Background: Ulcerative colitis is one of the two main forms of inflammatory bowel disease. Cuproptosis is reported to be a novel mode of cell death.
Methods: We examined clusters of cuproptosis related genes and immune cell infiltration molecules in 86 ulcerative colitis samples from the GSE179285 dataset. We identified the differentially expressed genes according to the clustering method, and the performance of the SVM model, the random forest model, the generalized linear model, and the limit gradient enhancement model were compared, and then the optimal machine model was selected. To assess the accuracy of the learning predictions, the nomogram and the calibration curve and decision curve analyses showed that the subtypes of ulcerative colitis have been accurately predicted.
Results: Significant cuproptosis-related genes and immune response cells were detected between the ulcerative colitis and control groups. Two cuproptosis-associated molecular clusters were identified. Immune infiltration analysis indicated that different clusters exhibited significant heterogeneity. The immune scores for Cluster2 were elevated. Both the residual error and root mean square error of the random forest machine model had clinical significance. There was a clear correlation between the differentially expressed genes in cluster 2 and the response of immune cells. The nomogram and the calibration curve and decision curve analyses showed that the subtypes of ulcerative colitis had sufficient accuracy.
Conclusion: We examined the complex relationship between cuproptosis and ulcerative colitis in a systematic manner. To estimate the likelihood that each subtype of cuproptosis will occur in ulcerative colitis patients and their disease outcome, we developed a promising prediction model.
Keywords: Cuproptosis; Machine learning; Molecular clusters; Prediction model; Ulcerative colitis.
© 2023. The Author(s).