Abstract 2497: Topological data analysis revealed complexity behind ER positive breast cancer subtypes in TCGA transcriptome data

Abstract

Abstract Introduction: Breast cancer subtyping is still not fully agreed among oncologists. Today, many guidelines separate patients into four subtypes, luminal A, luminal B, Her2 positive, and triple negative breast cancers. However, distinction between luminal A and luminal B cases is not clear cut although many use Ki67 index. As more treatment options including immune checkpoint blockade are available, it is imperative to define strata and to find optimum treatments for each stratum. Method: TCGA BRCA transcriptome data (BRCA-UQ) is used, which has a sample size of 1089. We employed topological data analysis (TDA) which can handle high dimensional nature of transcriptome data in elegant manner. For characterization of the clusters, we used clinical data such as metastasis status, tumor sizes, age of diagnosis, survival, and pathological data and molecular data including PIK3CA, TP53, GATA3 mutations, MYC amplification, 8p23.2 deletion, 16q24.3 deletion, and tumor mutation burden (TMB). In addition, state of tumor microenvironment is assessed using CD45 expression for tumor infiltrating leukocytes (TIL) and HOXD12 expression for cancer associated fibroblast (CAF). Results: TDA cleanly clustered breast cancer transcriptome data separating ER+, Her2+ and triple negative cases. However, among ER+ cases, boundary between Luminal A and Luminal B is vague except for ER+/Her2+ cases. In ER+ cases, there are four clusters with poor prognosis. One has PIK3CA mutation, high MYC, and low TIL. Another cluster is characterized by GATA3 mutations. The other has low TIL and PIK3CA mutations. Also, high CAF cluster coincides with poor prognosis ones. Conclusions: The study revealed complexity behind ER+ breast cancer stratification from molecular and clinical perspectives. A TDA based method will likely to classify a patient into a right stratum better than conventional methods. Identified clusters with high TIL and high TMB which are good candidates for immune checkpoint blockade. Further study on identification of optimal treatments for each stratum will be warranted. Citation Format: Takahiko Koyama, Aldo G. Saenz, Laxmi Parida. Topological data analysis revealed complexity behind ER positive breast cancer subtypes in TCGA transcriptome data [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2497.