The variancePartition workflow and implementation makes the rich linear mixed model framework easily applicable for interpreting drivers of variation in complex gene expression data. variancePartition provides a general statistical and visualization framework for studying drivers of variation in RNA-seq datasets in many types of high-throughput genomic assays including RNA-seq (gene-, exon- and isoform-level quantification, splicing efficiency), protein quantification, metabolite quantification, metagenomic assays, methylation arrays and epigenomic sequencing assays. Although we have focused here on large-scale studies, variancePartition analysis has given valuable insight into RNA-seq datasets with as few as 20 samples. The variancePartition software is an open source R package and is freely available on Bioconductor. The software can easily be applied to RNA-seq quantifications from featureCounts [46], HTSeq [47], kallisto [48], sailfish [49], salmon [50], RSEM [51] and cufflinks [52] which have been processed in R with limma/voom [15], DESeq2 [16], tximport [53] and ballgown [54]. The software provides a user-friendly interface for analysis and visualization with extensive documentation, and will enable routine application to a range of genomics datasets.
