Identification of Biomarkers and Target Drugs for Melanoma: A Topological and Deep Learning Approach

Xiwei Cui ^1,2 Jipeng Song ^3* Qingfeng Li ^4* Jieyi Ren ^4*

• ¹ Research Center of Plastic Surgery Hospital, Chinese Academy of Medical Science & Peking Union Medical College, Beijing, P. R. China., Beijing, China
• ² Key Laboratory of External Tissue and Organ Regeneration, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, P. R. China., Beijing, China
• ³ Comprehensive Ward of Plastic Surgery, Plastic Surgery Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College., Beijing, China
• ⁴ Department of Plastic and Reconstructive Surgery, Shanghai Ninth People’s Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China

Melanoma is a highly malignant cancer with rapid metastasis and high mortality rates, primarily affecting the skin. Despite advancements in surgery, immunotherapy, and targeted treatments, the prognosis for advanced melanoma remains poor. The incidence of melanoma has been rising globally, with over 100,000 new cases and more than 7,000 deaths annually in the United States alone. Recent advancements in next-generation sequencing (NGS) have led to an exponential growth in tumor data, enhancing our understanding of cancer pathology. However, current analysis methods often focus on single genes, ignoring complex interactions. Our study constructed a bidirectional, weighted, signed, and directed topological immune gene regulatory network for cutaneous melanoma, comparing benign nevi and melanomas. We observed significant differences between the immune gene regulatory network patterns of melanocytic nevi and melanoma. Specifically, the dominant modules shifted from cell cycle-related modules to those associated with DNA repair and cell migration. Additionally, we identified a set of key genes, including AURKA, CCNE1, APEX2, and EXOC8, which may play crucial roles in the immune activity shifts from nevi to melanoma. Furthermore, we employed deep learning models to predict drug-target interactions, offering new insights for clinical treatment and melanoma research.