Poster abstracts
Poster number 7 submitted by Julie Bove
Decoding RNA cis-regulatory elements in transcription
Julie M. Bove (Ray and Stephanie Lane Computational Biology Department, Carnegie Mellon University, Pittsburgh, PA 15213), Lydia Phillips (Department of Biological Sciences, Carnegie Mellon University, Pittsburgh PA 15213), Dervla Moore-Frederick (Department of Biological Sciences, Carnegie Mellon University, Pittsburgh PA 15213), Jonathan E. Henninger (Department of Biological Sciences, Carnegie Mellon University, Pittsburgh PA 15213)
Abstract:
Genomic regulatory elements, including enhancers and promoters, canonically regulate transcription by recruiting transcription factors (TFs) through protein-DNA interactions. Emerging evidence suggests that RNA molecules can similarly regulate transcription. RNA synthesis at enhancers remains the best predictor of enhancer function, RNA identity contributes to transcriptional output, and key proteins that regulate transcription also bind to RNA. How specific features of RNA molecules, including sequence and structure, contribute to this regulation remains unknown. Here, we find that a majority of super enhancers, clusters of enhancers that regulate key cell identity genes, overlap introns in all sampled cell types. This enrichment is significantly more than expected by chance, suggesting that RNA-encoded cis regulatory elements may contribute to enhancer function. We also describe an approach that uses RNA-specific embeddings to train a deep learning model to predict nascent transcription from RNA sequence. Using feature attribution methods, we can nominate RNA sequence and structural features capable of regulating transcription, with the ultimate goal of testing them in cells. Uncovering RNA-mediated mechanisms will define a novel layer of transcriptional control and provide a framework to interpret a subset of the 88% of disease-associated SNPs that are intronic or intergenic.
Keywords: Deep Learning, Enhancers, Non-coding RNA
