A novel integrative machine learning-based prognostic model reveals lactylation regulation in hepatocellular carcinoma progression.

BACKGROUND: Hepatocellular carcinoma (HCC) is one of the most common malignant liver tumor with poor clinical outcomes. Accumulated evidence has demonstrated lactylation plays a vital role in the metabolic reprogramming. However the mechanisms underlying the role of lactylation in the regulation of HCC progression remain largely unknown. This study aims to construct a prognostic model based on lactylation-related metabolism genes, and further explore its prognostic significance and biological function in HCC. METHODS: In this study, a robust prognostic prediction model has been constructed employing a complex machine learning framework using public bulk RNA-seq and proteomic HCC dataset. Moreover, the clinical application of this model was explored, and its biological functions were validated using several in vitro experiments. Subsequently, we performed functional analysis, survival analysis, tumor immune microenvironment analysis and drug sensitivity to demonstrate our model's potential in translational cancer medicine. RESULTS: We developed an integrative machine learning-based computational framework to generate a predictive Metabolism-related Lactylation Index (MRLI) within four independent HCC cohorts and validated its prognostic accuracy through various algorithms. Notably, compared to published gene signatures, MRLI demonstrated robust predictive capability. In addition, single-cell analysis demonstrated that the MRLI is predominantly localized within HCC cells and correlates with tumor malignancy. Mechanistically, Gene Set Enrichment Analysis (GSEA) suggested that the MRLI may be associated with cellular proliferation and metabolic reprogramming, which was further confirmed by experimental evidence. Subsequently, public spatial transcriptomics and bulk RNA-seq analysis revealing that the highly MRLI predicts a tumor immunosuppressive microenvironment, which was further verification in a cohort of 40 HCC samples by multiple immunofluorescence. Additionally, groups with highly MRLI showed decreased sensitivity to sorafenib, immune checkpoint inhibitors, and TACE, highlighting the potential of MRLI in facilitating personalized treatment strategies. CONCLUSION: Our study has developed a novel MRLI as a predictive marker for prognosis and therapeutic outcomes in patients with HCC. These findings indicate that lactylation promotes malignant cell phenotypes and immune microenvironment remodeling partially through metabolic regulation, suggesting it as a potential clinical therapeutic target.