Intelligent double-disease diagnosis method based on information big data

Abstract

The invention discloses an intelligent double-disease diagnosis method based on information big data, and belongs to the field of disease diagnosis, and the diagnosis method comprises the following specific steps: I, collecting and processing sample data, and screening differential expression genes; iI, constructing a protein interaction network and screening Hub genes based on the network; iII, performing traditional Chinese medicine prediction analysis and tissue specific expression analysis on the Hub gene; iV, constructing a double-disease diagnosis model, and analyzing and applying the double-disease diagnosis model; according to the method, a new thought can be provided for combination of traditional medicine and modern medicine, treatment selection can be expanded, a large amount of gene expression data can be rapidly processed and analyzed, the diagnosis efficiency is remarkably improved, the diagnosis accuracy is enhanced, and the recognition capacity for complex diseases is improved.

Description

Technical Field

[0001] The present invention relates to the field of disease diagnosis, and in particular to an intelligent dual disease diagnosis method based on bioinformatics big data. Background Art

[0002] Sarcopenia is a degenerative syndrome associated with aging, characterized by decreased muscle mass, decreased muscle strength, and (or) gradual decline in physical function. Sarcopenia is associated with a variety of diseases, such as Alzheimer's disease and major depression, and may increase the risk of falls and fractures, seriously affecting the quality of life of patients. It is estimated that there are currently about 50 million people suffering from sarcopenia in the world, and it is expected to reach 500 million by 2050. Alzheimer's disease (AD), a neurodegenerative disease, is characterized by cognitive impairment, mental and behavioral abnormalities, and impaired social life ability. As the most common cause of dementia in the elderly, AD has become the fourth cause of death in the elderly population after heart disease, malignant tumors and stroke, posing a huge threat to the life, health and quality of life of patients. It is estimated that the number of people affected will reach 65.7 million by 2030, and the number of AD patients will increase to 152 million by 2050. More and more clinical studies have found a correlation between the two. For example, a cohort study with a follow-up of 5.6 years in 2021 found that more severe sarcopenia at baseline was associated with a higher risk of AD (HR=1.50, 95%CI:1.20-1.86, P<0.001), and imbalance in myokine secretion may play a role in this. A cross-sectional study in 2023 found that as the stage of AD progressed, the prevalence of sarcopenia increased (OR=0.79, 95%CI:0.69-0.91, P=0.001), and peripheral deposition of amyloid-β (Aβ) outside neurons may be one of the reasons for AD patients suffering from sarcopenia. Although there are many clinical data suggesting that the two are related, and people have done a lot of research on sarcopenia and AD respectively, there is still a lack of basic research directly related to sarcopenia and AD, and the biological connection and potential molecular mechanism between them are still unclear; at the same time, there is a lack of models on the market that can diagnose both at the molecular level; therefore, it is particularly important to invent an intelligent dual disease diagnosis method based on bioinformatics big data.

[0003] The existing intelligent dual-disease diagnosis method based on bioinformatics big data cannot provide new ideas for the combination of traditional medicine and modern medicine, and the processing and analysis of large amounts of gene expression data is slow, which reduces diagnostic efficiency, low diagnostic accuracy, and reduced recognition of complex diseases. To this end, we propose an intelligent dual-disease diagnosis method based on bioinformatics big data. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose an intelligent dual disease diagnosis method based on bioinformatics big data.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent dual disease diagnosis method based on bioinformatics big data, the specific steps of the diagnosis method are as follows:

[0007] Ⅰ. Collect and process sample data and screen differentially expressed genes;

[0008] II. Construct a protein interaction network and screen Hub genes based on the network;

[0009] III. Perform TCM prediction analysis and tissue-specific expression analysis on Hub genes;

[0010] IV. Construct a dual disease diagnosis model and conduct analytical applications.

[0011] As a further solution of the present invention, the specific steps of screening differentially expressed genes in step I are as follows:

[0012] S1.1: Extract the microarray datasets GSE1428 and GSE5281 from the preprocessed sample data, use the GEOquery package in R to read the data, and use the limma package to construct the design matrix of each microarray dataset, specify the control group and the experimental group, use the lmFit function to fit the linear model of the control group and the experimental group, and calculate the expression levels of the genes in the two groups;

[0013] S1.2: Use the eBayes function to perform Bayesian adjustment on the results of the linear model and calculate the statistical significance of each gene. According to the difference screening criteria: |log2(FC)|>0.5, P<0.05, extract the genes with significant differential expression, where log2(FC) represents the degree of expression change of the treatment group relative to the control group, and P represents the confidence value. The P value is adjusted by using the Benjamini-Hochberg method to control the false positive rate.

[0014] S1.3: Use ggplot2 to draw a volcano plot to display the log2 FC and P value of each gene, and use the ComplexHeatmap package to display the screened DEGs. Then use the Venn diagram drawing tool to analyze the intersection of DEGs of the two datasets and determine the co-expressed genes.

[0015] As a further solution of the present invention, the specific calculation formula of log2(FC) in S1.1 is as follows:

[0016]

[0017] In the formula, log2(FC) represents the expression change degree of the treatment group relative to the control group; log 2 Represents the logarithm operation with base 2; Mean Treatment Represents the average expression level of any gene in the treatment group; Mean Control Represents the average expression level of the same gene selected in the control group as in the control group.

[0018] As a further scheme of the present invention, the collection and processing of sample data specifically includes: using groups of keywords corresponding to sarcopenia and Alzheimer's disease to search in public databases, screening corresponding disease research and data sets, and downloading the required data sets according to the screening results, including gene expression matrix files and corresponding metadata, checking missing values ​​in the gene expression matrix, calculating the missing rate, removing samples and genes with missing values ​​exceeding 30%, extracting sample data to be analyzed from the gene expression matrix that has passed quality control, arranging each group of sample data in ascending order, calculating the median, first quartile Q1 and third quartile Q3 of the arranged sample data, calculating the interquartile range IQR of the sample data based on Q1 and Q3, determining the upper and lower bounds of the outlier according to the IQR, drawing a box plot through the data visualization tool matplotlib, marking values ​​less than the lower bound or greater than the upper bound as outliers and eliminating them, and normalizing the remaining sample data.

[0019] As a further solution of the present invention, the specific steps of screening the Hub gene in step II are as follows

[0020] S2.1: Extract the DEGs screened from the differentially expressed gene analysis to construct a DEGs list, which includes the gene name or ID. Retrieve the corresponding gene / protein interaction information database through the provided DEGs list, calculate the confidence value of each group of interaction data retrieved, and screen out the data that do not meet the standard based on the screening standard of confidence value > 0.4;

[0021] S2.2: Genes / proteins in the interaction data obtained from the database were used as nodes of the network, and the corresponding interactions were used as edges of the network to construct the PPI network. The constructed PPI network was then imported into Cytoscape v3.9.1 for visualization analysis.

[0022] S2.3: The cytoHubba plug-in of Cytoscape was used to identify the key genes in the PPI network, and the scores of each gene were calculated using the MCC, MNC, EPC, DMNC, Degree, and Stress algorithms, respectively. The top 30 genes were selected based on the scores of each algorithm, and the key genes screened out by the six algorithms were visualized through the UpSet graph, and the Hub genes were identified.

[0023] As a further solution of the present invention, the specific steps of the Chinese medicine prediction analysis in step III are as follows:

[0024] S3.1: Extract the screened Hub gene list, access the medical ontology information retrieval database, and enter each Hub gene into the database for retrieval to find the Chinese medicine data that are related to each gene, and screen out the Chinese medicine data that meet the conditions based on the screening standard of confidence <0.05;

[0025] S3.2: Based on the screening criteria of key Chinese medicines with corresponding Hub genes ≥ 2, further screen out key Chinese medicine data from the Chinese medicine data that meet the conditions, and organize the information on the names, properties, meridians, and efficacy of Chinese medicines according to the Pharmacopoeia of the People's Republic of China (2020 edition), the Chinese Materia Medica Dictionary (Second Edition), and Chinese Materia Medica (New Century Second Edition);

[0026] S3.3: Based on the collated TCM information, use the ggplot2 package of R language to draw the four-Qi rectangular tree diagram, the five-flavor circular strip diagram, the meridian one-dimensional bar diagram, and the efficacy radial bar diagram.

[0027] As a further solution of the present invention, the specific steps of the tissue-specific expression analysis in step III are as follows:

[0028] S4.1: Log in to BioGPS and import the Hub gene list, query the expression levels of these genes in different tissues, extract the expression level data of Hub genes in various tissues from BioGPS, and use FPKM or TPM as the unit;

[0029] S4.2: Calculate the specific expression level of each group of genes in each Hub gene list based on the expression level of the target tissue and the median of the expression levels of all tissues, and compare the highest expression level and the second highest expression level of the Hub gene in the target tissue;

[0030] S4.3: When the specific expression level is greater than 10 times the median, and the highest expression level is greater than 2 times the second highest expression level, the corresponding Hub gene in the original Hub gene list is retained. Otherwise, the Hub genes that do not meet the conditions are screened out. After the screening is completed, the expression level data of the remaining genes in the Hub gene list are extracted, and each group of data is divided into a disease group and a control group;

[0031] S4.4: The Wilcoxon rank sum test was used to examine and compare the differences in expression between the disease group and the control group. Based on the Wilcoxon test results, a group comparison chart was drawn to show the differences in gene expression between the disease group and the control group. The sensitivity and specificity of each gene were then calculated, and the receiver operating characteristic curve was drawn and the area under the curve was calculated. The pROC package was used to perform receiver operating characteristic curve analysis. Finally, the receiver operating characteristic curve and the area under the curve were visualized to determine the diagnostic efficacy of each gene.

[0032] As a further solution of the present invention, the specific steps of constructing the dual disease diagnosis model in step IV are as follows:

[0033] S5.1: From the samples of sarcopenia and Alzheimer's disease, the expression data of each gene in the Hub gene list after screening were sorted out, and the expression matrix of the expression levels in different samples of the disease and control groups was constructed;

[0034] S5.2: Use the glm function of R language to build binary logistic regression models for sarcopenia and Alzheimer's disease, respectively. Use the pROC package to analyze the model prediction results of the two groups of models for the internal GEO datasets sarcopenia GSE1428 and Alzheimer's disease GSE5281. Receiver operating characteristic curves were constructed, and the area under the curve was calculated to verify the diagnostic efficacy of the model.

[0035] S5.3: Use the glm function in the rms package to calculate the net yield of each model, and use the ggplot2 package to draw the net yield curve. According to the difference between the predicted probability of each model and the actual probability, use the rms package to draw and construct a calibration curve to evaluate the consistency between the predicted probability of each group of models and the actual probability of occurrence;

[0036] S5.4: Perform ROC analysis on the external GEO dataset, repeat the internal GEO dataset validation steps, calculate the area under the curve and evaluate the diagnostic performance of the model, and use the ggplot2 package to draw the ROC curve, decision curve, and calibration curve to show the performance of the model on different datasets;

[0037] S5.5: After verification, the two sets of logistic regression models were converted into Nomogram models using the functions in the rms package, and the results of the models were displayed in the form of Nomogram to show the impact of different molecules on the predicted probability.

[0038] As a further solution of the present invention, the specific calculation formula of the Logistic regression model in S5.2 is as follows:

[0039] logit(L(Y=1))=β 0 +β 1 ·X 1 +β 2 ·X 2 +...+β n ·X n

[0040] Where L(Y=1) represents the probability that the sample belongs to the disease group; β 0 , β 1 , β 2 , ..., β n Represents the regression coefficient corresponding to each gene in the Hub gene list; X 1 , X 2 , …, X n represents the expression level of each gene in the Hub gene list; n represents the total number of genes in the Hub gene list.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] This intelligent dual-disease diagnosis method based on bioinformatics big data collects gene expression data of each group from public databases, screens for sarcopenia and Alzheimer's disease, and then uses the "limma" package in R language to perform differential expression analysis on the collected data sets, screens out DEGs genes with significant differences between the disease group and the control group, imports the screened DEGs into the protein-protein interaction database, constructs a PPI network, and uses Cytoscape software for visualization. The key genes are identified through the cytoHubba plug-in, and the identified Hub genes are analyzed for Chinese medicine relatedness, and the medical ontology information retrieval database is used to predict Chinese medicines that may have therapeutic effects. , and screened out traditional Chinese medicines with high correlation with the Hub gene. Then, the expression specificity of the Hub gene in different tissues was analyzed, and genes with significantly high expression in specific tissues were screened out. Then, the Logistic regression model was used to construct two diagnostic models for sarcopenia and Alzheimer's disease. The performance of the model was evaluated by drawing ROC curves and AUC values ​​based on internal and external data sets. The actual application effect of the model was evaluated based on the decision curve analysis and calibration curve of the model. This can provide new ideas for the combination of traditional medicine and modern medicine, expand treatment options, realize rapid processing and analysis of large amounts of gene expression data, significantly improve diagnostic efficiency, enhance diagnostic accuracy, and improve the ability to identify complex diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0044] Figure 1 This is a flowchart of an intelligent dual disease diagnosis method based on bioinformatics big data proposed by the present invention;

[0045] Figure 2 This is a diagram of the sarcopenia diagnosis model of the intelligent dual-disease diagnosis method based on bioinformatics big data proposed in the present invention;

[0046] Figure 3 This is a diagram of the Alzheimer's disease diagnosis model of an intelligent dual-disease diagnosis method based on bioinformatics big data proposed by the present invention;

[0047] Figure 4 This is a sarcopenia decision curve analysis diagram of an intelligent dual-disease diagnosis method based on bioinformatics big data proposed in the present invention;

[0048] Figure 5 This is a decision curve analysis diagram of Alzheimer's disease based on an intelligent dual-disease diagnosis method based on bioinformatics big data proposed in the present invention;

[0049] Figure 6This is a calibration curve diagram of sarcopenia for an intelligent dual-disease diagnosis method based on bioinformatics big data proposed in the present invention;

[0050] Figure 7 This is a calibration curve diagram of Alzheimer's disease for the intelligent dual-disease diagnosis method based on bioinformatics big data proposed in the present invention. DETAILED DESCRIPTION

[0051] Reference Figure 1-7 , an intelligent dual disease diagnosis method based on bioinformatics big data, the specific steps of the diagnosis method are as follows:

[0052] Collect and process sample data and screen differentially expressed genes.

[0053] Specifically, the microarray datasets GSE1428 and GSE5281 were extracted from the preprocessed sample data, the GEOquery package in R was used to read the data, and the design matrix of each microarray dataset was constructed by the limma package. The control group and the experimental group were specified at the same time. The linear model of the control group and the experimental group was fitted by the lmFit function, and the expression levels of the genes in the two groups were calculated. The results of the linear model were Bayesian adjusted using the eBayes function, and the statistical significance of each gene was calculated. According to the difference screening criteria: |log2(FC)|>0.5, P<0.05, the genes with significant differential expression were extracted, where log2(FC) represents the degree of expression change of the treatment group relative to the control group, P represents the confidence value, and the P value was adjusted by using the Benjamini-Hochberg method to control the false positive rate. The volcano map was drawn using ggplot2 to display the log2 FC and P value of each gene, and the screened DEGs were displayed by the ComplexHeatmap package. The Venn diagram drawing tool was used to analyze the intersection of the DEGs of the two datasets to determine the commonly expressed genes.

[0054] It should be further explained that the specific calculation formula of log2(FC) is as follows:

[0055]

[0056] In the formula, log2(FC) represents the expression change degree of the treatment group relative to the control group; log 2 Represents the logarithm operation with base 2; Mean Treatment Represents the average expression level of any gene in the treatment group; Mean Control Represents the average expression level of the same gene selected in the control group as in the control group.

[0057] In addition, it should be noted that the collection and processing of sample data specifically includes: searching in public databases using groups of keywords corresponding to sarcopenia and Alzheimer's disease, screening corresponding disease studies and data sets, and downloading the required data sets according to the screening results, including gene expression matrix files and corresponding metadata, checking missing values ​​in the gene expression matrix, calculating the missing rate, removing samples and genes with missing values ​​exceeding 30%, extracting sample data to be analyzed from the gene expression matrix that has passed quality control, arranging each group of sample data in ascending order, calculating the median, first quartile Q1, and third quartile Q3 of the arranged sample data, calculating the interquartile range IQR of the sample data based on Q1 and Q3, determining the upper and lower bounds of the outliers according to the IQR, drawing a box plot through the data visualization tool matplotlib, marking values ​​less than the lower bound or greater than the upper bound as outliers and removing them, and normalizing the remaining sample data.

[0058] The above method was used to perform gene differential expression analysis on the dataset GSE1428, and 695 DEGs were screened out, including 348 up-regulated genes and 347 down-regulated genes; the gene differential expression analysis was performed on the dataset GSE5281, and 6021 DEGs were screened out, including 2773 up-regulated genes and 3248 down-regulated genes.

[0059] At the same time, the two groups of differentially expressed genes were plotted as volcano plots, and the top 50 genes with significant differences were selected to draw heat maps. The up-regulated and down-regulated DEGs in the two data sets were intersected by Venn diagrams, and 177 DEGs with common expression trends were obtained, including 84 up-regulated DEGs and 93 down-regulated DEGs.

[0060] Construct a protein interaction network and screen Hub genes based on the network.

[0061] Specifically, the DEGs screened out from the differentially expressed gene analysis were extracted to construct a DEGs list, which included gene names or IDs. The corresponding gene / protein interaction information database was retrieved through the provided DEGs list, and the confidence values ​​of each group of interaction data retrieved were calculated. Based on the screening criterion of confidence value > 0.4, the data that did not meet the standard were screened out. The genes / proteins in the interaction data obtained from the database were used as nodes of the network, and the corresponding interactions were used as edges of the network to construct a PPI network. The constructed PPI network was imported into Cytoscape v3.9.1 for visualization analysis. The key genes in the PPI network were identified through the cytoHubba plug-in of Cytoscape, and the scores of each gene were calculated using the MCC, MNC, EPC, DMNC, Degree and Stress algorithms respectively. According to the scores of each algorithm, the top 30 genes were selected respectively, and the key genes screened out by the six algorithms were visualized through the UpSet graph, and the Hub genes among them were identified.

[0062] Through the above screening method, the information of 177 co-expressed DEGs obtained by screening was imported into the STRING database, and the isolated nodes were removed to construct the PPI network, including 109 nodes and 524 edges. Using the six common algorithms of cytoHubba, 12 overlapping Hub genes were screened, namely DLD, UQCRFS1, NDUFS1, NDUFA10, CDKN1A, GOT2, LDHA, CEBPD, ERBB2, PGK1, SAMM50, and ZBTB16, among which CDKN1A, CEBPD, ERBB2, and ZBTB16 were up-regulated genes, and DLD, UQCRFS1, NDUFS1, NDUFA10, GOT2, LDHA, PGK1, and SAMM50 were down-regulated genes.

[0063] Traditional Chinese medicine prediction analysis and tissue-specific expression analysis of Hub genes were performed.

[0064] Specifically, the screened Hub gene list was extracted, the medical ontology information retrieval database was accessed, and each Hub gene was entered into the database for retrieval to find the Chinese medicine data that were related to each gene, and based on the screening criteria of confidence level <0.05, the qualified Chinese medicine data were screened out, and based on the screening criteria of key Chinese medicines with corresponding Hub genes ≥2, the key Chinese medicine data were further screened out from the Chinese medicine data that met the conditions, and the information on the name, nature and taste, meridians and efficacy of Chinese medicines was standardized and sorted out according to the "Pharmacopoeia of the People's Republic of China" (2020 edition), "Dictionary of Chinese Materia Medica" (Second Edition) and "Chinese Materia Medica" (New Century Second Edition). According to the sorted Chinese medicine information, the ggplot2 package of R language was used to draw the four-qi rectangular tree diagram, five-flavor circular strip diagram, meridian one-dimensional bar diagram and efficacy radial bar diagram.

[0065] Specifically, log in to BioGPS and import the Hub gene list, query the expression levels of these genes in different tissues, extract the expression level data of Hub genes in various tissues from BioGPS, and calculate the specific expression level of each group of genes in each Hub gene list in units of FPKM or TPM according to the expression level of the target tissue and the median expression level of all tissues. At the same time, compare the highest expression level and the second highest expression level of the Hub gene in the target tissue. When the specific expression level is greater than 10 times the median, and the highest expression level is greater than 2 times the second highest expression level, retain the corresponding Hub gene in the original Hub gene list. Otherwise, screen out the Hub genes that do not meet the conditions. After the screening is completed, extract the expression level data of the remaining genes in the Hub gene list, and divide each group of data into a disease group and a control group. Use Wilcoxon rank The statistical method of sumtest was used to test and compare the expression differences between the disease group and the control group. Then, according to the results of the Wilcoxon test, a group comparison chart was drawn to show the expression differences of each gene in the disease group and the control group. Then, the sensitivity and specificity of each gene were calculated, and the receiver operating characteristic curve was drawn and the area under the curve was calculated. The receiver operating characteristic curve analysis was then performed using the pROC package. Finally, the receiver operating characteristic curve and the area under the curve were visualized to determine the diagnostic efficacy of each gene.

[0066] It should be further explained that, through the above-mentioned prediction and analysis of traditional Chinese medicine, 12 Hub genes were introduced into CoremineMedical to predict traditional Chinese medicines with potential therapeutic effects, and 21 key traditional Chinese medicines with corresponding Hub genes ≥ 2 were further screened out, namely, rhubarb, fish bladder glue, sophora japonica, salvia miltiorrhiza, clam shell, lithospermum officinale, panax notoginseng, seven-leaf flos, dandelion, buffalo horn, sophora flavescens, scutellaria baicalensis, fish brain stone, Ligusticum chuanxiong, wolfberry, silkworm moth, silkworm feces, inulin, dandelion, and Evodia rutaecarpa. Among them, the four properties are mainly concentrated in cold, the five flavors are mainly concentrated in bitter, the meridians are mainly the liver meridian of foot Jueyin, the spleen meridian of foot Taiyin, and the kidney meridian of foot Shaoyin, and the main effect is clearing heat.

[0067] According to the above tissue-specific expression analysis, 12 Hub genes were introduced into BioGPS, and 4 genes with specific expression in specific tissues or organ systems were identified, namely DLD, NDUFS1, ERBB2, and ZBTB16. The specific expression system of DLD and NDUFS1 is the blood / immune system, the specific expression tissue of ERBB2 is the placenta, and the specific expression system of ZBTB16 is the respiratory system. In sarcopenia, the four molecules have differences in expression between the disease and control groups, and the trends are consistent. DLD and NDUFS1 are downregulated in the disease group, and ERBB2 and ZBTB16 are upregulated in the disease group; the verification results of AD are consistent, and the four molecules have shown good diagnostic efficacy in both diseases. In sarcopenia, the AUC value of the DLD molecule is 0.858, the AUC value of the NDUFS1 molecule is 0.812, the AUC value of the ERBB2 molecule is 0.850, and the AUC value of the ZBTB16 molecule is 0.800; in AD, the AUC value of the DLD molecule is 0.784, the AUC value of the NDUFS1 molecule is 0.752, the AUC value of the ERBB2 molecule is 0.626, and the AUC value of the ZBTB16 molecule is 0.684.

[0068] Construct a dual disease diagnosis model and conduct analytical applications.

[0069] Specifically, refer to Figure 1-7It can be seen that the expression data corresponding to each gene in the Hub gene list after screening were sorted out from the samples of sarcopenia and Alzheimer's disease, and the expression matrix of the expression in different samples of the disease and control groups was constructed. The two-classification Logistic regression models of sarcopenia and Alzheimer's disease were constructed respectively using the glm function of R language. The model prediction results of the two groups of models for the internal GEO data sets sarcopenia GSE1428 and Alzheimer's disease GSE5281 were analyzed using the pROC package. The receiver operating characteristic curve was constructed, and the area under the curve was calculated to verify the diagnostic efficacy of the model. The net rate of return of each model was calculated using the glm function in the rms package, and the net rate of return curve was plotted using the ggplot2 package. According to the difference between the predicted probability of each model and the actual probability, the calibration curve was constructed using the rms package to evaluate the consistency between the predicted probability of each group of models and the actual probability of occurrence. ROC analysis was performed on the external GEO dataset, and the internal GEO dataset verification steps were repeated to calculate the area under the curve and evaluate the diagnostic efficacy of the model. The ROC curve, decision curve and calibration curve were drawn using the ggplot2 package to demonstrate the performance of the model on different datasets. After the verification, the two groups of Logistic regression models were converted into Nomogram models using the functions in the rms package, and the results of the model were displayed in the form of Nomogram to show the influence of different molecules on the predicted probability.

[0070] In addition, by Figure 2-7 It can be seen that in the internal validation, the two groups of diagnostic models showed high diagnostic efficacy in both diseases. In sarcopenia, the AUC value of the diagnostic model was 0.958; in AD disease, the AUC value of the diagnostic model was 0.817. The diagnostic models showed high net returns in both diseases and good fitting in both diseases. In the external validation, the diagnostic models showed high diagnostic efficacy in both diseases. In sarcopenia, the AUC value of the diagnostic model was 0.817; in AD disease, the AUC value of the diagnostic model was 0.780.

Claims

1. An intelligent dual disease diagnosis method based on bioinformatics big data, characterized in that: The specific steps of this diagnostic method are as follows: Ⅰ. Collect and process sample data and screen differentially expressed genes; II. Construct a protein interaction network and screen Hub genes based on the network; III. Perform TCM prediction analysis and tissue-specific expression analysis on Hub genes; IV. Construct a dual disease diagnosis model and conduct analytical applications.

2. According to claim 1, an intelligent dual disease diagnosis method based on bioinformatics big data is characterized in that: The specific steps of screening differentially expressed genes in step I are as follows: S1.1: Extract the microarray datasets GSE1428 and GSE5281 from the preprocessed sample data, use the GEOquery package in R to read the data, and use the limma package to construct the design matrix of each microarray dataset, specify the control group and the experimental group, use the lmFit function to fit the linear model of the control group and the experimental group, and calculate the expression levels of the genes in the two groups; S1.2: Use the eBayes function to perform Bayesian adjustment on the results of the linear model and calculate the statistical significance of each gene. According to the difference screening criteria: |log2(FC)|>0.5, P<0.05, extract the genes with significant differential expression, where log2(FC) represents the degree of expression change of the treatment group relative to the control group, and P represents the confidence value. The P value is adjusted by using the Benjamini-Hochberg method to control the false positive rate. S1.3: Use ggplot2 to draw a volcano plot to display the log2 FC and P value of each gene, and use the ComplexHeatmap package to display the screened DEGs. Then use the Venn diagram drawing tool to analyze the intersection of DEGs of the two datasets and determine the co-expressed genes.

3. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 2 is characterized in that: The specific calculation formula of log2(FC) described in S1.1 is as follows: In the formula, log2(FC) represents the expression change of the treatment group relative to the control group; log2 represents the logarithmic operation with base 2; Mean Treatment Represents the average expression level of any gene in the treatment group; Mean Control Represents the average expression level of the same gene selected in the control group as in the control group.

4. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 2 is characterized in that: The specific steps for screening Hub genes in step II are as follows S2.1: Extract the DEGs screened from the differentially expressed gene analysis to construct a DEGs list, which includes the gene name or ID. Retrieve the corresponding gene / protein interaction information database through the provided DEGs list, calculate the confidence value of each group of interaction data retrieved, and screen out the data that do not meet the standard based on the screening standard of confidence value > 0.4; S2.2: Genes / proteins in the interaction data obtained from the database were used as nodes of the network, and the corresponding interactions were used as edges of the network to construct the PPI network. The constructed PPI network was then imported into Cytoscape v3.9.1 for visualization analysis. S2.3: The cytoHubba plug-in of Cytoscape was used to identify the key genes in the PPI network, and the scores of each gene were calculated using the MCC, MNC, EPC, DMNC, Degree, and Stress algorithms, respectively. The top 30 genes were selected based on the scores of each algorithm, and the key genes screened out by the six algorithms were visualized through the UpSet graph, and the Hub genes were identified.

5. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 4 is characterized in that: The specific steps of the prediction and analysis of traditional Chinese medicine described in step III are as follows: S3.1: Extract the screened Hub gene list, access the medical ontology information retrieval database, and enter each Hub gene into the database for retrieval to find the Chinese medicine data that are related to each gene, and screen out the Chinese medicine data that meet the conditions based on the screening standard of confidence <0.05; S3.2: Based on the screening criteria of key Chinese medicines with corresponding Hub genes ≥ 2, further screen out key Chinese medicine data from the Chinese medicine data that meet the conditions, and organize the information on the names, properties, meridians, and efficacy of Chinese medicines according to the Pharmacopoeia of the People's Republic of China (2020 edition), the Chinese Materia Medica Dictionary (Second Edition), and Chinese Materia Medica (New Century Second Edition); S3.3: Based on the collated TCM information, use the ggplot2 package of R language to draw the four-Qi rectangular tree diagram, the five-flavor circular strip diagram, the meridian one-dimensional bar diagram, and the efficacy radial bar diagram.

6. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 4 is characterized in that: The specific steps of tissue-specific expression analysis in step III are as follows: S4.1: Log in to BioGPS and import the Hub gene list, query the expression levels of these genes in different tissues, extract the expression level data of Hub genes in various tissues from BioGPS, and use FPKM or TPM as the unit; S4.2: Calculate the specific expression level of each group of genes in each Hub gene list based on the expression level of the target tissue and the median of the expression levels of all tissues, and compare the highest expression level and the second highest expression level of the Hub gene in the target tissue; S4.3: When the specific expression level is greater than 10 times the median, and the highest expression level is greater than 2 times the second highest expression level, the corresponding Hub gene in the original Hub gene list is retained. Otherwise, the Hub genes that do not meet the conditions are screened out. After the screening is completed, the expression level data of the remaining genes in the Hub gene list are extracted, and each group of data is divided into a disease group and a control group; S4.4: The Wilcoxon rank sum test was used to examine and compare the differences in expression between the disease group and the control group. Based on the Wilcoxon test results, a group comparison chart was drawn to show the differences in gene expression between the disease group and the control group. The sensitivity and specificity of each gene were then calculated, and the receiver operating characteristic curve was drawn and the area under the curve was calculated. The pROC package was used to perform receiver operating characteristic curve analysis. Finally, the receiver operating characteristic curve and the area under the curve were visualized to determine the diagnostic efficacy of each gene.

7. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 6 is characterized in that: The specific steps of constructing the dual disease diagnosis model described in step IV are as follows: S5.1: From the samples of sarcopenia and Alzheimer's disease, the expression data of each gene in the Hub gene list after screening were sorted out, and the expression matrix of the expression levels in different samples of the disease and control groups was constructed; S5.2: Use the glm function of R language to build binary logistic regression models for sarcopenia and Alzheimer's disease, respectively. Use the pROC package to analyze the model prediction results of the two groups of models for the internal GEO datasets sarcopenia GSE1428 and Alzheimer's disease GSE5281. Receiver operating characteristic curves were constructed, and the area under the curve was calculated to verify the diagnostic efficacy of the model. S5.3: Use the glm function in the rms package to calculate the net yield of each model, and use the ggplot2 package to draw the net yield curve. According to the difference between the predicted probability of each model and the actual probability, use the rms package to draw and construct a calibration curve to evaluate the consistency between the predicted probability of each group of models and the actual probability of occurrence; S5.4: Perform ROC analysis on the external GEO dataset, repeat the internal GEO dataset validation steps, calculate the area under the curve and evaluate the diagnostic performance of the model, and use the ggplot2 package to draw the ROC curve, decision curve, and calibration curve to show the performance of the model on different datasets; S5.5: After verification, the two sets of logistic regression models were converted into Nomogram models using the functions in the rms package, and the results of the models were displayed in the form of Nomogram to show the impact of different molecules on the predicted probability.

8. The intelligent dual disease diagnosis method based on bioinformatics big data according to claim 7 is characterized in that: The specific calculation formula of the Logistic regression model described in S5.2 is as follows: logit(L(Y=1))=β0+β1·X1+β2·X2+…+β n ·X n Where L(Y=1) represents the probability that the sample belongs to the disease group; β0, β1, β2, …, β n Represents the regression coefficient corresponding to each gene in the Hub gene list; X1, X2, …, X n represents the expression level of each gene in the Hub gene list; n represents the total number of genes in the Hub gene list.

Images