A Gleason score-related outcome model for human prostate cancer: a comprehensive study based on weighted gene co-expression network analysis

• Published in: Cancer cell international 2020-05, Vol.20 (1), p.159-15, Article 159
• Main Authors: Wang, Yongzhi, Yang, Zhonghua
• Format: Article
• Language: English
• Subjects: Accuracy; Bioinformatics; Biomarkers; Calibration; Cancer therapies; CD4 antigen; Cdc45 protein; Datasets; Dendritic cells; Gene expression; GEO; Immunological memory; LASSO; Leukocytes (eosinophilic); Lymphocytes; Lymphocytes B; Lymphocytes T; Mast cells; Medical prognosis; Medical research; Memory cells; Metastases; Metastasis; Nomograms; Patients; Plasma cells; Primary Research; Prognosis; Prostate cancer; Regression analysis; Survival analysis; TCGA; WGCNA
• ISSN: 1475-2867; 1475-2867
• DOI: 10.1186/s12935-020-01230-x

Prostate cancer (PCa) is the second leading cause of cancer death in men in 2018. Thus, the evaluation of prognosis is crucial for clinical treatment decision of human PCa patients. We aim to establishing an effective and reliable model to predict the outcome of PCa patients. We first identified differentially expressed genes between prostate cancer and normal prostate in TCGA-PRAD and then performed WGCNA to initially identify the candidate Gleason score related genes. Then, the candidate genes were applied to construct a LASSO Cox regression analysis model. Numerous independent validation cohorts, time-dependent receiver operating characteristic (ROC), univariate cox regression analysis, nomogram were used to test the effectiveness, accuracy and clinical utility of the prognostic model. Furthermore, functional analysis and immune cells infiltration were performed. Gleason score-related differentially expressed candidates were identified and used to build up the outcome model in TCGA-PRAD cohort and was validated in MSKCC cohort. We found the 3-gene outcome model (CDC45, ESPL1 and RAD54L) had good performance in predicting recurrence free survival, metastasis free survival and overall survival of PCa patients. Time-dependent ROC and nomogram indicated an ideal predictive accuracy and clinical utility of the outcome model. Moreover, outcome model was enriched in 28 pathways by GSVA and GSEA. In addition, the risk score was positively correlated with memory B cells, native CD4 T cells, activated CD4 memory T cells and eosinophil, and negatively correlated with plasma cells, resting CD4 memory T cells, resting mast cells and neutrophil. In summary, our outcome model proves to be an effective prognostic model for predicting the risk of prognosis in PCa.