Blind spots of quantitative RNA-seq: the limits for assessing abundance, differential expression, and isoform switching

Table 1 Workflow steps towards quantitative RNA-seq together with example applications

Input material	Unbiased random transcript fragments (hom. coverage)	Coverage bias (inhom. coverage)	Variable transcript start + poly-A (mod. TSS + polyA)
Alignment	Global; Transcriptome + Genome [tophat]	Local; Transcriptome + Genome [STAR]	Global; Transcriptome only [RSEM]
Abundance	Read count (include multi-reads) [GenomicRanges: countOverlaps]	Read count(ignore multi-reads [HTSeq]	Isoform abundance model (resolve multi-reads [RSEM]
Differential expression	Significance using a negative binomial count model [edgeR:exactTest]	Log-ratio effect size
Isoform switching	Differential isoform fractions [cuffdiff]	Differential splicing modules [DiffSplice]	Differential exons [DEXSeq]

We analyze the impact of the different types of input material and the subsequent data analysis steps on the results of quantitative RNA-seq. For each step we investigate approaches that we consider as representative for a given analysis strategy.

ISSN: 1471-2105