Abstract 718P
Background
Bioinformatics tools were used to identify prognosis-related molecular subtypes and biomarkers of hepatocellular carcinoma (HCC).
Methods
The TCGA datasets and GEO datasets (GSE14520, GSE76427, and GSE25097) were screened for differentially expressed genes (DEGs) between HCC and normal tissues. DEGs in the same direction across the four datasets were analyzed for enrichment. Non-negative matrix decomposition to identify subtypes of HCC with different prognosis. Cox regression and Kaplan-Meier curve analyses were performed to identify overlapping DEGs associated with survival defined as prognosis-related genes. An area under the curve > 0.80 of genes used to construct random survival forest and least absolute shrinkage and selection operator (LASSO) models to identify feature genes. We constructed a Gaussian mixture model (GMM) to identify feature genes with ability to diagnose HCC recurrence. Key gene associated with OS were determined by univariate Cox regression analysis. Nomograms mode was used to evaluate the predictive power. The mutation and methylation of key gene were analyzed in TCGA. The relative levels of immune cell infiltration were determined by single-sample gene set enrichment.
Results
Four datasets identified 3,330 DEGs in the same direction that were involved in cell cycle, and FOXO signaling pathway. Subtype C2 showing better overall survival than subtype C1. Seven feature genes (SORBS2, DHRS1, SLC16A2, RCL1, IGFALS, GNA14, and FANCI) that may be involved in HCC occurrence and prognosis. A univariate Cox model identified FANCI as a key gene involved mainly in the cell cycle, and mismatch repair. FANCI had two mutation sites and may undergo methylation. ssGSEA showed that Th2 and Th cells are significantly high-infiltrated in HCC patients.
Conclusions
We defined two molecular subtypes of HCC that are associated with different prognosis, and we identified FANCI as a good prognostic indicator in HCC.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
National Natural Science Foundation of China (NO. 81803007, 82060427, 82103297), Guangxi Key Research and Development Plan (NO. GUIKEAB19245002), Guangxi Scholarship Fund of Guangxi Education Department, Guangxi Natural Science Foundation (NO. 2020GXNSFAA259080), Guangxi Medical University Training Program for Distinguished Young Scholars, Science and Technology Plan Project of Qingxiu District, Nanning (NO. 2020037, 2020038).
Disclosure
All authors have declared no conflicts of interest.