Construction of immune related genes predictive models of elderly rectal cancer patients based on machine learning

There were not many studies on predictive indicators of elderly rectal cancer (ERC) based on transcriptome level. Immune related genes (IRGs) have been shown to be associated with cancer, however, their role in ERC had not been fully explored. For the first time, IRGs associated with ERC prognosis were identified using univariate Cox regression analysis. Based on the screened IRGs, consistent clustering analysis and prognostic difference analysis were conducted and a prognostic model was constructed through multivariate Cox regression. The model was validated on an independent dataset. The patients were grouped based on the median risk score. Simultaneously conduct immune infiltration analysis, enrichment analysis and drug sensitivity evaluation. The prognostic value of IRGs in ERC patients had been comprehensively analyzed for the first time and identified 3 IRGs with prognostic values. Consensus cluster analysis based on 3 IRGs revealed that the two identified clusters had prognostic differences. A prognostic model was constructed based on 3 IRGs and validated in an independent dataset. The signature was an independent risk factor for ERC patients, 3 IRGs were related to the immune microenvironment of ERC. This research provides a fundamental theoretical basis for guiding the treatment of ERC patients.