Comprehensive Analysis of Microsatellite-Related Transcriptomic Signature and Identify Its Clinical Value in Colon Cancer

BACKGROUND: Microsatellite has been proved to be an important prognostic factor and a treatment reference in colon cancer. The transcriptome profile and tumor microenvironment of different microsatellite statuses are different. Metastatic colon cancer patients with microsatellite instability-high (MSI-H) are sensitive to immune checkpoint inhibitors (ICIs), but not fluorouracil. Efforts have been devoted to identify the predictive factors of immunotherapy. METHODS: We analyzed the transcriptome profile of different microsatellite statuses in colon cancer by using single-cell and bulk transcriptome data from publicly available databases. The immune cells in the tumor microenvironment were analyzed by the ESTIMATION algorithm. The microsatellite-related gene signature (MSRS) was constructed by the least absolute shrinkage and selection operator (LASSO) Cox regression based on the differentially expressed genes (DEGs) and its prognostic value and predictive value of response to immunotherapy were assessed. The prognostic value of the MSRS was also validated in another cohort. RESULTS: The MSI-H cancers cells were clustered differentially in the dimension reduction plot. Most of the immune cells have a higher proportion in the tumor immune microenvironment, except for CD56 bright natural killer cells. A total of 238 DEGs were identified. Based on the 238 DEGs, a neural network was constructed with a Kappa coefficient of 0.706 in the testing cohort. The MSRS is a favorable prognostic factor of overall survival, which was also validated in another cohort (GSE39582). Besides, MSRS is correlated with tumor mutation burden in MSI-H colon cancer. However, the MSRS is a barely satisfactory factor in predicting immunotherapy with the area under the curve (AUC) of 0.624. CONCLUSION: We developed the MSRS, which is a robust prognostic factor of overall survival in spite of a barely satisfactory immunotherapy predictor. Further studies may need to improve the predictive ability.