Chronic kidney disease and colorectal cancer: analysis of public databases and introduction to the “Spatial radiomics and transcriptomics to the discovery of the cross-link between colon cancer and chronic kidney disease in the SIRIO study”

Abstract

Background:
Colorectal cancer (CRC) and chronic kidney disease (CKD) are major contributors to global morbidity and mortality. Emerging evidence suggests a potential biological interplay between these two conditions, yet integrative datasets linking renal dysfunction with colorectal tumorigenesis are lacking.

Methods:
This study evaluated public datasets including TCGA-COAD and an open-access GitHub transcriptomic-survival dataset to explore the feasibility of integrating clinical and genomic data for biomarker discovery. Unsupervised clustering, differential gene expression analysis, and multiple supervised machine learning classifiers were employed to identify prognostic gene signatures associated with disease-free survival (DFS).

Results:
TCGA-COAD confirmed the prominence of CRC driver mutations (e.g., KRAS, APC, TP53), but lacked renal data. In contrast, the 62-patient genomic dataset enabled integrative DFS analysis. Clustering revealed three transcriptionally distinct subgroups. Random Forest and Logistic Regression achieved the highest classification performance (AUC > 0.90). Top-ranked predictive genes included CYP2E1, RAB39A, and ZBTB3 highlighted for future molecular validation.

Conclusion:
This study underscores the need for comprehensive, multimodal datasets to disentangle CRC–CKD associations. Our findings provide a proof-of-concept framework for integrating transcriptomics and machine learning to inform prognostic stratification in CRC.