Colorectal cancer (CRC) is the fourth most prevalent cancer type worldwide. Although the 5-year survival rate of CRC is higher than many cancer types, prediction of prognosis and identification of accurate biomarkers still maintain their importance for chemotherapy benefits, thus survival of the patients. Current techniques to identify biomarkers for clinical use are based on building models with multi-gene signatures. However, the accuracy rates of such signatures are not high enough due to heterogeneity of the tumors and low sensitivity of gene expression measurement techniques, although cell lines can be predicted very well with such signatures. There has also been sufficient evidence that multi-gene signatures may not be better predictors than random signatures with the same size. Therefore, in this study, we aimed to develop two R-based statistical analysis tools, SSAT and USAT, to identify single-gene expression markers for prognosis with chemotherapy benefit prediction power. We identified two genes, ULBP2 and SEMA5A, with SSAT and 6 genes, PTRF, TGFB1I1, DUSP10, KLF9, CLCN7 and CLDN3, with USAT for colon cancer and CRC, respectively. We were able to validate independent prognostic power of ULBP2 and SEMA5A in an independent cohort. However, we could only validate CLCN7 among 6 genes that we identified by USAT. Those results showed that SSAT may be a better tool to identify prognostic gene markers and USAT needs to be improved to identify better candidate genes. We could also reveal the chemotherapy benefit prediction power of ULBP2 and SEMA5A in CCLE and CGP drug databases, although these in silico results should be validated by in vitro experiments. We believe that the approach that we used in this study may pioneer the studies to develop commercial theranostic tools for clinical use in various types of cancer.