Integrating pathology genomics and single-cell genomics to identify lactate metabolism-related prognostic features and therapeutic strategies for melanoma.

Cutaneous melanoma (SKCM) is highly malignant and prone to developing treatment resistance. Lactate metabolism in the tumor microenvironment (TME) plays a crucial role in SKCM progression, immune evasion, and therapy resistance. This study aimed to integrate multi-omics data to systematically characterize the molecular features of lactate metabolism in SKCM, construct an effective prognostic model, and explore potential therapeutic strategies. Quantitative pathological features were extracted using CellProfiler and combined with deep learning features obtained from a pre-trained ResNet50 convolutional neural network. Gene set variation analysis (GSVA) was used to calculate lactate metabolism scores and identify associated pathological features. Single-cell RNA sequencing was applied to assess lactate metabolic activity across different cell types. These data, together with spatial transcriptomics, genomic alterations, immune infiltration profiles, and immunotherapy response data, were integrated to construct a lactate metabolism signature (LMS) prognostic model (comprising 3 pathological features and 11 genes). The model was developed using 101 combinations of 10 machine learning algorithms. Furthermore, RAB32 knockdown experiments were performed to verify its effects on melanoma cell proliferation, migration, invasion, and metabolism. A total of 443 pathological imaging features significantly associated with lactate metabolism were identified. Single-cell analysis revealed that melanoma cells exhibited the highest lactate metabolic activity, with markedly enhanced intercellular communication in the high-metabolism group. The LMS model demonstrated excellent prognostic performance in both the TCGA training and validation cohorts. Patients in the high-LMS group had significantly shorter survival, showed immune evasion features, and exhibited activation of melanoma-related metabolic and signaling pathways (e.g., oxidative phosphorylation). In contrast, the low-LMS group had stronger immune infiltration and higher expression of immune checkpoint molecules. The key gene RAB32 was significantly correlated with all lactate metabolism-related pathological features, was highly expressed in the tumor core, and its high expression predicted poor prognosis. RAB32 knockdown markedly inhibited melanoma cell proliferation, migration, and invasion; reduced lactate production; suppressed the expression of glycolytic enzymes and lactate transporters; and decreased extracellular acidification rate (ECAR) and oxygen consumption rate (OCR). In addition, it significantly inhibited tumor growth in mouse xenograft models. This study developed a multi-omics-integrated prognostic model (LMS) based on lactate metabolism, providing a novel tool for risk stratification and therapeutic decision-making in SKCM patients. It also identified RAB32 as a central player in tumor metabolic reprogramming and invasiveness, with promising potential as a therapeutic target.