Evaluation of statistical methods for normalization and differential expression in mRNA-Seq experiments

By James H. Bullard, Elizabeth Purdom, Kasper D. Hansen, Sandrine Dudoit. BMC Bioinformatics, 2010

1,760 citations recorded by the source corpus.

BackgroundHigh-throughput sequencing technologies, such as the Illumina Genome Analyzer, are powerful new tools for investigating a wide range of biological and medical questions. Statistical and computational methods are key for drawing meaningful and accurate conclusions from the massive and complex datasets generated by the sequencers. We provide a detailed evaluation of statistical methods for normalization and differential expression (DE) analysis of Illumina transcriptome sequencing (mRNA-Seq) data.ResultsWe compare statistical methods for detecting genes that are significantly DE between two types of biological samples and find that there are substantial differences in how the test statistics handle low-count genes. We evaluate how DE results are affected by features of the sequencing platform, such as, varying gene lengths, base-calling calibration method (with and without phi X control lane), and flow-cell/library preparation effects. We investigate the impact of the read count normalization method on DE results and show that the standard approach of scaling by total lane counts (e.g., RPKM) can bias estimates of DE. We propose more general quantile-based normalization procedures and demonstrate an improvement in DE detection.ConclusionsOur results have significant practical and methodological implications for the design and analysis of mRNA-Seq experiments. They highlight the importance of appropriate statistical methods for normalization and DE inference, to account for features of the sequencing platform that could impact the accuracy of results. They also reveal the need for further research in the development of statistical and computational methods for mRNA-Seq.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/cb66bc153e0dc8c5b94032781b40d6ca00bfad88

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome (2011, methodology): “Following [37], a gene was considered expressed if 75% of its probes passed the detection threshold.”
baySeq: Empirical Bayesian methods for identifying differential expression in sequence count data (2010, methodology): “The library size is a measure of the total number of counts in a given library, or some surrogate measure of library size as discussed by Bullard et al [12], and is used as a scaling factor for the observed data.”
Computational methods for transcriptome annotation and quantification using RNA-seq (2011, methodology): “Note: Supplementary information is available on the Nature Methods website. aCKnowLedGments expression depends on the sequencing depth of the sample, the expression of the gene, and even the length of the gene 64,68 .”
A comprehensive evaluation of normalization methods for Illumina high-throughput RNA sequencing data analysis (2013, methodology): “Even in cases where a strong positive association between gene counts and length is observed, scaling counts by gene length with RPKM is not sufficient for removing this bias [16, 20].”
cn.MOPS: mixture of Poissons for discovering copy number variations in next-generation sequencing data with a low false discovery rate (2012, background): “Despite these successes, quantitative analyses of NGS data, for instance, determination of the expression levels of genes, are still challenging (4,5).”
Comprehensive evaluation of differential gene expression analysis methods for RNA-seq data (2013, background): “This can result in erroneous differential expression [24,25].”
RNA Sequencing and Analysis (2015, neutral): “Each model makes speci ﬁ c assumptions that may be violated in the context of the observed data; therefore, an understanding of the model parameters and their constraints is critical for drawing meaningful and accurate biological conclusions (Bullard et al. 2010).”
De novo transcript sequence reconstruction from RNA-Seq: reference generation and analysis with Trinity (2013, methodology): “For example, Trinity now supports tools (e.g., RSEM 14 , edgeR 15 and DESeq 16 ) that take its output transcripts and test for differential expression while accounting for both technical and biological sources of variation 17–19 and correcting for multiple hypothesis testing.”
Biases in Illumina transcriptome sequencing caused by random hexamer priming (2010, background): “sapiens (9), we found that reweighting the reads improves their uniformity along expressed transcripts, although substantial heterogeneity remains (Figure 4 and Supplementary Figures S6–S8).”
Normalization and microbial differential abundance strategies depend upon data characteristics (2017, methodology): “(none, proportion, rarefying without replacement as in the hypergeometric model [75], CSS [4], logUQ [28], DESeqVS [30], and edgeR-TMM [27]) and the beta”
svaseq: removing batch effects and other unwanted noise from sequencing data (2014, methodology): “For sequencing data, which are often represented as counts, a more suitable model may involve the use of a moderated log function [17, 18].”