036] (2) A preliminary screening of the expression profiles of 8478 genes was performed using biological analysis methods. Genes with differential expression differences between symptomatic and asymptomatic hosts were screened out through differential expression analysis, and genes in the co-expression module associated with seasonal influenza A / H3N2 susceptibility were obtained through co-expression network analysis. Based on the results of comprehensive gene analysis of differentially expressed genes and co-expression modules, genes that were differentially expressed between symptomatic and asymptomatic hosts after seasonal influenza A / H3N2 infection and belonged to the co-expression module associated with seasonal influenza A / H3N2 susceptibility were extracted as the results of the preliminary screening;

[0037] (3) Use the co-expression module-based screening method to further screen candidate genes: In each co-expression module, use the random forest model to further screen the two genes that are most correlated with influenza susceptibility in each co-expression module. By combining the results of each module, the candidate genes with the highest correlation with influenza susceptibility can be screened.

[0038] (4) Based on the obtained candidate gene expression profile data, a classification model is constructed using the random forest method to predict whether an individual is susceptible.

[0039] In the present invention, the prediction model is based on the baseline blood transcriptome of healthy people. By screening influenza susceptibility-related genes, a targeted gene-based influenza susceptibility prediction model for the general population is constructed. The data basis and judgment criteria of the influenza risk assessment model in the prior art are completely different, and it is the only prediction model for influenza susceptibility for the general population.

[0040] Early screening is extremely important for the control of influenza. For individuals with a high risk of developing the disease, targeted measures can be taken to target high-risk factors for influenza A infection, especially environmental factors, to conduct more precise preventive interventions. The key prevention strategy for high-risk groups is more in line with the principle of cost-effectiveness, which can maximize the benefits of limited medical resources and reduce the waste of medical resources and social economy to a great extent. At the same time, by blocking the infection of individuals who are highly susceptible to influenza A, multiple transmission chains of influenza are cut off, protecting other individuals who may be infected with influenza. It can further reduce the risk of influenza A H3N2 in the population, reduce the disease burden of influenza, and protect the health of the population.

[0041] In the present invention, the term "susceptibility" refers to the risk of an individual getting a disease, that is, the risk of different individuals getting a disease under the same environment may be related to multiple factors such as genetics and environment.

[0042] The third aspect of the present invention provides use of a product for quantitatively detecting the influenza susceptibility marker described in the first aspect of the present invention in preparing a product for predicting a high-risk population for influenza.

[0043] In some embodiments of the present invention, the influenza high-risk population prediction product includes a detection reagent, a detection kit, a detection chip, and a risk prediction system.

[0044] In some embodiments of the present invention, in the risk prediction system, the product for quantitatively detecting the influenza susceptibility markers described in the first aspect of the present invention is used as the detection end, and the data obtained by the detection end is transmitted to the analysis end carrying the influenza high-risk population prediction model of the present invention for analysis, so as to obtain the prediction result, and the population determination result is directly outputted based on the prediction result.

[0045] In some embodiments of the invention, the influenza comprises H3N2.

[0046] In some embodiments of the invention, the influenza is H3N2.

[0047] In some embodiments of the present invention, the method for using the influenza high-risk population prediction product includes the following steps: quantifying the content of influenza susceptibility markers in the sample to be tested, bringing it into the prediction model, and determining whether it belongs to the influenza high-risk population based on the results obtained by the prediction model.

[0048] In some embodiments of the present invention, the prediction model is a random forest classification model characterized by genes "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY". The model can determine whether the patient is at high risk of influenza by simply inputting the expression levels of these genes in the patient's blood.

[0049] In some embodiments of the present invention, the criteria for determining a high-risk group for influenza are: if the model output is 1, then the group is determined to be a high-risk group for influenza; if the model output is 0, then the group is determined to be a low-risk or no-risk group for influenza.

[0050] In the present invention, the present invention aims to provide a group of genes associated with influenza susceptibility, so that blood transcriptome screening based on these genes at the baseline level can predict the risk of healthy adults to seasonal influenza A / H3N2 through a model.

[0051] The beneficial effects of the present invention are:

[0052] 1. The present invention only uses the data of the baseline transcriptome of healthy adults to discover a marker combination associated with influenza susceptibility: "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY", and constructs a random forest classification model based on the markers for screening influenza A H3N2 susceptible individuals in healthy adults with good accuracy; and in the screening of the markers, the biological significance is fully considered, and it is also considered that the genes in the same module have similar expression patterns and the current large number of input features will easily lead to the problem of overfitting caused by too much input data, thereby achieving effective optimization.

[0053] 2. The construction method of the present invention is based on baseline transcription information, screens genes related to influenza susceptibility, and constructs a random forest model to predict the susceptibility of healthy adults to seasonal influenza A / H3N2. The method is simple and convenient, and the obtained model has high sensitivity; moreover, the method focuses on the differences in differentially expressed genes and co-expression modules, and the obtained model can be adjusted and optimized according to actual conditions, and preliminary screening can be performed or not performed according to other principles, which has high flexibility.

[0054] 3. The model in the present invention is based on the gene expression at the baseline level of the transcriptome of healthy adults to predict the susceptibility of individuals to influenza A virus H3N2. It can be widely used in the society for screening the susceptibility of the population to influenza A without population limitation. The AUC of the model is 0.94, the accuracy is 0.91, the precision is 1, the recall rate is 0.75, and the F1 score is 0.86, indicating that it has a good prediction effect.

[0055] 4. The model in the present invention uses peripheral blood transcriptome data to predict individuals who are susceptible to influenza in healthy adults. Obtaining peripheral blood tissue is a relatively non-invasive medical examination method, which is highly accepted by the public and has a high degree of feasibility. It can also timely detect and screen out people who are highly susceptible to influenza in the healthy population, thereby determining the high-risk population for influenza A in terms of individual factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is the ROC curve obtained by using the prediction model of the present invention with the independent dataset GSE61754 as the validation set. DETAILED DESCRIPTION

[0057] In order to make the invention purpose, technical solution and technical effect of the present invention clearer, the present invention is further described in detail below in conjunction with specific implementation methods. It should be understood that the specific implementation methods described in this specification are only for explaining the present invention, not for limiting the present invention.

[0058] Unless otherwise specified, the experimental materials and reagents used are consumables and reagents that can be obtained from conventional commercial channels.

[0059] Construction of a prediction model for high-risk populations of influenza

[0060] In this example, seasonal influenza A / H3N2 is used as an example to demonstrate how to predict the susceptibility of healthy adults to influenza using baseline transcription information through a model, which specifically includes the following steps:

[0061] 1. Data Collection:

[0062] Three microarray data sets were collected from the Gene Expression Omnibus (GEO): GSE73072_DEE2, GSE73072_DEE5, and GSE61754. The above three microarray data sets are all from human influenza challenge experiments. The experiment is based on blood samples collected from volunteers at baseline and several other time points after vaccination with A / H1N1 or A / H3N2 influenza virus, and the data obtained after sequencing the collected samples at the transcriptional level. For details, see Table 1. In this embodiment, only the baseline data related to A / H3N2 influenza are extracted for use.

[0063] Based on the symptom status of influenza laboratory testing and self-reported modified Jackson scores, the individuals in the above data set were defined as hosts with symptoms (representing susceptible individuals) and asymptomatic (representing non-susceptible individuals). Specifically, the data set used included a total of 49 subjects, including 30 symptomatic hosts and 19 asymptomatic hosts.

[0064] Table 1. Flu challenge experiment dataset information

[0065] GEO ID Virus Asymptomatic Have symptoms chip References GSE61754 H3N2 3 8 Illumina Emma E et al.2015 GSE73072_DEE2 H3N2 8 9 Affymetrix Yu Liu, et al. 2016 GSE73072_DEE5 H3N2 8 13 Affymetrix Yu Liu, et al. 2016

[0066] 2. Data preprocessing:

[0067] In order to eliminate the influence of experimental platform and processing method on the data, the data sets in the above three microarray experiments were preprocessed. The specific steps are as follows: First, after annotating the array probes, all data were merged into a single matrix, in which rows represent genes and columns represent samples, and the matrix was logarithmically transformed. After retaining the gene information in all three data, the final gene expression matrix contained a total of 8478 genes from 49 samples. The obtained gene information was normalized using the "normalized between arrays" method in the R package limma (version 3.48.3.). Then, the batch effect was removed using the ComBat method in the R package sva with default parameters (version 3.40.0). Finally, the datasets GSE73072_DEE2 and GSE73072_DEE2 were selected as the training set (containing a total of 38 samples, including 16 asymptomatic samples and 22 symptomatic samples), and the independent dataset GSE61754 was used as the independent validation set (containing 3 asymptomatic samples and 8 symptomatic samples) to evaluate the model.

[0068] 3. Initial screening of susceptibility-related genes:

[0069] In the embodiment of the present invention, the screening of susceptibility-related genes is based on a bioinformatics analysis method, specifically, preliminary screening of pre-processed gene expression data using differentially expressed genes and co-expression network analysis.

[0070] (1) Differential expression analysis (DEGs):

[0071] Differentially expressed genes (DEGs) between asymptomatic and symptomatic hosts were determined using the RankProd (version 3.18.0) method (percentage of false predictions (pfp) <= 0.05).

[0072] In this example, a total of 223 differentially expressed genes were screened out from 8478 genes.

[0073] (2) Co-expression network analysis:

[0074] The co-expression network was constructed using the R package WGCNA using the top 20% variant genes (1696 genes) among the 8478 genes.

[0075] The specific steps are:

[0076] In order to test whether there are outliers, the samples are first clustered using the Hclust function hierarchical clustering method in R. Then the soft threshold is selected using the function in the WGCNAR package to calculate the soft threshold power β. The threshold β is set to 6, and R 2 It can reach 0.9, and the average connectivity is relatively high, which meets the standards of scale-free networks. Pearson correlation was calculated, and the Pearson correlation results were used to construct a weighted adjacency matrix with soft threshold power β. The weighted adjacency matrix was converted into a topological overlap measure (TOM). The obtained TOM can be used to estimate the network connectivity of any gene. Average linkage hierarchical clustering was performed based on the difference metric of TOM, and genes with similar expression patterns were divided into modules. Among them, the relevant parameters of average linkage hierarchical clustering were set as: TOM Type = "unsigned", min Module Size = 15, reassign Threshold = 0, merge Cut Height = 0.25, deepsplit = 2.

[0077] Finally, 18 modules were obtained, including 5 modules closely related to susceptibility and non-susceptibility, namely "MECyan", "MEMidnightblue", "MEBlue", "MEGreenyellow", and "MEGrey60", with a total of 466 genes.

[0078] (3) Preliminary screening of susceptibility-related genes:

[0079] Through the steps in (1) and (2) above, we found modules related to influenza susceptibility based on the co-expression network, and these modules contain potential genes related to influenza susceptibility. In addition, differential expression analysis also found genes that are expressed differently in symptomatic and asymptomatic groups at baseline. By comparing the genes related to influenza susceptibility found in the two, only genes that are differentially expressed at the same time are more meaningful for distinguishing between influenza susceptible and non-susceptible populations. Finally, by comparing the gene sets of the five significant co-expression modules obtained after comparison analysis and the intersection genes between DEGs, a total of 119 genes were selected as candidate genes for the next step of model construction.

[0080] 4. Module-based susceptibility gene screening:

[0081] In order to further screen genes related to influenza susceptibility while avoiding bias and potential overfitting caused by a single data set, considering that genes in the same module have similar expression patterns and similar functions, the present invention adopts module-based screening of representative genes.

[0082] The specific screening method is: randomly select two genes in a module, use the expression profiles of the selected two genes and all genes in the other two modules to build a random forest model, and use five-fold cross validation to evaluate the accuracy of the model. Test the random forest model composed of all two genes in this module and genes in the other two modules, compare their accuracy, and select the model with the highest accuracy in this module. The two genes that make up this model are the two genes selected from this module that are most correlated with influenza susceptibility.

[0083] By repeating the above screening operation in each co-expression module, the 9 genes with the highest correlation were finally selected from the 5 co-expression modules as the final prediction model features. The 9 important genes related to the susceptibility of seasonal influenza A virus A / H3N2 were finally obtained: "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5", and "PRKY".

[0084] We further used manual parameter adjustment and grid parameter adjustment to debug the model parameters, optimize the prediction model, and found the model with the highest prediction accuracy.

[0085] The resulting model code is:

[0086]

[0087]

[0088]

[0089] The prediction model is a random forest classification model composed of genes "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY". By simply inputting the expression levels of these genes in the patient's blood, the model can determine whether the patient is at high risk of influenza. If the model output is 1, it is determined to be a high-risk group for influenza; if the model output is 0, it is determined to be a low-risk or no-risk group for influenza.

[0090] Effect verification

[0091] The prediction effect was verified using the random forest prediction model based on 9 important genes related to the susceptibility of seasonal influenza A virus A / H3N2 obtained in the above example.

[0092] The specific steps are:

[0093] The independent data set GSE61754 was used as the validation set. Based on the 9 important genes screened in the above steps, the expression data corresponding to these genes were extracted from the gene expression profile and a random forest classification prediction model was constructed in the training set using the R package randomForest according to the method in the above embodiment.

[0094] The independent dataset GSE61754 was used to evaluate the model effect. The accuracy, precision, recall, F1 score, and area under the curve (AUC) were calculated. The corresponding ROC curve was also drawn. The results are shown in Table 2 and Figure 1 .

[0095] Table 2 Performance of the random forest prediction model in predicting the susceptible population of influenza A / H3N2 on the selected training dataset

[0096] Classification Model accuracy Accuracy Recall F1 Rating AUC Random Forest 0.91 1 0.75 0.86 0.94

[0097] It can be seen from Table 1 that the influenza high-risk population prediction model obtained in the above embodiment can effectively distinguish influenza high-risk populations with high accuracy and strong specificity, and the AUC can reach 0.94.

[0098] Practical Usage

[0099] In actual use, nucleic acid molecules (DNA, commercially available nucleic acid extraction kits or other products can be used) are extracted from the collected subject samples, and the content of the corresponding gene in the subject sample is obtained through nucleic acid quantitative detection technology (such as PCR amplification technology or sequencing technology for the above-mentioned ADORA3, CD36, ANXA3, PPP1R15A, ABHD2, ANKRD11, RPS23, UQCRH, ACOX1, and DSC1), which is then brought into the above-mentioned model. When the model output is 1, the subject is judged to be a high-risk group for influenza; if the model output is 0, the subject is judged to be a low-risk or no-risk group for influenza.

[0100] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A set of influenza susceptibility markers, characterized in that: The influenza susceptibility markers are a combination of "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY"; the influenza is H3N2.

2. A method for constructing a prediction model for high-risk populations of influenza, comprising the following steps: (1) Collect genetic data from subject samples and screen out influenza-related baseline data for use; (2) Screen out the data of symptomatic and asymptomatic subjects from the baseline data, perform differential expression analysis and co-expression network analysis respectively, and select the intersection genes after differential expression analysis and co-expression network analysis as candidate genes; (3) Based on candidate genes, a random forest was used to construct a prediction model for high-risk populations for influenza; in, The candidate gene includes the influenza susceptibility marker according to claim 1; The influenza is H3N2; Wherein, the step of co-expression network analysis comprises: (1) The samples were clustered using the Hclust function hierarchical clustering method in R; (2) Calculate the soft threshold power β by selecting the soft threshold function in the WGCNAR package, calculate the Pearson correlation, and use the Pearson correlation result to construct a weighted adjacency matrix with the soft threshold power β; (3) The weighted adjacency matrix is ​​converted into a topological overlap measure, and average linkage hierarchical clustering is performed based on the difference metric of the topological overlap measure, so that genes with similar expression patterns are grouped into the same module; The parameters of the average linkage hierarchical clustering are: TOM Type="unsigned", min ModuleSize=15, reassign Threshold-0, merge Cut Height- 0.25, deepsplit=2; In the differential expression analysis, genes with a false prediction percentage ≤ 0.05 were used to determine differentially expressed genes between asymptomatic and symptomatic hosts.

3. Use of a product for quantitatively detecting the influenza susceptibility marker of claim 1 in preparing a product for predicting high-risk populations for influenza, wherein the influenza is H3N2.

4. The use according to claim 3, characterized in that: The influenza high-risk population prediction products include detection reagents, detection kits, detection chips, and risk prediction systems.

5. The use according to claim 3, characterized in that: The method for using the influenza high-risk population prediction product comprises the following steps: quantifying the content of influenza susceptibility markers in the sample to be tested, bringing it into the prediction model, and determining whether it belongs to the influenza high-risk population based on the results obtained by the prediction model.

6. The use according to claim 5, characterized in that: The prediction model is a random forest classification model characterized by genes "AMFR", "HBQ1", "DHRS9", "SLC35E2A", "BANK1", "CD79A", "TXNDC5", "H2BC5" and "PRKY". The model can determine whether the patient is at high risk of influenza by simply inputting the expression levels of these genes in the patient's blood.

7. The use according to claim 6, characterized in that: The criteria for determining high-risk groups for influenza are: if the model output is 1, they are determined to be at high risk for influenza; if the model output is 0, they are determined to be at low risk or no risk for influenza.