Integrated single-cell and bulk transcriptomic analysis leverages liver metastasis-related genes to develop a prognostic model for colorectal cancer patients
Xu, Y.; Zhang, X.; Chen, W.; Li, Y.; Lu, L.; Huang, R.; Liao, J.; Li, H.; Zheng, W.
Show abstract
PurposeDifferentially expressed genes (DEGs) between colorectal cancer liver metastasis (CRLM) epithelium and primary colorectal cancer (CRC) epithelium (LMR DEGs) identified based on single-cell RNA sequencing (scRNA-seq) data may become new biomarkers for CRC prognosis. MethodsAn scRNA-seq dataset was used to describe the cellular landscape of primary CRC and CRLM and identify LMR DEGs. Prognostic LMR DEGs were identified in the bulk RNA-seq dataset. Based on the prognostic LMR DEGs, multiple machine learning algorithm combinations were compared in terms of their C-index, and the best model was selected for the construction of the LMR score. ResultsAmong the 2070 LMR DEGs, 426 prognostic LMR DEGs were ultimately obtained. The combination of the randomized survival forest (RSF) model and ridge regression had the highest C-index and was therefore used to construct a 15-gene scoring system (LMR score). In the external validation set, the 1- and 5-year AUCs of the LMR score were greater than those of the AJCC stage and other scoring systems constructed with a similar dataset. In addition, the LMR score was closely associated with factors that influence CRC outcomes, such as immune infiltration. ConclusionThe LMR score may be a reliable new biomarker for predicting the prognosis of patients with CRC.
Matching journals
The top 7 journals account for 50% of the predicted probability mass.