Identification of the platelet related DEGs (plt-DEGs)
A total of 3770 DEGs were found through between Sepsis and Control comparison, including 2062 up- and 1708 down-regulated DEGs (Supplementary Table 1). The distribution and between Sepsis and Control comparison of the genes were shown in Fig. 1A&B. Moreover, in order to screen platelet-related genes from the above DEGs, we used the Venn tool to intersect them with 351 platelet-related genes to obtain differential platelet genes (Supplementary Table 2). As shown in Fig. 1C, a total of 85 plt-DEGs were obtained, including 56 up- and 29 down-regulated plt-DEGs. Among them, THBS1 was plt-DEGs with upregulation pattern (FC = 0.76, adjusted p-value = 0.0000315).
Function Enrichment Analysis Of Plt-degs
The results from enrichment analyses revealed that a total of 100 GO BPs, 37 GO CCs, 27 GO MFs and 33 KEGG pathways were enriched (Supplementary Table 3). Among them, immune-related pathways such as B cell receptor signaling pathway, T cell receptor signaling pathway, innate immune response, positive regulation of B cell proliferation and neutrophil chemotaxis were found, and platelet-related pathways included platelet activation, platelet degranulation, platelet formation, and platelet-derived growth factor receptor signaling are enriched. Moreover, platelet aggregation, platelet activating factor biosynthetic process, cellular response to platelet-derived growth factor stimulus and positive regulation of platelet activation were enriched (Fig. 2A-D). In addition, THBS1 was found to be involved in platelet degranulation, response to drug and activation of MAPK activity.
Ppi Network Construction
In order to further observe the interaction relationship between the above-mentioned plt-DEGs, we combined the STRING database to predict and analyze whether there is an interaction relationship between the proteins encoded by the genes. As a result, 331 PPI pairs were obtained, including 76 protein nodes, and the detailed PPI network was shown in Fig. 3. In addition, THBS1 was found to be interacted with ORM1, SRGN, HGF and TIMP1.
Prognosis Model Construction And Verification
The relationship between plt-DEGs and survival statue
In order to explore the relationship between plt-DEGs and survival statue, the expression level of plt-DEGs in 34 sepsis death samples and 68 alive samples were evaluated. Finally, a total of 16 genes showed a significant correlation between death and alive samples (Fig. 4A).
Lasso Regression Analysis
Based on the above obtained 16 survival related plt-DEGs, we adopted LASSO (least absolute shrinkage and selection operator) algorithm to screen for prognostic markers using the GSE95233 dataset. As shown in Fig. 4B. finally 10 survival correlated prognostic markers were found: PLA2G4A, GNAQ, PIK3CB, LHFPL2, SCCPDH, PRKCD, VEGFA, CCNA2, PRKDC and SLC9A3R1, and the corresponding coefficient parameters were 0.040911573, 0.6538434, 2.699257553, 1.045291704, 0.239707039, -2.157664384, 1.497152256, 0.225601144, 2.194029479 and − 0.646211922.
Prognosis Prediction Model Establishment
The ROC curve was further drawn to evaluate the predictive efficacy of this prognostic model. The results are shown in Fig. 4C, where the AUC was above 0.9 using the training dataset, indicating that good diagnostic efficacy.
In order to further verify the efficacy of the model, we employed dataset GSE54514 (validation set), and the AUC value of this prognostic model in the validation set is above 0.6, which further illustrates the effectiveness of the diagnostic model to distinguish between death and alive.
Risk Classification Based Go And Kegg Analyses
Based on the eigenvalues of all sepsis disease samples in the above dataset GSE95233, all disease samples were divided into high and low risk groups according to the median value of the eigenvalues (34.79136573), of which 51 samples were in the high risk group and 51 in the low risk group. According to the normalized p-value, the TOP5 of the up- and down-regulated GO and KEGG pathway enrichment results of each phenotype were shown in Fig. 5A-D.
Immune Cell Infiltration Analyses
According to the scores of each immune cell from all samples of the training set, the score heat map of 22 kinds of immune cells is shown in Fig. 6A. Furthermore, as shown in Fig. 6B, there are 10 types of immune cells with significantly difference (p < 0.05) in sepsis and normal samples: B cells naïve, plasma cells, T cells CD8, T cells gamma delta, NK cells resting, monocytes, macrophages M0, mast cells resting, eosinophils and neutrophils. Moreover, the relationship between survival related plt-DEGs and the immune cells were also illustrated in Fig. 6C. A total of 43 correlated gene-cell pairs were found.
The Expression Of Survival Related Plt-degs In Training And Validation Set
In order to further understand the expression of the above 10 survival related plt-DEGs between sepsis and normal samples in both dataset GSE95233 (training set) and the validation sets GSE69528 and GSE28750. As shown in Fig. 7A&C, all the 10 DEGs showed significant changes in all 3 datasets. Except SLC9A3R1, other 9 DEGs showed upregulated trend.
In Vivo Verification Of The Expression Thbs1
Compared with the control group, the expression level of THBS1 from the serum of LPS mice was significantly decreased (P < 0.001, Fig. 8). These data indicated possible critical role of THBS1.