Why weight? Modelling sample and observational level variability improves power in RNA-seq analyses

Abstract

Variations in sample quality are frequently encountered in small RNA-sequencing experiments, and pose a major challenge in a differential expression analysis. Removal of high variation samples reduces noise, but at a cost of reducing power, thus limiting our ability to detect biologically meaningful changes. Similarly, retaining these samples in the analysis may not reveal any statistically significant changes due to the higher noise level. A compromise is to use all available data, but to down-weight the observations from more variable samples. We describe a statistical approach that facilitates this by modelling heterogeneity at both the sample and observational levels as part of the differential expression analysis. At the sample level this is achieved by fitting a log-linear variance model that includes common sample-specific or group-specific parameters that are shared between genes. The estimated sample variance factors are then converted to weights and combined with observational level weights obtained from the mean–variance relationship of the log-counts-per-million using ‘voom’. A comprehensive analysis involving both simulations and experimental RNA-sequencing data demonstrates that this strategy leads to a universally more powerful analysis and fewer false discoveries when compared to conventional approaches. This methodology has wide application and is implemented in the open-source ‘limma’ package.

Extracellular vesicles from early stagePlasmodium falciparum-infected red blood cells contain PfEMP1 and induce transcriptional changes in human monocytes

“…Counts were transformed to log 2 ‐CPM with associated precision weights using voom (Law, Chen, Shi, & Smyth, ). Sample‐specific quality weights were also calculated using limma's arrayWeights function (Liu et al, ). Differential expression was assessed using linear models and robust empirical Bayes moderated t statistics (Phipson, Lee, Majewski, Alexander, & Smyth, ).…”

limma: Linear Models for Microarray Data

“…Nucleic Acids Research 43, e97. Law et al (2014) describe the voom and limma-trend pipelines for RNA-seq, while Liu et al (2015) describe the voomWithQualityWeights function.…”

“…This capability is implemented in the voomWithQualityWeights function. The example below shows its use on an RNA-seq data set where the epigenetic regulator Smchd1 has been knocked-out in lymphona cell-lines (GEO series GSE64099) [21]. Overall we obtain more differential expression by applying this combined weighting strategy and the raw p-value and false discovery rate for the Smchd1 gene, which has been knocked out, is smaller.…”

“…The limma approach to RNA-seq explained in the articles by Law et al [15] and Liu et al [21]. See also the article by Law et al [14], which gives a complete workflow case study.…”

RNA-seq analysis is easy as 1-2-3 with limma, Glimma and edgeR

“…Where sample-level variation is evident from earlier inspections of the MDS plot, the function can be used to simultaneously incorporate sample-level weights together with the abundance dependent weights estimated by 14 . For an example of this, see Liu et al (2016) 15 .…”