An Anoikis-related Gene Signature for Forecasting Prognosis, Tumor Microenvironment, and Treatment Response in Hepatocellular Carcinoma

Abstract Background: As hepatocellular carcinoma (HCC) is molecularly heterogeneous, this study aimed to generate an anoikis-related gene signature for prognosis prediction and precision medication. Methods: In the TCGA-LIHC cohort, anoikis-related genes with prognostic values were determined by univariate Cox regression, followed by prognostic model construction based on the LASSO-Cox method. Independent prognostic factors were integrated to develop a nomogram. The diagnostic performances of the established model and the nomogram were assessed in training (TCGA-LIHC) and validation (GSE14520 and ICGC-LIRP-JP) cohorts using time-dependent receiver operating characteristic curve (ROC), Kaplan-Meier (KM) plot, calibration plot, principal components analysis (PCA), concordance index, and decision curve analysis (DCA). Seven published methods such as CIBERSORT were used to analyze the tumor microenvironment. The drug sensitivity was analyzed using Tumor Immune Dysfunction and Exclusion (TIDE) algorithm, IMvigor210 cohort, CellMiner database, and "oncoPredict" package. Results: Eight genes were included in the signature, including PTRH2, ITGAV, BSG, DAP3, SKP2, LGALS3, SFN, and PLK1, which could successfully divide patients into two subgroups with survival differences. The nomogram comprised of the TNM stage and our gene signature could further improve predictive accuracy. The patients in the low-risk group could benefit less more from immunotherapy, and this gene signature was also conducive to tailoring chemotherapy and molecular-targeted therapy. Conclusions: The anoikis-related gene signature provides reliable evidence for prognosis prediction, treatment optimization, and new therapeutic targets for HCC patients.