Fig. 8

Predicting the 3-year survival probability of colorectal cancer patients using ATG7-HMGCR expression patterns combined with machine learning algorithms. (A) Representative images and quantification of IHC staining for ATG7, HMGCR, and CD8 expression were obtained from 12 clinical samples, which were randomly selected from 6 different groups of CRC patients with varying levels of ATG7 expression. (high: high ATG7 expression group, low: high ATG7 expression group). Scale bars, 50 μm. (B) The prognostic value of combining ATG7 and HMGCR levels was analyzed by Kaplan-Meier analysis in the Huashan COAD/READ cohort. (C) The OS survival analysis of patients who ever or never used statins was analyzed by Kaplan-Meier analysis in the Huashan COAD/READ cohort. (D) ROC curve analysis of six machine learning algorithms for prediction the 3-year survival probability of COAD patients in the TCGA-COAD/READ cohort. (E) Matthews correlation coefficient of six machine learning algorithms. (F) The established random forest model was used to rank the importance of clinical parameters. (G) Prediction of the 3-year survival events for Huashan COAD/READ cohort patients using the established random forest model