Predicting master transcription factors from pan-cancer expression data
Jessica Reddy, Marcos A. Fonseca, Rosario I. Corona, Robbin Nameki, Felipe Segato Dezem, Isaac A. Klein, Heidi Chang, Daniele Chaves‐Moreira, Lena K. Afeyan, Tathiane M. Malta, Xianzhi Lin, Forough Abbasi, Alba Font‐Tello, Thaís S. Sabedot, Paloma Cejas, Norma I. Rodríguez-Malavé, Ji-Heui Seo, De‐Chen Lin, Ursula A. Matulonis, Beth Y. Karlan, Simon A. Gayther, Bogdan Paşaniuc, Alexander Gusev, Houtan Noushmehr, Henry W. Long, Matthew L. Freedman, Ronny Drapkin, Richard A. Young, Brian J. Abraham, Kate Lawrenson
Abstract
Critical developmental “master transcription factors” (MTFs) can be subverted during tumorigenesis to control oncogenic transcriptional programs. Current approaches to identifying MTFs rely on ChIP-seq data, which is unavailable for many cancers. We developed the CaCTS (Cancer Core Transcription factor Specificity) algorithm to prioritize candidate MTFs using pan-cancer RNA sequencing data. CaCTS identified candidate MTFs across 34 tumor types and 140 subtypes including predictions for cancer types/subtypes for which MTFs are unknown, including e.g. PAX8, SOX17, and MECOM as candidates in ovarian cancer (OvCa). In OvCa cells, consistent with known MTF properties, these factors are required for viability, lie proximal to superenhancers, co-occupy regulatory elements globally, co-bind loci encoding OvCa biomarkers, and are sensitive to pharmacologic inhibition of transcription. Our predictions of MTFs, especially for tumor types with limited understanding of transcriptional drivers, pave the way to therapeutic targeting of MTFs in a broad spectrum of cancers.