Visualization methods for differential expression analysis

Authors: RUTTER, LINDSAY - Iowa State University; Moran Lauter, Adrienne; Graham, Michelle; COOK, DIANNE - Monash University

Submitted to: BMC Bioinformatics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/24/2019
Publication Date: 9/6/2019
Citation: Rutter, L., Moran Lauter, A.N., Graham, M.A., Cook, D. 2019. Visualization methods for differential expression analysis. BMC Bioinformatics. 20:458. https://doi.org/10.1186/s12859-019-2968-1.
DOI: https://doi.org/10.1186/s12859-019-2968-1

Interpretive Summary: RNA-sequencing (RNA-seq) compares gene expression in response to a treatment across all genes in an organism's genome. While many programs are available for identifying differentially expressed genes, few provide researchers with opportunities to visualize their data and identify problems that could impact their results and interpretation. We have developed data visualization tools to allow researchers to easily explore their data and identify potential problems. We use publicly available data to demonstrate how scientists can detect problems in sample tracking, data normalization, identification of differentially expressed genes and other issues.

Technical Abstract: Motivation: Despite the availability of many ready-made testing software, reliable detection of differentially expressed genes in RNA-seq data is not a trivial task. While the data collection might be considered high- throughput, data analysis has intricacies that require careful human attention. In light of this, researchers should use effective and modern data analysis techniques, and verify and enhance the appropriateness of their models with visual feedback. There is a need to make it easier for researchers to use models and visuals in a complimentary fashion during RNA-seq data analysis. Results: We use several public RNA-seq data sets to show that our visualization tools can detect normalization issues, DEG designation problems, and common analysis errors. We also show that our visual tools can identify genes of interest in ways undetectable with models. In this paper, we propose that users slightly modify their approach to RNA-seq analysis by incorporating statistical graphics into their usual analysis pipelines. Availability: Interactive versions of graphics are available at shinyapps.io, as specified in the paper.