Machine learning-based biomarker discovery for melanoma-related distant metastasis 2766

Abstract Description Objective Melanoma is a lethal type of skin cancer that causes high mortality and morbidity in patients due to distant metastasis. Identifying specific proteins and genetic biomarkers that can improve the prognosis for melanoma patients and predict and prevent distant metastasis is crucial. Methods In this study, we employed machine learning techniques to systematically analyze multi-omics datasets from patient samples and discovered multiple genes and pathways differentially expressed in metastatic tumors compared to non-metastatic tumors. To further validate a set of potential biomarkers, we assessed the metastatic potential of several melanoma cell lines, including A375, SK-MEL-2, and A431, by cell invasion and migration assays and measured the gene expression/protein level of newly identified markers in these cell lines. Results Using these approaches, we have identified proteins Histocompatibility Minor Antigen HA-1, Aurora-ABC, U-Histone, and genes CALML, PI3, RHBDD2, HLA-DQA1, PLA2G2D, and TGM2 as new potential genetic and proteomic biomarkers associated with high-risk distant metastasis of melanoma patients. Conclusion By integrating machine learning techniques with biological validation, this study has the potential to help improve the clinical management of melanoma patients and design more specific targeted therapies that can inhibit the disease progression and increase the survival rate of melanoma patients. Topic Categories Computational and Systems Immunology (COMP)