Prognostic model development for classification of colorectal adenocarcinoma by using machine learning model based on feature selection technique boruta

Colorectal cancer (CRC) is the third most prevalent cancer type and accounts for nearly one million deaths worldwide. The CRC mRNA gene expression datasets from TCGA and GEO (GSE144259, GSE50760, and GSE87096) were analyzed to find the significant differentially expressed genes (DEGs). These significant genes were further processed for feature selection through boruta and the confirmed features of importance (genes) were subsequently used for ML-based prognostic classification model development. These genes were analyzed for survival and correlation analysis between final genes and infiltrated immunocytes. A total of 770 CRC samples were included having 78 normal and 692 tumor tissue samples. 170 significant DEGs were identified after DESeq2 analysis along with the topconfects R package. The 33 confirmed features of importance-based RF prognostic classification model have given accuracy, precision, recall, and f1-score of 100% with 0% standard deviation. The overall survival analysis had finalized GLP2R and VSTM2A genes that were significantly downregulated in tumor samples and had a strong correlation with immunocyte infiltration. The involvement of these genes in CRC prognosis was further confirmed on the basis of their biological function and literature analysis. The current findings indicate that GLP2R and VSTM2A may play a significant role in CRC progression and immune response suppression.