Online Program Home
My Program

Abstract Details

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
Abstract #329918 Presentation
Title: A Smoothed EM-Algorithm for Modeling DNA Methylation Profiles from Bisulfite Sequencing Data
Author(s): Karim Oualkacha* and Celia M.T. Greenwood and Kaiqiong Zhao and Lajmi Lakhal-Chaieb
Companies: Universite Du Quebec a Montreal and Lady Davis Research Institute, McGill University and Epidemiology, Biostatistics and Occupational Health, and Human Genetics, McGill University and Université Laval
Keywords: Epigenetic; DNA methylation; Whole-genome bisulfite sequencing; Local smoothing techniques; EM-algorithm; single cell type and cell lines
Abstract:

DNA methylation is an important epigenetic modification mechanism that is known to play a crucial role in tissue differentiation and cancer development. Whole-genome bisulfite sequencing (WGBS) is becoming a popular method for measuring DNA methylation profiles. However, modeling and analyzing WGBS data are very challenging. WGBS data are subject to experimental errors arising in both bisulfite-treatment and sequencing processes. DNA methylation often displays a pattern related to an important exposure or covariate. However, inaccurate methods with the aim to detect such associations are likely to be underpowered. In this work, we propose a smoothed EM-algorithm capable to correct for experimental errors and to take into account spatial correlations at neighboring CpG sites in order to provide accurate estimates of DNA methylation profiles and to make inference about smooth exposure/covariate effects. We evaluate the proposed methodology performance via simulation studies. We illustrate the use of the inference procedure through the analysis of a dataset from a cell line of human embryonic stem cells and also a data set for a genomic region in three different immune cell types.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program