Alternative classifier to TCGA molecular subgroups to identify recurrence risk for endometrial cancer

The Cancer Genome Atlas (TCGA) endometrial cancer molecular subgroups, POLE mutated (POLEmut), mismatch repair deficient (MMRd), p53 mutated (p53mut, or copy number high), and p53 wild type (p53wt or copy number low) have been associated with clinical outcomes (survival and recurrence). However, there are caveats to the use of these subgroups as classifiers: model 1) was built with clustering methods; 2) has overlapping molecular features between subgroups; and 3) subgroups identify extreme phenotypes, but intermediate risk patients remain largely unclassified (p53wt, up to 50–60%) or with unclear risk assessment. In previous studies from TCGA data, we have identified potential molecular features that more accurately stratify endometrial cancer recurrence risk. Our objective was to create a prediction model for recurrence with these molecular features in an independent dataset.