Screening candidate DEARGs in PAH
In this study, a total of 580 differentially expressed genes (DEGs: 485 upregulated and 95 downregulated) were identified between PAH and control samples in the GSE113439 dataset, using |log2FC| > 1 and adjusted P < 0.05 as cutoff values (Fig. 2A). The 580 DEGs were compared with 502 ARGs in the Aging Atlas database, and 23 genes were identified in both datasets (Fig. 2B, 2C). Among these 23 genes, 22 were upregulated (RB1CC1, TOP1, WRN, TOP2B, CDK1, FGF7, HSPA9, HIF1A, CLOCK, NBN, ITGA2, EPS8, CFLAR, CCL20, IGF1, MMP1, CXCL8, BIRC3, SERPINB2, TOP2A, HSP90AA1, HSPD1), and 1 were downregulated (PECAM1; Fig. 3).
Validation DEARGs in GSE53409 dataset
Expression changes of the 23 genes in GSE113439 and GSE53409 are listed in Table 1. PECAM1 and CFLAR were not involved in the differential genes in the dataset GSE53408, and the other 21 genes were equally highly expressed in PAH patients. Therefore, we will carry out the next research on these 21 differential genes.
Table 1
The analysis of 23 aging- and PAH-related genes in GSE113439 and GSE53408 datasets.
Gene | | GSE113439 | | | | GSE53408 | |
Log2FC | P value | Type | Log2FC | P value | Type |
PECAM1 | -1.213 | 0.000 | Down | | ---- | ---- | ---- |
RB1CC1 | 1.014 | 0.000 | Up | | 1.056 | 0.000 | Up |
TOP1 | 1.073 | 0.000 | Up | | 1.066 | 0.000 | Up |
WRN | 1.076 | 0.000 | Up | | 1.075 | 0.000 | Up |
TOP2B | 1.089 | 0.000 | Up | | 1.122 | 0.000 | Up |
CDK1 | 1.095 | 0.001 | Up | | 1.215 | 0.000 | Up |
FGF7 | 1.096 | 0.000 | Up | | 1.376 | 0.000 | Up |
HSPA9 | 1.111 | 0.000 | Up | | 1.165 | 0.000 | Up |
HIF1A | 1.152 | 0.000 | Up | | 1.262 | 0.000 | Up |
CLOCK | 1.182 | 0.000 | Up | | 1.155 | 0.000 | Up |
NBN | 1.193 | 0.000 | Up | | 1.184 | 0.000 | Up |
ITGA2 | 1.200 | 0.000 | Up | | 1.179 | 0.000 | Up |
EPS8 | 1.261 | 0.000 | Up | | 1.35 | 0.000 | Up |
CFLAR | 1.269 | 0.000 | Up | | ---- | ----- | ---- |
CCL20 | 1.270 | 0.020 | Up | | 1.545 | 0.005 | Up |
IGF1 | 1.283 | 0.000 | Up | | 1.491 | 0.000 | Up |
MMP1 | 1.319 | 0.000 | Up | | 1.601 | 0.000 | Up |
CXCL8 | 1.357 | 0.000 | Up | | 1.599 | 0.001 | Up |
BIRC3 | 1.362 | 0.000 | Up | | 1.397 | 0.000 | Up |
SERPINB2 | 1.441 | 0.017 | Up | | 1.634 | 0.007 | Up |
TOP2A | 1.476 | 0.000 | Up | | 1.672 | 0.000 | Up |
HSP90AA1 | 1.584 | 0.000 | Up | | 1.643 | 0.000 | Up |
HSPD1 | 1.600 | 0.000 | Up | | 1.626 | 0.000 | Up |
Functional enrichment analysis of DEARGs
Five most significantly enriched GO terms are shown in Fig. 4A, 4B and 4C. The enriched biological processes were DNA conformation change, cell cycle checkpoint and DNA integrity checkpoint. The enriched cellular components were axon part, chromosome, telomeric region and replication fork. The enriched molecular functions were ATPase activity, catalytic activity, acting on DNA and ATPase activity, coupled. The KEGG enrichment analysis showed that the differentially expressed aging-related genes played a key role in IL-17 signaling pathway, progesterone-mediated oocyte maturation, rheumatoid arthritis and platinum drug resistance (Fig. 4D).
Analysis of PPI networks and identifification of hub genes
Construct PPI networks using STRING database to identify interactions between DEARGs and further visualize the results comprising 19 nodes and 41 edges via Cytoscape software. Figure 5A shows PPI network of DEARGs. Then, hub genes were explored using eight algorithms of the Cytohubba plug-in (Fig. 5B). The first 8 genes obtained by MCC, MNC, DMNC, EPC, BottleNeck, EcCentricity and Closeness were intersected to obtain 6 hub genes, namely TOP1, TOP2A, HSP90AA1, HIF1A, CDK1 and CXCL8 (Table 2).
Table 2
Top eight hub genes obtained by seven algorithms of Cytohubba
BottleNeck | Closeness | DMNC | EcCentricity | EPC | MCC | MNC |
HIF1A | CXCL8 | TOP2B | HIF1A | HIF1A | CDK1 | HSP90AA1 |
HSP90AA1 | HIF1A | WRN | CDK1 | HSP90AA1 | TOP1 | CXCL8 |
MMP1 | HSP90AA1 | TOP2A | HSP90AA1 | CXCL8 | TOP2A | HIF1A |
TOP2A | MMP1 | CDK1 | TOP1 | CDK1 | CXCL8 | CDK1 |
CXCL8 | CDK1 | TOP1 | HSPD1 | TOP1 | WRN | TOP1 |
HSPD1 | TOP1 | CCL20 | CXCL8 | MMP1 | TOP2B | MMP1 |
WRN | IGF1 | SERPINB2 | MMP1 | IGF1 | HSP90AA1 | IGF1 |
IGF1 | TOP2A | FGF7 | TOP2B | TOP2A | HIF1A | TOP2A |
Identification and validation of diagnostic feature biomarkers
We verified the expression level of hub gene in validation set GSE53408. Figure 6A-F shows the expression levels of six hub genes in GSE53408. The gene expression difference between PHA group and control group was consistent with dataset GSE113439, and the difference was statistically significant, among which the p-values of TOP1, TOP2A, HSP90AA1 and HIF1A were all < 0.001, and the p-values of CDK1 and CXCL8 were all < 0.01. The diagnostic value of six hub genes in PAH was measured by receiver operating characteristic (ROC) curves (Fig. 6G-L). The results show that TOP1(area under the curve (AUC): 1.000) and TOP2A (AUC: 1.000), HSP90AA1 (AUC: 1.000), HIF1A(AUC: 0.977) have high accuracy. CDK1 and CXCL8 had slightly lower AUC of 0.939 and 0.864, respectively. These results suggest that TOP1, TOP2A, HSP90AA1 and HIF1A may be the ultimate diagnostic genes for PAH.
Prediction of diagnostic genes target miRNAs and construction of miRNA-mRNA network
The NetworkAnalyst tool was used to predict target miRNAs of the diagnostic genes (TOP1, TOP2A, HSP90AA1 and HIF1A). Ultimately, we obtained 192 miRNAs of 4 diagnostic genes. TOP1 was regulated by 8 miRNAs, TOP2A was regulated by 17 miRNAs, HSP90AA1 was regulated by 100 miRNAs and HIF1A was regulated by 84 miRNAs, and these four genes were simultaneously regulated by hsa-miRNA-186-5p (Fig. 7).