A pathomics model for predicting lactate-related subtypes and their biological mechanisms in lung adenocarcinoma

Background Lung adenocarcinoma (LUAD) exhibits considerable heterogeneity in postoperative survival, underscoring the need for robust prognostic biomarkers. Given the emerging role of lactate metabolism in cancer progression, this study aimed to develop a pathomics model for prognostic stratification by predicting lactate-related subtypes. Methods Using data from TCGA-LUAD project, we performed non-negative matrix factorization to stratify patients based on lactate-related genes. A lactate-related prognostic model was constructed via LASSO-Cox regression. Patients were then randomly split into training and validation sets to develop and evaluate an XGBoost-based pathomics model for subtype prediction. GSVA, immune cell infiltration analysis, and mutational landscape analysis, were further conducted to explore the underlying mechanisms. Results A four-gene lactate-related prognostic signature (GAPDH, PGK1, SOCS3, and CXCL9) was constructed to define high-risk and low-risk subtypes. The pathomics model achieved an AUC of 0.825 (95% CI: 0.773–0.877) in the training set and 0.700 (95% CI: 0.592–0.808) in the validation set for predicting lactate-related subtypes. The calibration curve indicated excellent agreement between the pathomics model’s predictions and actual observations. A high pathomics score (PS) was identified as an independent risk factor for overall survival in LUAD. Moreover, the high-PS phenotype, characterized by elevated M2 macrophage infiltration and a higher TP53 mutation rate, exhibited significant enrichment of tumor-related pathways. Conclusions We developed a pathomics model for predicting lactate-related subtypes in LUAD. The model showed robust predictive performance, providing a useful tool for prognostic stratification and immunotherapy guidance.