Machine learning-based integration develops a lactate metabolism related gene signature for improving outcomes in pancreatic ductal adenocarcinoma.

OBJECTIVES: Given its high global mortality rate, pancreatic ductal adenocarcinoma (PDAC) remains a significant area of investigation. However, a robust gene signature linked to lactate metabolism for PDAC patients has not yet been established. Our objective was therefore to construct a novel lactate metabolism related gene signature (LMRGS) capable of predicting patient outcomes and informing therapeutic decisions. METHODS: Genes associated with lactate metabolism were sourced from the Molecular Signatures Database (MsigDB). The LMRGS was constructed using distinct algorithmic combinations and its performance was subsequently verified in 8 separate patient cohorts. Multiomics analyses were employed to evaluate the signature's impact on biological functions and to investigate its relationship with the immune microenvironment. EdU, colony formation and wound-healing assays were used to demonstrate the effects of lactate on pancreatic cancer cells. RESULTS: An artificial intelligence framework enabled the creation of an LMRGS that serves as an independent prognostic predictor for individuals with PDAC. This signature demonstrated considerable accuracy in forecasting overall survival. When patients were stratified into high- and low-risk groups, the high-risk group showed reduced immune cell infiltration and a poorer response to immunotherapy. Further investigation confirmed a strong correlation between the LMRGS and the immune milieu in PDAC. In vitro experiments demonstrated that lactate promotes the proliferation and migration of pancreatic cancer cells. CONCLUSION: We have formulated a new LMRGS for PDAC which holds potential for informing personalized treatment plans. Interventions aimed at the lactate metabolic pathway could represent a promising strategy to boost therapeutic effectiveness and extend survival for patients diagnosed with this disease.