Comprehensive analysis of an autophagy-related prognostic model for predicting survival based on TCGA and ICGC database in hepatocellular carcinoma patients

Abstract

There is accumulating evidence that autophagic activity is crucial to the development of hepatocellular carcinoma (HCC). Thus, we sought to develop a predictive model based on autophagy-related genes (ARGs) to forecast the prognosis of HCC patients. Based on expression data from The Cancer Genome Atlas (TCGA) and ARGs from Human Autophagy Database (HADb), the differentially expressed ARGs were screened. The prognosis-related ARGs were identified using a univariate Cox regression analysis. Using multivariate Cox regression analysis, a prognostic model was developed. To assess the predictive value of the model, receiver operating characteristic (ROC) curve, Kaplan-Meier curve, and multivariable Cox regression analyses were conducted. A data cohort gathered independently from the International Cancer Genome Consortium (ICGC) database further verified the model's predictive accuracy. The immune landscape was generated using the TIMER and CIBERSORT algorithms. Finally, the correlation between the prognostic signature and gene mutation status was analyzed by employing "maftools" package. We identified a novel prediction model based on the ARGs of PLD1 and SLC36A1 with significant prognostic values for HCC in both univariate and multivariate Cox regression analysis, and patients were classified into high- or low-risk groups based on their risk scores. High-risk patients had significantly shorter overall survival (OS) times than low-risk patients (P=5e-4). According to the ROC curve analysis, the risk score had a higher predictive value than the other clinical characteristics. Prognostic nomograms were also performed to visualize the relationship between individual predictors and survival rates in patients with HCC. Further, an external independent cohort of ICGC patients provided additional confirmation of the predictive efficacy of the model. We subsequently analyzed the differential immune densities of the two groups and discovered that various immune cells, including naïve B cells, resting memory cluster of differentiation (CD)4 T cells, regulatory T cells, M2 macrophages, and neutrophils, had considerably larger infiltrating densities in the high-risk group than the low-risk group. We established a robust autophagy-related risk model having a certain prediction accuracy for predicting the prognosis of HCC patients. Our findings will contribute to the definition of prognosis and establishment of personalized treatment interventions for HCC patients.