2.1 Data from the GEO database
We obtained the transcriptome expression profiles GSE25504, GSE54514 and GSE13904 from the GEO database (Table.1). Additionally, the three datasets included 115 sepsis samples and 71 normal samples, which were used to identify DEGs between sepsis and normal groups.
2.2 Screening of DEGs
Three GEO series were merged and normalized by Perl. R packages “affy” and “limma” were employed to perform probe summarization and background correction of GSE25504, GSE54514 and GSE13904. The Benjamini-Hochberg method was used to adjust the original p-values. Fold-changes (FC) were calculated using the false discovery rate (FDR). The cut-off criteria for DEGs were |log FC| > 0.585 and p value < 0.05.
2.3 WGCNA analysis
The WGCNA analysis of all genes in the merged series was performed using R package “WGCNA” (https://cran.r-project.org/web/packages/WGCNA/index.html). A co-expression network for all genes were constructed, and the algorithm filtered the 25% genes with the largest variation for further analysis. Then, 115 sepsis samples and 71 normal samples were used for WGCNA analysis. The adjacency matrix was created by samples, and then it was transformed into a topological overlap matrix (TOM). Genes were divided into different gene modules using TOM-based difference measurement. The minimal gene module > 200 and the threshold to merge similar modules=0.1 were used to search modules that play an important role in sepsis.
2.4 Functional annotation of DEGs.
The GO and KEGG analyses are computational methods that evaluate gene functions and biological pathways. Metascape (http://metascape.org/gp/index.html) database can provide a comprehensive gene list annotation and analysis resource. Gene Set Enrichment Analysis (GSEA) (http://software.broadinstitute.org/gsea/index.jsp) is a computational method that could execute GO and KEGG analysis with a given gene list. In this study, GO and KEGG analyses of target genes were performed to explore the potential mechanism by which module genes promotes the progression of sepsis.
2.5 Construction and analysis of PPI network
Funrich (http://www.funrich.org/) is a biological analysis software. The DEGs in the yellow model genes were analysis by Funrich. STRING (http://string-db.org), an online database, could predict and provide the protein-protein interaction (PPI) network after importing the yellow model DEGs. Cytoscape is an analysis tool, which could provide biological network analysis and two-dimensional (2D) visualization for biologists. In our study, the PPI network were construct and analyzed by SRTING database and Cytoscape, and the hub DEGs was screened by four algorithms. And then, the Venn plots were used to intersect the four groups hub DEGs to obtain the common hub DEGs by Funrich.
2.6 Identification of hub DEGs associated with cardiovascular diseases
Comparative toxicogenomics database (CTD database, http://ctdbase.org/) is a web-based database. The relationship between gene/protein and disease could be predicted by the CTD database. In our study, the relationships between genes products and sepsis were analyzed by this database.
2.7 MiRNA of hub DEGs prediction
TargetScan (www.targetscan.org) is an online database that performs predictive analysis and identifies possible mechanisms for co-regulating the expression of hundreds of genes expressed in different cell types. In our study, TargetScan was used to screen for miRNAs that regulate the hub DEGs.
2.8 Statistical Analysis
Univariate and multivariate logistic regression analysis were perform by R. The ROC curves were provided by MedCalc software.
2.9 Identification of candidate molecule drugs for hub genes
Connectivity map (CMap, https://portals.broadinstitute.org/cmap/) database could connect genes and genomic information with human disease and drugs that treat them. In this study, the molecule drugs that could regulate hub genes was searched by CMap database.
2.10 Construction of mouse model of sepsis
This procedure used the C57BL/6 mice (8 weeks old). Animal adaptation feeding for 1 week. The mice could be anesthetized with inhaled isoflurane (Shanghai Civic Chemical Technology Co., Ltd, Shanghai). Shave the abdomen of the mouse and disinfect area by first applying betadine solution. A 1.5cm incision was made at the midline and the cecum was exposed. After ligation (6-0 PROLENE, 8680G; Ethicon), the cecum is perforated once or twice on the ipsilateral side of the cecum with a 19 needle. Gently squeeze the cecum with your hands to squeeze out a small amount of stool. We then return the cecum to the peritoneal cavity and close the peritoneum. Close the skin wound with a clamp. Mice were subcutaneously injected with preheated 0.9% normal saline. Food and water were not restricted .
This study was approved by the Ethics Committee of the Tianjin chest hospital. All experiments were approved by Animal Care and Use Committee of the Tianjin chest hospital. All institutional and national guidelines for the care and use of laboratory animals were followed.
2.11 ELISA assay
The used antigen was diluted to an appropriate concentration with coating diluent (generally, the required amount of antigen coating is 100μg per well), and 100μ L was added to each well, and placed at 37℃ for 4h. The liquid in the well was discarded. 5% calf serum was placed at 37℃ and sealed for 40min. When sealing, fill each reaction hole with sealing liquid and remove bubbles in each hole. After sealing, wash the hole with washing liquid for 3 times, 3min each time. The dilution of 1:200 is used for detection, and a larger dilution volume should be used to ensure the sample absorption amount>20ul.Add the diluted samples to the enzyme-conjugate reaction well, add at least two wells to each sample, 100μ L to each well, and place at 37℃ for 40-60min. Use washing solution to wash the well 3 times, 3min each time. Enzyme-labeled antibody was added at 37℃ for 30-60min. Add 100μL to each well. Enzyme-labeled antibody was added at 37℃ for 30-60min. Add 100μL to each well. Add stop solution 50μl to each well to stop the reaction, and determine the experimental results within 20min. After OPD color rendering, 492nm wavelength was used, and TMB reaction product detection required 450nm wavelength.
The main reagents were that: Mouse IL-6 ELISA kit (KE10007, proteintech, Rosemont, USA), Mouse tumor necrosis factor (TNF) alpha ELISA kit (KE10002, proteintech, Rosemont, USA), Mouse lactic acid ELISA kit (LS-F55829, LSBio, USA), Mouse HCK ELISA Kit (Abbexa, abx389525, UK), Mouse IL1RN ELISA kit (ab238258, abcam, Cambridge, UK).
2.12 RT-PCR assay
The extraction of total RNA from the peripheral blood of sepsis mice was made by RNA extraction kit (RN5301, Hangzhou Haoxin Biotechnology Co., LTD). The reaction solution for reverse transcription was configured, and the negative control was treated with DEPC water instead of RNA template. The reaction tubes were placed on the PCR amplifiers, denaturated at 37℃ for 60min and 70℃ for 15min. The reverse transcription products were stored at 4℃ for later use. Each PCR reaction tube was put into the DNA amplifier at 94℃ for 5min, and then thermal cycling was conducted under the following conditions: 94℃, 45s; 56 ℃, 30 s; 72℃, 30s. Repeat this cycle 30 times, and extend for 10min at 72℃. RT-PCR products were detected by 1%-2% agarose gel electrophoresis. PCR product 8μ L was mixed with sample loading buffer 2μ L, and DNA Marker of appropriate size was used for electrophoresis separation. Ethidium bromide was stained, and the results were recorded and analyzed by image analyzer.
Primers of HCK, IL1RN and their sequences for PCR analysis were followed:
2.13 Immunofluorescence assay
The slides were washed successively in xylene Ⅰ15min-xylene Ⅱ 15min-anhydrous ethanol Ⅰ 5min-anhydrous ethanol Ⅱ 5min-85% alcohol 5min-75% alcohol 5min-distilled water. The slides were placed in a repair box filled with EDTA antigen repair buffer (PH8.0) for antigen repair. After natural cooling, the slides were placed in PBS (PH7.4) and washed by shaking on a decolorizing shaker for 3 times, and 5min each time. Add BSA drops and incubate for 30min. The sealing solution was gently removed, and the prepared primary antibody was added to the sections with drops in a certain proportion. The sections were placed horizontally in a wet box at 4°C for incubation overnight. The slides were placed in PBS (PH7.4) and washed 3 times on a decolorizing shaker, 5min each. The slices were dried by shaking, then the second antibody was added to cover the tissue in a certain proportion, and incubated at room temperature for 50min. DAPI was used to restain the nuclei and incubate for 10min at room temperature, away from light. The self-quenching agent was added for 5min, and the water was rinsed for 10min.The slides were placed in PBS(PH7.4) and washed 3 times on a decolorizing shaker, 5min each. The slices were briefly shaken dry and sealed with anti-fluorescence quenching sealing tablets. The slices were placed under a scanner for image collection or a fluorescence microscope for photo taking. The DAPI stained nucleus is blue under ultra violet excitation, and the positive expression is corresponding fluorescein labeled red.
The main reagents were that: Mouse Anti-HCK antibody (ab61055, abcam, Cambridge, UK); Anti-IL1RN antibody (ab175392, abcam, Cambridge, UK).
2.14 Construction of BP neural network model
The calculation process of BP neural network consists of forward calculation process and reverse calculation process. Neural network is a network formed by connecting multiple neurons together according to certain rules. In the forward propagation process, the input mode is processed layer by layer from the input layer through the hidden unit layer and then transferred to the output layer. The state of neurons in each layer only affects the state of neurons in the next layer. If the desired output cannot be obtained at the output layer, reverse propagation is carried out to return the error signal along the original connection path, and the error signal is minimized by modifying the weights of each neuron. In this study, the input value of neural network is the relative expression value of HCK and IL1RN, and the output variable is the expression quantity of IL6. Define a BP neural network class, set network parameters. A BP network with 3 dimensions of output and 1 dimension of output is constructed with 3 hidden layers (10 nodes for each hidden layer). When the BP neural network is initialized, the weight, weight momentum and initial error of network nodes of each layer are initialized. Introduction of learning and training data; Input and output data iterated 3000 times. 3000 times continuously forward layer by layer to compute the output node data. At the same time, the error is calculated layer by layer and the weight value is reversed until the iteration is completed. Note that the error function must be decreasing. Introduce the data to predict the results and bring the data back to the model.
2.15 Build the support vector machine model among HCK, IL1RN and IL6
Support Vector Machine (SVM) is a kind of generalized linear classifier that classifies data by supervised learning method. The decision boundary is the maximum-margin hyperplane solved for the learning sample. SVM uses hinge loss function to calculate empirical risk and adds regularization term to the solving system to optimize structural risk. It is a classifier with sparsity and robustness. SVM is one of the common kernel learning methods, can be used for nonlinear classification by kernel method. In this study, the input value of neural network is the relative expression value of HCK and IL1RN, and the output variable is the expression quantity of IL6.
2.16 Statistical analysis
The data was expressed as mean ± SD. Independent-samples T test was used, and when the equal variances not assumed, Brown-Forsythe was performed. All statistical analyses were conducted using SPSS software, version 24.0 (IBM Corp., Armonk, NY, USA), and MATLAB (R2014a, MathWorks. Inc, New Mexcico, USA). A p-value < 0.05 was considered statistically significant.