Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium Hybrid
11 May 2021
Untangling batch and biological effects with RLE plots
Luke GandolfoWEHI Bioinformatics
Data visualisation is an essential part of analysing high-dimensional gene expression data, e.g. RNA-seq data. One reason for visualising data is to detect the presence of technical variation, e.g. batch effects, and to check the success of our attempts to remove such variation via normalisation. However, in situations where the biological conditions of interest are significantly confounded with the factors producing technical variation, common visualisation approaches, e.g. PCA plots, can be poor tools for this purpose, since such plots are sensitive to both technical and biological effects. Relative log expression (RLE) plots are sensitive to technical effects but insensitive (or highly robust) to biological effects, which makes them a useful tool for visualising technical variation in situations involving confounding. While RLE plots are a simple extension of standard boxplots, they have several significant advantages.