Construction of a protein-based model for prognosis prediction of kidney renal clear cell carcinoma: an investigation based on functional proteomics data

Abstract

Background Several studies have shown prognostic value of gene-based models at mRNA level in kidney renal clear cell carcinoma (KIRC). However, protein-based models for prognosis prediction of KIRC are rarely reported, then we conduct this study. Methods Proteomics and clinical data of KIRC were acquired from The Cancer Proteome Atlas (TCPA) and The Cancer Genome Atlas (TCGA), respectively. Prognosis-associated proteins were screened by univariate Cox regression analysis and log-rank test. Patients were grouped into training set and testing set. The protein-based prognostic model was constructed by lasso Cox regression analysis in training set and validated in test set and whole set. Results A five-protein model (ACC1, IGFBP2, MIG6, PEA15 and RAD51) was constructed for KIRC prognosis prediction. It could well classify KIRC patients into high-risk and low-risk group with significantly different survival. The results was validated in the testing set and whole set. Age, AJCC stage and risk score based on the five-protein model were identified as independent prognostic parameters and they were used to construct a nomogram. The calibration plot showed the nomogram had good agreement between predicted and actual outcomes. Time-dependent ROC curves revealed the nomogram performed best in predicting the 1-year, 3-year and 5-year overall survival (OS) compared with other independent prognostic parameters. DCA demonstrated the nomogram had obviously clinical net benefit. Furthermore, several cancer-related biological signaling pathways were enriched in the functional enrichment analysis. Conclusion Our study developed an effective protein-based model to predict the OS of KIRC, which may help clinicians to offer individual treatment.