Online Program Home
My Program

Abstract Details

Activity Number: 648 - Are Statistical Methods Developed for Bulk RNAseq Data Appropriate for Single Cell Data Sets?
Type: Topic Contributed
Date/Time: Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #306518 Presentation
Title: Unlocking Bulk RNA-Seq Tools for Single Cell Applications
Author(s): Lieven Clement*
Companies: Ghent University
Keywords: generalized additive models; single cell RNA-seq; zero inflation; differential expression; developmental trajectories

Single-cell RNA-sequencing (scRNA-seq) allows transcript profiling at the level of single cells, enabling cell identity delineation and the study of cell-level heterogeneity of gene expression. Through differential expression (DE) analysis, marker genes can be discovered that define different cell identities, or that are (differentially) associated with a particular developmental process, e.g. the development of a stem cell to a differentiated cell type. First, we show that adopting bulk RNA-seq methods for DE analysis in scRNA-seq data suffers from excess zeros, especially for full-length protocols. We therefore introduce observation weights derived from zero inflated negative binomial (ZINB) models that effectively down-weight excess zeros in the analysis, and unlock bulk RNA-seq tools for DE analysis in single cell applications. Second, we introduce a generalized additive model (GAM) framework for DE analysis downstream of trajectory inference as genes often exhibit nonlinear expression profiles along development. We show how the NB-GAMs provide a powerful approach to assess differences in gene expression along and between lineages of a trajectory.

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

Back to the full JSM 2019 program