Abstract
Recent developments in sequencing have allowed consortiums like ENCODE and ROADMAP to probe the epigenome of normal and cancer cell lines. Applying these approaches in a broad fashion to patient data has been difficult but a pioneering TCGA study using ATAC-Seq, an assay for probing open chromatin and thus regulatory regions, provides an opportunity to analyze cancer patient epigenomes for key regulators and potential drug targets.
Utilizing this TCGA dataset, we identify the regulatory targets of open chromatin regions as assayed by ATAC-Seq and RNA-Seq across 371 patients on a patient-specific basis. We subsequently utilize transcription factor (TF) footprinting to identify regions bound by 850 TF &TF-families in each patient. The final output of this approach identifies which TF regulates individual genes for every patient. Using a combination of network metrics and TF abundance, we can identify known and rare subtypes of multiple cancers with similar underlying networks. We next identify a set of TFs with greater importance and predicted activity within each cluster of samples relative to others. To expand the set of potentially druggable genes, we use data from the Achilles/DepMap project to identify a set of genes whose essentiality is most highly correlated with the set of critical TFs for each cluster, i.e. “coessential genes”. We next combine our list with two datasets, the list of known druggable genes and the Drug Interaction Database to identify compounds that target this gene set. We prioritize genes highly correlated with our key TFs' essentiality, are not pan-cancer essential genes or targeted by drugs with high inferred toxicity. We identify a set of neuroendocrine-like cancers in lung and stomach adenocarcinoma, rare disease subtypes in their representative cancers. Further investigation finds these samples expressing high levels of known neuroendocrine markers synaptophysin and PCSK1. Within well-studied cancers like breast and colon adenocarcinoma we identify separate ER+ and ER- sample clusters as well as MSI-H samples as separate and featuring different key TFs than MSI-L and MSS clusters. Our drug candidate pipeline identified known targeted drugs in breast cancer such as Tamoxifen and Fulvestrant, non-specific chemotherapies like doxorubicin in addition to the novel candidates Nelfinavir and Indinavir as potential therapeutic agents. In summary we identify key TFs in multiple cancers and novel drugs for targeting those TFs or downstream effectors.
Citation Format: Andre Neil Forbes, Duo Xu, Ekta Khurana. Identifying potential drug targets using patient-derived, tissue specific, gene regulatory networks [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 235.