Detecting Developmental Expression Switches from Transcriptomic and Epigenomic Data

• Activity Number: 397 - Statistical Learning for Epigenomics Data
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: SSC
• Author(s): Claudia Kleinman* and Marie Forest and Selin Jessa and Celia M.T. Greenwood
• Companies: McGill University and Lady Davis Research Institute, McGill University and McGill University and Lady Davis Research Institute, McGill University
• Keywords: epigenomics; genomics; ChIP-Seq; RNA-Seq; functional data analysis

Abstract

Precise, sustained control of gene expression is essential for brain development and function. Chromatin state plays a major role in achieving this regulation, particularly during development. Here, we leverage comprehensive transcriptomic and epigenomic datasets spanning the full course of brain development to identify developmental switches. At the expression level, we present an approach to detect variable exon usage based on functional data analysis with spline functions, estimating exon expression trajectories of 16 brain regions followed by hierarchical clustering of these trajectories. We used a distance matrix incorporating spatial information so as to detect brain-region specific patterns, and built a predictive model based on simulated training data and a random forest classifier to identify genes containing at least two sets of exons. At the chromatin level, in turn, we propose an approach to integrate epigenomic information based on hierarchical clustering and external cluster validity metrics to predict chromatin state switches. The combination of these tools may assist in the discovery of regulatory events that are critical to developmental transitions in the brain.