Method for analyzing and identifying diseases by using gene and co-disease

Abstract

The invention discloses a method for analyzing and identifying diseases by using gene combination co-disease. The method comprises the following steps: screening and extracting differentially expressed genes (DEGs) of two or more common diseases by using DESeq2; subsequently, this process results in shared genes between diseases. The shared DEG is submitted to a DSigDB database, and related drug molecules are retrieved. And the MCODE application program and the CytoHubba application program in the Cytoscape are used for identifying the central gene. A machine learning algorithm including LASSO analysis and random forest analysis is adopted to identify potential feature genes, the diagnosis potential is further evaluated through an ROC curve, and finally the feature center differential expression genes are determined. According to the method, the characteristic center differential expression genes which can be used for co-disease analysis and identification of diseases can be screened out, so that doctors are expected to identify more potential causes, and more accurate medical services are provided for patients.

Description

Technical Field

[0001] The present invention relates to the technical field of gene data analysis, and in particular to a method for identifying diseases by using gene combined with comorbidity analysis. Background Art

[0002] Although many diseases appear different in phenotype, they may actually share some common genetic basis or pathological mechanisms. In recent years, with the development of genomic technology, people have begun to realize that there are potential commonalities between many diseases. These commonalities may arise from common genetic variations, similar environmental exposures, or shared pathophysiological processes. For example, EPL and EP are both part of pregnancy complications, although they have different pathological characteristics - the former occurs when the embryo that implants normally in the uterus fails to develop successfully, while the latter is when the fertilized egg implants and develops outside the uterus. However, in actual clinical practice, many patients do not have a clear cause, which makes it difficult for clinicians to provide targeted treatment plans.

[0003] Traditional genetic research focuses on single disease models, that is, trying to find genetic variants that are closely related to a specific disease. However, this method is powerless when faced with complex diseases with unknown causes. Especially for cases such as EPL and EP that involve multiple factors, single disease research may not be able to fully reveal the genetic mechanisms behind them. In addition, traditional research methods often ignore the potential correlation between different diseases, which limits our understanding of the nature of the disease and our ability to develop effective treatments.

[0004] In order to overcome the above limitations, a new research method has emerged in recent years - Gene Joint Comorbidity Analysis. This method integrates the genomic data of patients with different diseases to find common genetic variations that exist in multiple diseases, thereby revealing the potential connection between different diseases. Through this comprehensive analysis, researchers can more accurately identify key genes or biomarkers that play a role in multiple diseases, thereby providing a more solid foundation for clinical practice. Therefore, it is necessary to develop a method to identify diseases using gene joint comorbidity analysis in order to diagnose diseases more accurately and ultimately improve existing treatment strategies. Summary of the invention

[0005] The purpose of the present invention is to provide a method for identifying diseases by combining gene and comorbidity analysis, which can screen out characteristic central differentially expressed genes that can be used for comorbidity analysis and identification of diseases, and is expected to help doctors identify more potential causes of diseases and provide patients with more accurate medical services.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] A method for identifying diseases by combining gene and comorbidity analysis, characterized by comprising the following steps:

[0008] S1. Screening genes with significant differential expression for two or more common diseases respectively, and deleting the same genes with significant differential expression between the diseases to obtain shared differentially expressed genes;

[0009] S2. The protein interaction network of shared differentially expressed genes in all the above diseases was constructed using the STRING database. The Cytoscape plug-in MCODE was used to identify the top modules in the protein interaction network. The top modules were defined as core modules, which consisted of multiple shared differentially expressed genes.

[0010] S3. Using the CytoHubba application in Cytoscape 3.8.2, we used a variety of different algorithms to select the top genes from each algorithm and sorted them according to their frequency of occurrence. We then crossed these genes with the shared differentially expressed genes in the core module, and finally obtained multiple hub differentially expressed genes.

[0011] S4. Based on the DSigDB database, the above-mentioned hub differentially expressed genes were uploaded to the database, and multiple candidate drugs were predicted and ranked according to the comprehensive scores. The molecular docking of the TOP1 drug with the above-mentioned hub differentially expressed genes was performed using Autodock Tools;

[0012] S5. LASSO regression analysis was performed on the above-mentioned hub differentially expressed genes to identify characteristic genes, and random forest analysis was performed to obtain the MeanDiscoveryGini values ​​of multiple differentially expressed genes, which were then sorted according to importance. The results of LASSO regression and random forest analysis were combined to finally determine the characteristic hub differentially expressed genes;

[0013] S6. The ROC curve was used to evaluate the diagnostic performance of the characteristic hub differentially expressed genes. Finally, the characteristic hub differentially expressed genes could be used for comorbidity analysis and identification of the above diseases.

[0014] Further preferably, in step S1, differential expression is evaluated by DESeq2 to perform differentially expressed gene screening.

[0015] Further preferably, in step S1, differentially expressed genes are screened with >1 and FDR<0.05 as critical criteria.

[0016] Further preferably, in step S3, the algorithms used include MCC, MNC, EPC, stress, DMNC, degree, spacing, bottleneck, clustering coefficient, eccentricity, compactness and radiation.

[0017] Further preferably, in step S5, the parameters used in the LASSO analysis are α=1 and λ.

[0018] Further preferably, a "RandomForest" function is used to perform random forest analysis, the minimum error is selected as the mtry node value, and the image value that tends to be stable is selected as ntree.

[0019] In summary, the present invention has the following beneficial effects:

[0020] First, through the identification and functional enrichment analysis of differentially expressed genes (DEGs), we can gain in-depth understanding of the changing patterns of gene expression and their biological functions in specific disease states, such as inflammatory response, immune response, etc., which helps to reveal the underlying mechanism of the disease.

[0021] Second, constructing a protein-protein interaction (PPI) network and identifying key hub genes provides us with a new perspective to understand the molecular basis and regulatory network of complex diseases, thereby helping to identify new therapeutic targets.

[0022] Third, by using machine learning techniques such as LASSO regression and random forest analysis, characteristic genes that are highly correlated with the disease can be screened out, which improves the accuracy and predictive ability of disease diagnosis, and provides a strong scientific basis and technical support for the early diagnosis of the disease, prognosis assessment, and even the formulation of personalized treatment strategies.

[0023] Fourth, the combination of experimental verification and immune cell infiltration analysis not only confirmed the role of key genes in the disease, but also further revealed the dynamic changes of immune cells in the disease process and their relationship with gene expression. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is the immune-related genetic feature of EPL.

[0025] Figure 2 (A) Heat map of differentially expressed genes (DEGs) in early pregnancy loss (EPL); (B) Heat map of DEGs in ectopic pregnancy (EP); (C) Volcano plot of DEGs in EPL; (D) Volcano plot of DEGs in EP; (E) DEGs shared between EPL and EP.

[0026] Figure 3(A) Circular plot of gene ontology (GO) functional analysis of shared DEGs; (B) GO functional analysis bar graph of shared DEGs; (C) Circular plot of KEGG analysis of shared DEGs; (D) Bubble plot of KEGG analysis of shared DEGs; (E) Protein interaction network of shared DEGs.

[0027] Figure 4 (A) Module 1 of the PPI network; (B) Module 2 of the PPI network; (C) KEGG analysis of module 1; (D) GO analysis of module 1; (E) Overlap of DEGs between MCODE and CytoHubba; (F) Functional analysis of hub genes using GeneMANIA.

[0028] Figure 5 (A) Schematic diagram of molecular docking between cholecalciferol and ITGB2; (B) Schematic diagram of molecular docking between cholecalciferol and TLR2; (C) Schematic diagram of molecular docking between cholecalciferol and TLR4.

[0029] Figure 6 (AB) PCA plots before and after batch correction using the dataset GSE123719; (CD) Machine learning of LASSO analysis; (EF) Machine learning of LASSO analysis; (G) Gene overlap in machine learning; (H) Diagnostic potential of CYBB gene in EPL; (I) Diagnostic potential of CYBB gene in EP.

[0030] Figure 7 (A) Relative expression levels in CYBB were verified by RT-qPCR in a sample set of 7 EPL patients, 7 EP patients, and 7 controls; (B) Stacked histogram of immune cells in EP patients; (C) Stacked histogram of immune cells in EPL patients; (D) Comparison of immune cell components between EPL patients and healthy controls; (E) Comparison of immune cell components between EP patients and healthy controls; (F) Correlation of central DEGs with immune cell components in EPL patients; (G) Correlation of central DEGs with immune cell components in EP patients; (H) Single-cell map of CYBB genes in placental tissue. DETAILED DESCRIPTION

[0031] The following is a further detailed description of a method for identifying diseases by gene combined comorbidity analysis proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer according to the following description. It should be noted that the accompanying drawings are all in a very simplified form and are not in precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0032] Example, in order to explore the pathological mechanism of EPL and EP and explore the diagnostic and therapeutic value of differentially expressed genes (DEGs) in placental villus tissue, we attempted to study the mutual potential interactions of DEGs between EPL and EPL according to this workflow, especially regarding the immune-related genetic features of EPL ( Figure 1 ).

[0033] Taking early pregnancy loss (EPL) and ectopic pregnancy (EP) as examples, we collected chorionic villus tissue from patients who visited the gynecology clinic of Zhejiang University Women's Hospital School of Medicine for EPL and unexpected pregnancy. Twelve EPL patients, 12 EP patients and 12 healthy controls were included. 1. Identification and functional enrichment analysis of DEGs

[0034] Differential expression was assessed by DESeq2 using read counts as input. Using >1 and FDR <0.05 as critical criteria, a volcano plot showing the expression of all differentially expressed genes (DEGs) was created using the ggplot2 package in R, while a heatmap was generated using the pheatmap package. Significant differential expression of 1220 genes could be observed in the villus tissue of EPL, of which 1210 DEGs were significantly upregulated and 10 DEGs were significantly downregulated ( Figure 2 -AB). At the same time, we observed 622 genes significantly upregulated and 18 genes significantly downregulated in the EP group compared with the control group (P<0.05) ( Figure 2 -CD). We finally obtained 241 shared DEGs between the EPL and EP groups ( Figure 2 E).

[0035] To further investigate the biological functions of these shared DEGs, we performed GO term and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on the shared DEGs using the DAVID database (https: / / david.ncifcrf.gov / ). These shared DEGs were significantly enriched in the terms (including molecular function, biological process, and cellular component) obtained from the KEGG pathway and GO analysis, with an adjusted P value < 0.05.

[0036] GO analysis results showed that DEGs were enriched in various biological processes, including inflammatory response, immune response, and innate immune response. In addition, they were related to molecular functions such as transmembrane signaling receptor activity, signaling receptor activity, and complement receptor activity. In addition, DEGs were also involved in cellular components such as plasma membrane and secretory granule membrane. Figure 3 AB shows the top ten results of MF, CP, and BP ranked by FDR in GO function analysis.

[0037] KEGG pathway analysis results showed that DEGs were widely enriched in various immune and inflammation-related pathways, such as Toll-like receptor signaling pathway, natural killer cell-mediated cytotoxicity, and NOD-like receptor signaling pathway (FDR < 0.05) ( Figure 3 CD).

[0038] 2. Construction of PPI network and identification of hub genes

[0039] We constructed a protein-protein interaction (PPI) network of shared DEGs in EPL and EP using a search tool for retrieving interacting genes (STRING, http: / / string-db.org / , version 12.0) to highlight complex regulatory connections. Interactions with a combined score of more than 0.4 were considered statistically significant ( Figure 3 E).

[0040] The PPI network was visualized using Cytoscape (http: / / www.cytoscape.org / , version 3.8.2). The Cytoscape plug-in MCODE (Molecular Complex Detection) was used to identify key functional modules, with the following standard settings: K-core = 2, degree cutoff = 2, maximum depth = 100, and node score cutoff = 0.2. MCODE was used to identify the top two modules in the PPI network, and the top-ranked modules were defined as core modules ( Figure 4 AB). The core module consisted of 18 shared DEGs with a score of 17.294. The KEGG functional results of the core module showed that these 18 shared DEGs were significantly enriched in various immune signaling pathways, including Toll-like receptor signaling pathway, NOD-like receptor signaling pathway, and neutrophil extracellular trap formation ( Figure 4 C).

[0041] Figure 4 D summarizes the top 10 results of the GO functional analysis of the core module genes (sorted by FDR value), highlighting the close association with immune-related functions. Among them, only 7 MF classifications met the condition of FDR < 0.5.

[0042] Using the cytoHubba application in Cytoscape 3.8.2, we employed 12 algorithms (MCC, MNC, EPC, stress, DMNC, degree, inter-degree, bottleneck, clustering coefficient, eccentricity, compactness, and radiality) to select hub genes (Hub genes). We selected the top 10 genes from each algorithm and ranked them according to their frequency of occurrence. Finally, the top 10 genes were determined based on their frequency. Table 1 provides the information of the 10 genes, while Table S4 records the top 10 genes in each algorithm. We crossed these 10 genes with the 18 DEGs in the core module and finally obtained 7 hub DEGs (CCL2, CD86, CYBB, FCER1G, ITGB2, TLR2, and TLR4) ( Figure 4 E). Table 1. The top ten shared DEGs obtained by 12 algorithms.

[0043]

[0044] Based on the functional annotation model of GeneMANIA, we established a co-expression network to describe the genetic interactions of these 7 hub DEGs and their co-expressed genes, including co-expression, physical interaction, co-localization, prediction, pathway, genetic interaction, and shared protein domains. We obtained 20 predicted co-expressed genes, with a total of 27 predicted genes. Functional analysis results showed that they were closely related to immune inflammatory responses, including the migration and activation of immune cells, the production of chemokines, and the activation of immune-related signaling pathways (FDR < 0.05) ( Figure 4 F). Potential immune inflammatory response is considered to be one of the common underlying factors in patients with EPL and EP.

[0045] 3. Drug candidate prediction and molecular docking

[0046] Based on the Drug Signature Database (DSigDB, http: / / dsigdb.tanlab.org / DSigDBv1.0 / ), the shared DEGs were uploaded to DSigDB, and candidate drugs could be predicted to evaluate the medicinal activity of target genes, and 135 small molecule drugs were predicted (adjusted P value < 0.05). Sorted by comprehensive scores, the top ten predicted drugs are shown in Table 2.

[0047] Table 2 Top 10 drug candidates predicted using DSigDB

[0048]

[0049]

[0050] In order to further understand the effects of candidate drugs on drug target genes and the drug ability of target genes, this study further performed molecular docking at the atomic level to evaluate the binding energy and interaction pattern between candidate drugs and their targets. Using Autodock 4.0 (http: / / autodock.scripps.edu / ), a computerized protein-ligand docking software, molecular docking of important drugs and proteins encoded by corresponding target genes was performed. We performed molecular docking of the TOP1 drug cholecalciferol with the above-mentioned hub DEGs, including ITGB2, TLR4 and TLR2 ( Figure 5 AC).

[0051] 4. Machine Learning

[0052] Considering the sample size requirement, we merged the sequencing expression values ​​of EPL and NP with the GSE123719 dataset and performed batch correction. The results showed that the batch effect was effectively eliminated, and the dataset after batch removal was ready for further analysis ( Figure 6 AB).

[0053] Next, we performed LASSO regression analysis on these 7 hub genes to identify characteristic genes, and constructed the least absolute shrinkage and selection operator (LASSO) regression prediction model using the “cv.glmnet” function in the R package glmnet. The parameters used in the LASSO analysis were α = 1 and λ.min was selected as the optimal lambda. LASSO analysis identified 3 characteristic hub DEGs (CD86, CYBB, and FCER1G), and the most appropriate lambda.min was 0.044 ( Figure 6 CD). With ntree set to 500, we performed a random forest analysis using the “RandomForest” function. The minimum error was selected as the mtry node value and the image value that tended to be stable was selected as ntree to obtain the MeanDiscoveryGini values ​​of the 10 differentially expressed genes, which were then ranked according to importance ( Figure 6 -EF). According to the filtering criteria of MeanDiscoveryGini value > 2.5, we finally identified 2 characteristic DEGs: CYBB and ITGB2.

[0054] Combining the results of LASSO regression and random forest analysis, CYBB was identified as a characteristic hub DEG ( Figure 6 G). We used ROC curve to evaluate the diagnostic performance of CYBB gene. The results showed that CYBB gene was robust in identifying diseases between EPL and EP ( Figure 6 HI).

[0055] 5. CYBB Verification

[0056] We again collected placental villus tissues from 7 cases of EPL, EP and healthy women. RT-qPCR results also showed that the CYBB gene was significantly upregulated in the placental villus tissues of EPL and EP patients (P<0.05) ( Figure 7 A).

[0057] 6. Immune cell infiltration

[0058] IOBR is an immuno-oncology computational tool for tumor immunobiology research. Here, we used the xCell method in IOBR to calculate the immune infiltrating cell fractions of 64 different immune cell types in each sample based on expression profiles. The results obtained from xCell were visualized using the “corplot” and “ggplot2” packages in R. The results showed that in the EPL group, the infiltration levels of M1 macrophages, NK cells, and cDCs in the immune cell population were significantly higher than those in the control group ( Figure 7 -BC). In addition, in the EP group, a significant increase in the infiltration levels of monocytes, aDCs, and iDCs was observed (P<0.05) ( Figure 7 -DE).

[0059] In addition, we used the Spearman rank correlation test to perform correlation analysis between immune cells and central DEGs, and the score represented the degree of correlation. The results showed that there was a significant correlation between the 7 central DEGs and the level of immune cell infiltration ( Figure 7 -FG). In the immunomodulatory analysis of the EPL and EP groups, a positive correlation was observed between aDC, iDC, macrophages, and CYBB genes (P<0.05).

[0060] The Human Protein Atlas database (HPA, https: / / www.proteinatlas.org / ) is a systematic human proteomics project that aims to provide comprehensive data on human protein expression, including protein expression profiles in different tissues and cell types, as well as subcellular localization information. We obtained a single-cell map of the CYBB gene in placental tissue from the HPA database, revealing a close association between the CYBB gene and fetal-derived macrophages, Hofbauer cells ( Figure 7 H).

[0061] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0062] The above description is only a description of the preferred embodiments of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art in the field of the present invention based on the above disclosure shall fall within the scope of protection of the claims.

Claims

1. A method for identifying diseases by combining gene and comorbidity analysis, characterized in that: The following steps are involved: S1. Screening genes with significant differential expression for two or more common diseases respectively, and deleting the same genes with significant differential expression between the diseases to obtain shared differentially expressed genes; S2. The protein interaction network of shared differentially expressed genes in all the above diseases was constructed using the STRING database. The Cytoscape plug-in MCODE was used to identify the top modules in the protein interaction network. The top modules were defined as core modules, which consisted of multiple shared differentially expressed genes. S3. Using the CytoHubba application in Cytoscape 3.8.2, we used a variety of different algorithms to select the top genes from each algorithm and sorted them according to their frequency of occurrence. We then crossed these genes with the shared differentially expressed genes in the core module, and finally obtained multiple hub differentially expressed genes. S4. Based on the DSigDB database, the above-mentioned hub differentially expressed genes were uploaded to the database, and multiple candidate drugs were predicted and ranked according to the comprehensive scores. The molecular docking of the TOP1 drug with the above-mentioned hub differentially expressed genes was performed using Autodock Tools; S5. LASSO regression analysis was performed on the above-mentioned hub differentially expressed genes to identify characteristic genes, and random forest analysis was performed to obtain the MeanDiscoveryGini values ​​of multiple differentially expressed genes, which were then sorted according to importance. The results of LASSO regression and random forest analysis were combined to finally determine the characteristic hub differentially expressed genes; S6. The ROC curve was used to evaluate the diagnostic performance of the characteristic hub differentially expressed genes. Finally, the characteristic hub differentially expressed genes could be used for comorbidity analysis and identification of the above diseases.

2. A method for identifying diseases by combining gene and comorbidity analysis according to claim 1, characterized in that: In step S1, differential expression was evaluated by DESeq2 to perform differentially expressed gene screening.

3. A method for identifying diseases by combining gene and comorbidity analysis according to claim 2, characterized in that: In step S1, differentially expressed genes were screened with >1 and FDR<0.05 as critical criteria.

4. A method for identifying diseases by combining gene and comorbidity analysis according to claim 1, characterized in that: In step S3, the algorithms used include MCC, MNC, EPC, stress, DMNC, degree, spacing, bottleneck, clustering coefficient, eccentricity, compactness and radiation.

5. The method for identifying diseases by combining gene and comorbidity analysis according to claim 1, characterized in that: In step S5, the parameters used in the LASSO analysis are α=1 and λ.

6. The method for identifying diseases by combining gene and comorbidity analysis according to claim 1, characterized in that: Use the "RandomForest" function to perform a random forest analysis, select the minimum error as the mtry node value, and select the image value that tends to be stable as ntree.

Images