Integrating prognosis-related genes with immune-related gene signature for development and validation of a survival stratification model for early-stage colorectal cancer.

Background: Current prognostic factors are inadequate for identifying patients at risk of disease recurrence after treatment for early-stage colorectal cancer (CRC). Integrating genes into gene pairs, considering the relative ranking of gene expression levels, can effectively avoid the bias due to different sequencing platforms when using the absolute values of gene expression levels. The aim of this study is to build an individualized gene index based on gene-pairs to estimate prognosis in patients with early-stage CRC and to assist in the decision-making of adjuvant chemotherapy. Methods: Clinical samples and gene expression data are acquired from public datasets including Gene Expression Omnibus (GEO) GSE39582, GSE17536, GSE33113, GSE39084, and The Cancer Genome Atlas (TCGA). Gene expression profiles from the Cluster Identification Tool (CIT) microarray dataset of 309 patients, who did not receive adjuvant chemotherapy, were integrated with immune-related gene signature to generate a 10-gene pair index (GPI). Prognostic analysis was performed to assess the predictive value of GPI on 1,525 CRC patients who were assigned to a training cohort and two validation cohorts. Results: There were 309 prognostic-related genes integrated with 463 immune-related genes to construct a GPI consisting of 20 unique genes. The GPI significantly stratified patients into high- and low-risk groups in terms of disease-free survival (DFS) among early-stage CRC [hazard ratio (HR) =3.91, 95% confidence interval (CI): 2.36-6.5]. A strong prognostic value was observed according to time-dependent ROC curve analysis [area under the curve (AUC) =0.753 at 2 years; AUC =0.770 at 3 years; AUC =0.775 at 5 years]. Besides, patients with adjuvant chemotherapy showed worse DFS than those without (HR =4.74, 95% CI: 2.28-9.82) among the GPI-identified low-risk group in early-stage CRC. The GPI-identified high-risk group obtained significantly higher immune and stromal infiltration. Several biological processes including apoptosis and angiogenesis were enriched among genes in the GPI. Compared with the commercialized index (Oncotype DX), GPI achieved higher accuracy (mean C-index, 0.81 Conclusions: An individualized gene index based on gene-pairs, which was robust against the technical biases across different sequencing platforms, was built to estimate prognosis in patients with early-stage CRC. The GPI model provides a new basis for the clinical strategic decision of whether early CRC patients should receive adjuvant chemotherapy.