Purpose: To develop and validate a gene-related nomogram for predicting the risk of lymph node (LN) metastasis preoperatively in patients with colorectal cancer (CRC). Methods: RNA-seq data of 581 CRC and 51 normal cases with clinical features were downloaded from TCGA database. In the evaluation cohort with 381 CRC cases, the LASSO regression was used to reduce dimensionality of gene signatures extracted to build gene score. A gene-related nomogram was performed based on the multivariable logistic regression analysis. The performance of the nomogram was assessed by the discrimination, calibration, and clinical usefulness not only in the evaluation, but also in the validation cohort with 200 CRC cases. Results: A total of 12,590 differentially expressed genes were selected, in which 59 candidates associated with LN metastasis in differentially expressed genes set were screened by LASSO to form the gene score. Based on the analysis of multivariate logistic regression, the gene-related nomogram showed good calibration and discrimination not only in the evaluation cohort (concordance-index 0.93; 95%CI 0.91-0.96), but also in the validation cohort (concordance-index 0.70; 95%CI 0.63-0.78). The decision curve analysis of the gene-related nomogram also provides constructive guidance for the design of operation plan, preoperatively. Conclusions: The presented genes nomogram may predict the LN metastasis in CRC patients, preoperatively. And 59 hub genes were defined related to LN metastasis of CRC, which can serve as treatment targets for the further study. Preoperative biopsy and gene analysis are needed to develop the operation plan in clinical practice.
Keywords: CRC; LN metastasis; RNA-seq; logistic regression model; nomogram; the LASSO regression.