Normalization of RNA-Sequencing Data from Samples with Varying mRNA Levels

Abstract

Methods for normalization of RNA-sequencing gene expression data commonly assume equal total expression between compared samples. In contrast, scenarios of global gene expression shifts are many and increasing. Here we compare the performance of three normalization methods when polyA+ RNA content fluctuates significantly during zebrafish early developmental stages. As a benchmark we have used reverse transcription-quantitative PCR. The results show that reads per kilobase per million (RPKM) and trimmed mean of M-values (TMM) normalization systematically leads to biased gene expression estimates. Biological scaling normalization (BSN), designed to handle differences in total expression, showed improved accuracy compared to the two other methods in estimating transcript level dynamics. The results have implications for past and future studies using RNA-sequencing on samples with different levels of total or polyA+ RNA.

Variation in transcriptome size: are we getting the message?

“…Because mRNA is generally a small fraction of total RNA, the absolute expression of a given gene per unit of total RNA will differ from the absolute expression per unit of mRNA. Furthermore, the relative proportion of mRNA to total RNA varies (e.g., Aanes et al 2014). Consequently, the relationship between absolute expression per unit of total RNA and absolute expression per unit of mRNA will vary as well.…”

“…This approach, yielding what we call transcriptome-normalized expression, is based on the assumption that total expression per cell (= the total number of transcripts per cell, herein referred to as "transcriptome size") is equal in the samples being compared (Coate and Doyle 2010;Robinson and Oshlack 2010;Lovén et al 2012;Aanes et al 2014). As summarized below, however, transcriptome size variation appears to be widespread, having been observed in plants (Coate and Doyle 2010), animals (Aanes et al 2011(Aanes et al , 2014Islam et al 2011;Lin et al 2012;Lovén et al 2012;Nie et al 2012;Miettinen et al 2014), yeast (Zhurinsky et al 2010;, and bacteria (Chandler and Pritchard 1975). In comparisons between samples that differ in transcriptome size, transcriptome-normalized expression will differ, sometimes dramatically, from expression level per gene copy, per cell, or both.…”

“…For example, Illumina offers different RNA-Seq library kits that selectively target polyadenylated RNA, small RNA, or all nonribosomal RNA species. The various RNA pools differ dramatically in cellular abundance (Yang et al 2011;Qing et al 2013), and their abundances relative to each other can change across experimental conditions (Yang et al 2011;Aanes et al 2014). Thus, even transcriptome-normalized expression estimates will differ depending on the RNA pool to which they are normalized, as well as experimental context.…”

“…As described below, failure to recognize the distinction between expression per transcriptome and expression per gene copy, or the distinctions between expression estimates normalized to different circumscriptions of the transcriptome, has led to conceptual and terminological confusion regarding studies of gene expression in polyploid organisms and has also led to the misinterpretation of expression data examining a range of biological questions, including the relationship between gene dosage and expression in aneuploids and sex chromosomes (Birchler 2014) and the transcriptional responses of cells to cancers , Rett syndrome (Li et al 2013), embryonic development (Aanes et al 2014), and aging .…”

Perfil transcriptômico comparativo de macrófagos em cultura, infectados com isolados clínicos de Mycobacterium tuberculosis com diferentes perfis de resistência a quimioterápicos

“…mapeiam em mais de uma região é retirá-las das análises, dessa maneira foram consideradas apenas as sequências que mapearam em uma única região, evitando assim alterações nos níveis de expressão de alguns transcritos (LI et al, 2010 (AANES et al, 2014).…”