Abstract P5-12-04: A new method of data analysis to derive DNA methylation signatures that stratify risk of recurrence in triple negative breast cancer

Background: Triple negative breast cancer (TNBC) accounts for 10-17% of all breast cancer and is more likely to be of higher histological grade, poorly differentiated, associated with a higher recurrence rate and with decreased overall survival. The clinical course of a TNBC patient remains difficult to predict, as tumors with homogenous morphological characteristics may vary in response to therapy and have divergent outcomes. Therefore, additional analytical methods are needed to better classify TNBC. Our goal is to refine the analysis of methylome datasets to derive reliable molecular signatures that can distinguish TNBC patients with good outcomes who may benefit from less aggressive treatment, from those with poor outcomes who would be candidates for more aggressive treatments. Methods: Our laboratory has conducted and reported, in this meeting, results from analysis of 450k methylation array data on a discovery set of 53 high-risk TNBC cases and 62 low-risk controls treated by locoregional therapy alone, as well as 5 normal breast tissue samples. High-risk cases were defined as patients that relapsed within 0.5 to 6.5 years from the time of diagnosis, while low-risk controls had no relapse and >4 year recurrence-free intervals (RFI). In this work, we devised and applied a novel methylation biomarker discovery program named Hypermethylated Outlier Detector (HOD) that emphasizes the selection of highly methylated markers in cases compared to controls, to find a high-risk signature in the TNBC discovery set. The methylation signature identified by HOD was interrogated in a test set of 50 TNBCs (with 16 recurrences) that did not receive chemotherapy, and in a second test set of 131 TNBCs (with 33 recurrences) that did receive chemotherapy. Results: HOD identified 39 hypermethylated markers (beta >0.20) that could accurately distinguish between the high-risk cases and the low-risk controls in the discovery set of TNBCs (n=115) treated with locoregional therapy alone. In the test set of TNBC (n=50) with no chemotherapy the 39 markers distinguished high from low risk individuals (likelihood ratio test P=0.049). In a second test set of TNBC (n=131) that received chemotherapy the 39 hypermethylated markers again distinguished high from low risk individuals (likelihood ratio test P=0.0043). Conclusions: We have presented evidence that a methylation signature identified by HOD can be used to identify TNBC patients that have a high-risk of relapse regardless of receiving chemotherapy. This methylation signature could potentially be used to inform physician decisions on therapeutic strategies for TNBC patients. This could ultimately lead to less aggressive treatment given to patients possessing a methylation profile consistent with a better prognosis. Conversely, patients with hypermethylation in the 39 markers will likely benefit from a more aggressive course of treatment. Citation Format: Downs BM, Cope LM, Fackler MJ, Cho S, Wolff AC, Regan MM, Sukumar S, Umbricht CB. A new method of data analysis to derive DNA methylation signatures that stratify risk of recurrence in triple negative breast cancer [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P5-12-04.