Identification of Metabolism-Associated Prostate Cancer Subtypes and Construction of a Prognostic Risk Model

Abstract

BackgroundDespite being the second most common tumor in men worldwide, the tumor metabolism-associated mechanisms of prostate cancer (PCa) remain unclear. Herein, this study aimed to investigate the metabolism-associated characteristics of PCa and to develop a metabolism-associated prognostic risk model for patients with PCa.MethodsThe activity levels of PCa metabolic pathways were determined using mRNA expression profiling of The Cancer Genome Atlas Prostate Adenocarcinoma cohort via single-sample gene set enrichment analysis (ssGSEA). The analyzed samples were divided into three subtypes based on the partitioning around medication algorithm. Tumor characteristics of the subsets were then investigated using t-distributed stochastic neighbor embedding (t-SNE) analysis, differential analysis, Kaplan–Meier survival analysis, and GSEA. Finally, we developed and validated a metabolism-associated prognostic risk model using weighted gene co-expression network analysis, univariate Cox analysis, least absolute shrinkage and selection operator, and multivariate Cox analysis. Other cohorts (GSE54460, GSE70768, genotype-tissue expression, and International Cancer Genome Consortium) were utilized for external validation. Drug sensibility analysis was performed on Genomics of Drug Sensitivity in Cancer and GSE78220 datasets. In total, 1,039 samples and six cell lines were concluded in our work.ResultsThree metabolism-associated clusters with significantly different characteristics in disease-free survival (DFS), clinical stage, stemness index, tumor microenvironment including stromal and immune cells, DNA mutation (TP53 and SPOP), copy number variation, and microsatellite instability were identified in PCa. Eighty-four of the metabolism-associated module genes were narrowed to a six-gene signature associated with DFS, CACNG4, SLC2A4, EPHX2, CA14, NUDT7, and ADH5 (p <0.05). A risk model was developed, and external validation revealed the strong robustness our risk model possessed in diagnosis and prognosis as well as the association with the cancer feature of drug sensitivity.ConclusionsThe identified metabolism-associated subtypes reflected the pathogenesis, essential features, and heterogeneity of PCa tumors. Our metabolism-associated risk model may provide clinicians with predictive values for diagnosis, prognosis, and treatment guidance in patients with PCa.

A Robust Prognostic Gene Signature Based on eRNAs-Driven Genes in Prostate Cancer

“…Differential analysis of the MSI of each PRAD sample obtained from a previous study ( Zhang et al, 2020 ) showed that the low-risk group had a lower level of MSI than the high-risk group ( P < 0.05, Figure 5C ). We considered this to be MSI high in the high-risk group and MSI low in the low-risk group, due to the tumor’s disrupted function during DNA damage repair which increased gene instability.…”

“…Using the Tumor Immune Estimation Resource 2.0 (TIMER2.0), we downloaded the immune cell infiltration data of prostate adenocarcinoma (PRAD) tissues from the Cistrome Project 4 . We obtained the gene sets associated with epithelial-mesenchymal transformation (EMT), transforming growth factor β (TGF-β), and extracellular matrix (ECM) from the Molecular Signatures Database (MSigDB 5 ) ( Zhang et al, 2020 ). The copy number alterations (CNAs) analysis was performed by cBio.…”

Metabolic Classification and Intervention Opportunities for Tumor Energy Dysfunction

“…Cancer with fatty acid pathway changes will show different clinical-biological characteristics to others with altered protein or carbohydrate metabolism, despite all being recognized as oncological diseases regardless of organ or mutational or copy number aberration burden, depending on the degree of metabolic and energy dysfunction. Recent studies on hepatocarcinoma [ 12 ], prostate [ 13 ], and colon cancer [ 14 ] suggest classifying tumors according to their energy dysfunction degree. The group showing high metabolic activity tends to be associated with better prognosis, while those that show low metabolic activity have worse prognosis but high immune response, thus being more receptive to chemotherapy and immunotherapy [ 152 ].…”

“…Proposals for tumor classification and tumor heterogeneity according to the degree of metabolic alteration are particularly interesting from a metabolic reprogramming point of view [ 11 ]. Indeed, recent data on hepatocarcinoma [ 12 ], prostate [ 13 ], and colon cancer [ 14 ] support classifying tumors according to their energy dysfunction degree, independently of the tumor tissue origin. An important biological consideration in cancer, tumor heterogeneity, understood as differing morphological, immunophenotypic, and genotypic profiles that can occur in different areas of the same tumor, between the primary tumor and its metastases (spatial and/or temporal intratumoral heterogeneity), and/or between different tumors (intertumoral heterogeneity), can also be explained by varying degrees of energy dysfunction [ 15 ].…”

Artificial Intelligence in Bulk and Single-Cell RNA-Sequencing Data to Foster Precision Oncology

“…Similarly, the integration of pathway enrichment scores as input of random forest improved breast cancer classification is relative to single-gene signature-based methods [ 63 ]. Recently, partition around medoids clustering of metabolism-related gene set activity scores has been used to identify prostate cancer subtypes associated with patient prognosis and therapy response [ 64 ].…”