A Prognostic Immunoscore for Relapse-Free Survival Prediction in Colorectal Cancer

We aimed to establish a novel immunoscore (IS) model based on the transcriptomes of tumor tissues to improve the relapse-free survival (RFS) prediction of colorectal cancer (CRC). CIBERSORT was used to estimate the immune cell fractions based on the Gene Expression Omnibus (GEO) database. Then, a least absolute shrinkage and selection operator regression was applied to construct the IS model based on the immune cell fractions. After screening, four GEO databases were included in the CIBERSORT transformation. A total of 13 types of immune cells were selected and constructed an IS model. In the training set (n = 613) and test set (n = 262), the patients in the high-immunoscore group showed a significant poor RFS than that in the low-immunoscore group. Stratified analysis also found similar results in patients with identical age, sex, adjunctive chemotherapy, or TNM stage I-II. Multivariate Cox regression further demonstrated that the IS model was an independent predictor of RFS in CRC. In addition, the IS was highly associated with the expression of several immune checkpoints, inflammatory mediators, cell cycle, and epithelial-mesenchymal transformation regulators in CRC. We proposed a novel IS model for estimating RFS in CRC patients.