Online Program Home
My Program

Abstract Details

Activity Number: 29 - SPEED: An Ensemble of Advances in Genomics and Genetics
Type: Contributed
Date/Time: Sunday, July 29, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistics in Genomics and Genetics
Abstract #329083 Presentation
Title: A Novel Framework for Differential Gene Expression Analysis Using Robust Profile Likelihood Ratios
Author(s): Lehang Zhong* and Lisa Joanna Strug
Companies: Division of Biostatistics, Dalla Lana School of Public Health, University of Toronto and Genetics and Genome Biology, The Hospital for Sick Children
Keywords: RNA sequence; Likelihood Paradigm; Mean-Variance relationship; Dispersion

Differential gene expression analysis with RNA sequencing (RNA-seq) routinely assumes a negative binomial distribution for the read count to model the over dispersed mean-variance relationship for biological replicates. The negative binomial distribution assumes a quadratic variance function, yet more general patterns for over/under-dispersion could be observed. We propose a general solution for any variance pattern using robust adjusted pure likelihood inference. Poisson regression is used to model the relative abundance, with the estimated effective library size used as an offset term for normalization. Simulation shows that the profile likelihood ratio for the differential gene expression parameter, after a robust adjustment for departures from the Poisson variance assumption, can reliably measure the strength of evidence; the analogous Type I error probability is bounded by 0.02 in large samples and under a variety of scenarios. In RNA-seq from 63 patients with Cystic Fibrosis we determine differentially expressed genes associated with lung disease, and compare results to existing methods. Support intervals and novel graphics provide accessible genome-wide interpretations.

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

Back to the full JSM 2018 program