Genomic Landscape Defines Peritoneal Metastatic Pattern and Related Target of Peritoneal Metastasis in Colorectal Cancer.

The primary objective of this study is to develop a prediction model for peritoneal metastasis (PM) in colorectal cancer by integrating the genomic features of primary colorectal cancer, along with clinicopathological features. Concurrently, we aim to identify potential target implicated in the peritoneal dissemination of colorectal cancer through bioinformatics exploration and experimental validation. By analyzing the genomic landscape of primary colorectal cancer and clinicopathological features from 363 metastatic colorectal cancer patients, we identified 22 differently distributed variables, which were used for subsequent LASSO regression to construct a PM prediction model. The integrated model established by LASSO regression, which incorporated two clinicopathological variables and seven genomic variables, precisely discriminated PM cases (AUC 0.899; 95% CI 0.860-0.937) with good calibration (Hosmer-Lemeshow test p = .147). Model validation yielded AUCs of 0.898 (95% CI 0.896-0.899) and 0.704 (95% CI 0.622-0.787) internally and externally, respectively. Additionally, the peritoneal metastasis-related genomic signature (PGS), which was composed of the seven genes in the integrated model, has prognostic stratification capability for colorectal cancer. The divergent genomic landscape drives the driver genes of PM. Bioinformatic analysis concerning these driver genes indicated SERINC1 may be associated with PM. Subsequent experiments indicate that knocking down of SERINC1 functionally suppresses peritoneal dissemination, emphasizing its importance in CRCPM. In summary, the genomic landscape of primary cancer in colorectal cancer defines peritoneal metastatic pattern and reveals the potential target of SERINC1 for PM in colorectal cancer.