Landscape of Metabolite-Protein Interaction Networks Reveals Prognostic Subtypes of Ovarian Cancer

Abstract

Background Metabolic reprogramming, a hallmark of cancer, can promote tumorigenesis and tumour progression through metabolite-protein interactions (MPIs). However, MPI functions and related genes in ovarian cancer (OV) development and treatment remain largely unknown. Methods A TCGA-based metabolic heterogeneity analysis of pancancer was used to identify OV-specific metabolic altered genes (MIPros) and classify OV by MPIScore. MPIscores were based on hub genes intersecting the WGCNA module genes and DEGs of the PCA subtype and LASSO Cox regression analysis. A correlation analysis of the MPIscore, clinical features, functional and genomic characteristics, and the immune landscape was performed. The Gene Expression Omnibus (GEO) database was used for validation. Result In total, 323 OV-specific MIPros were identified by pancancer analysis and used for PCA. Two subtypes with different survival times, ages, and HRD scores were recognized. Five hub prognosis-related genes were included in the MPIscore, an independent prognostic factor (HR = 4.029, P = 0.0118) of patient survival, and possessed distinct metabolism-related pathways and clinical features. Genomic mutations were distributed diversely among MPIscore subgroups; comutations among frequently mutated were detected. Tumour microenvironment analyses correlated a high MPIscore with greater immune infiltration and TIDE scores, leading to poor responses to immunotherapy. Subtyping was consistent across multiple OV cohorts. Conclusion A new OV typing method was developed using specific MIPros, showing differences in metabolism, mutation, immune landscape, and drug response, improving understanding and clinical applications of OV metabolism heterogeneity.