Identification and Development of Subtypes With Poor Prognosis in Pan-Gynecological Cancer Based on Gene Expression in the Glycolysis-Cholesterol Synthesis Axis

Objective: Metabolic reprogramming is an important biomarker of cancer. Metabolic adaptation driven by oncogenes allows tumor cells to survive and grow in a complex tumor microenvironment. The heterogeneity of tumor metabolism is related to survival time, somatic cell-driven gene mutations, and tumor subtypes. Using the heterogeneity of different metabolic pathways for the classification of gynecological pan-cancer is of great significance for clinical decision-making and prognosis prediction. Methods: RNA sequencing data for patients with ovarian, cervical, and endometrial cancer were downloaded from The Cancer Genome Atlas database. Genes related to glycolysis and cholesterol were extracted and clustered coherently by using ConsensusClusterPlus. The mutations and copy number variations in different subtypes were compared, and the immune scores of the samples were evaluated. The limma R package was used to identify differentially expressed genes between subtypes, and the WebGestaltR package (V0.4.2) was used to conduct Kyoto Encyclopedia of Genes and Genomes pathway and Gene Ontology functional enrichment analyses. A risk score model was constructed based on multivariate Cox analysis. Prognostic classification efficiency was analyzed by using timeROC, and internal and external cohorts were used to verify the robustness of the model. Results: Based on the expression of 11 glycolysis-related genes and seven cholesterol-related genes, 1,204 samples were divided into four metabolic subtypes (quiescent, glycolysis, cholesterol, and mixed). Immune infiltration scores showed significant differences among the four subtypes. Survival analysis showed that the prognosis of the cholesterol subtype was better than that of the quiescent subtype. A nine-gene signature was constructed based on differentially expressed genes between the cholesterol and quiescent subtypes, and it was validated by using an independent cohort of the International Cancer Genome Consortium. Compared with existing models, our nine-gene signature had good prediction performance. Conclusion: The metabolic classification of gynecological pan-cancer based on metabolic reprogramming may provide an important basis for clinicians to choose treatment options, predict treatment resistance, and predict patients' clinical outcomes.