Mutation-informed gene pairs to predict melanoma metastasis

Abstract Background Metastasis causes over 90% of cancer-related deaths, including melanoma. However, most anti-cancer treatments focus on reducing tumor size rather than preventing metastatic spread. Therefore, there is a need to identify robust biomarkers that can predict and inhibit metastatic progression without inducing tumor cell death. Methods We introduce the novel concept of synthetic anti-metastasis (SAM), which builds on the idea of synthetic lethality (SL). SAM pairs are interactions whose simultaneous impairment suppresses metastasis without inducing cell death. We identified preliminary SAM pairs using somatic mutation and clinical data from The Cancer Genome Atlas (TCGA). We selected the final SAM pairs by excluding previously reported SL interactions and pairs having at least one essential gene from preliminary pairs. We validated these SAM pairs across multiple datasets and tested their clinical relevance using survival analysis and machine learning (ML). Candidate anti-metastatic drugs for melanoma were identified through LINCS-based gene signature analysis, network analysis, and literature review. Results We identified 325 final SAM pairs from 367 preliminary pairs. We found that patients with a high number of co-impairment or -inactivation of SAM pairs showed improved overall survival and reduced metastasis. The ML model based on SAM gene features accurately distinguished primary from metastases melanoma samples (AUROC: 0.940; HR: 0.724), outperforming models built from other known melanoma metastasis-associated genes. Finally, we discovered five compounds - MLN2480, pifithrin-µ, RO4929097, trametinib, and sorafenib - as potential anti-metastatic drugs for melanoma. Conclusions This study provides SAM pairs as a novel type of biomarkers that could predict metastatic melanoma prognosis and as therapeutic targets in terms of reducing metastasis risk. Our framework to identify SAM pairs could offer a data-driven strategy to improve the prediction and discover potential treatment for melanoma metastasis through integrated genomic, transcriptomic, and pharmacogenomic analysis.