Clinicopathologic and protein markers distinguishing POLE and copy-number low group in endometrial cancer.

Abstract

e18110 Background: Requirement of DNA sequencing to diagnose POLE group is a barrier for adoption of molecular subtyping in endometrial cancer. We aimed to identify clinicopathologic and protein markers distinguishing POLE and copy-number (CN) low group in endometrial cancer. Methods: Ninety-one samples (POLE: 15, CN low 76) classified as POLE or CN low groups by integrative clustering were selected from The Cancer Genome Atlas endometrial cancer dataset. Clinicopathologic variables and normalized reverse phase protein array expression data were extracted. We first selected clinicopathologic variables associated with group (POLE vs CN low) via univariate analysis. Then, we identified protein makers using the logistic regression model with significant clinicopathologic variables as adjusting covariates. The differentially expressed proteins were selected based on q-value of the false discovery rate for multiple comparison. With various q-value cut-off ( < 5%, 10%, 20%), several logistic models including differentially expressed markers were constructed by stepwise selection method using the area under curve (AUC) from 5-fold cross-validation (CV). Results: Among clinicopathologic variables, body mass index (BMI) and tumor grade were associated with group (p = 0.02 and p < 0.01, respectively). Being adjusted for BMI and tumor grade, 5 proteins were associated with group in q-value cut-off of < 5%. The model including clinicopathologic variables and 5 proteins identified BMI, Cyclin B1, Caspase 8, and XBP1 as markers distinguishing POLE and CN low groups. The mean of CV AUC, sensitivity and specificity of the selected model were 0.97, 0.97, and 0.60, respectively. Conclusions: BMI and expression levels of Cyclin B1, Caspase 8, XBP1 are candidate markers distinguishing POLE and CN low group. A further validation study using immunohistochemical staining is necessary to facilitate the adoption of molecular subtyping as daily practice.