Lactylation-based machine algorithm combined with multi-omics analysis to predict prognosis in cervical cancer.

Although lactylation has been investigated in cancer biology, its mechanistic role in cervical cancer remains unclear. This study integrated RNA-sequencing data from TCGA, three GEO datasets, and single-cell data (GSE44001) to identify lactylation-associated genes (LAGs) involved in cervical cancer. Differential expression analysis, WGCNA, and lactylation-related gene sets were combined to identify candidate genes. Multiple machine learning algorithms were employed to construct a prognostic model, which was further validated using Cox regression, receiver operating characteristic analysis, immune infiltration profiling, functional enrichment, and cell-cell communication analysis. A total of 43 overlapping co-expressed genes were identified, and 14 LAGs strongly associated with prognosis were incorporated into a risk-scoring system. The model demonstrated robust predictive performance and enrichment in pathways associated with carbon metabolism and glycolysis, with notable immune differences between risk groups, particularly in mast cells and neutrophils. Drug sensitivity analysis showed positive correlations between the risk score and IC(50) values of paclitaxel and rapamycin, and a negative correlation with midostaurin. Mendelian randomization revealed a causal association between HMGN1 and cervical cancer risk. In vitro assays demonstrated that HMGN1 inhibition significantly suppressed SiHa and HeLa cell proliferation and induced S-phase arrest, highlighting its potential as a therapeutic target. In conclusion, this study developed a reliable LAG-based prognostic model and uncovered key lactylation-related mechanisms in cervical cancer, providing new insights for biomarker discovery and personalized therapeutic strategies.