Development and experimental validation of a machine learning-based disulfidptosis-related ferroptosis score for hepatocellular carcinoma

Abstract

Disulfidoptosis and ferroptosis are two distinct programmed cell death pathways that have garnered considerable attention due to their potential as therapeutic targets. However, despite their significance of these pathways, the role of disulfidoptosis-related ferroptosis genes in hepatocellular carcinoma (HCC) remains unclear. In this study, we employed a comprehensive approach that utilized various sophisticated techniques such as Pearson analysis, differential analysis, uniCox regression, lasso, ranger, and multivariable Cox regression to develop the disulfidoptosis-related ferroptosis (DRF) score. We then classified patients with HCC into high- and low-score groups to examine the association between the DRF score and various outcomes, including prognosis, functional enrichment, immune infiltration, immunotherapy, TACE sensitivity, drug sensitivity, and single-cell level function. Finally, we conducted in vitro experiments to validate the function of KIF20A. Our analysis revealed that KIF20A, G6PD, SLC7A11, and SLC2A1 were integral to constructing the DRF score. Our findings showed that patients with low DRF scores had significantly better prognoses and were more responsive to immunotherapy, TACE, and chemotherapy than those with high DRF scores. Based on our results obtained from bulk RNA-seq, single-cell RNA-seq, and in vitro experiments, we identified the cell cycle pathway as the primary distinguished factor between high-score and low-score groups. This study sheds light on the contribution of disulfidoptosis-related ferroptosis genes to the development and progression of HCC. The information gleaned from this study can be leveraged to improve our understanding of their potential as therapeutic targets for HCC treatment.

Keywords: Disulfidoptosis; Ferroptosis; HCC; Machine learning; Prognosis.