Compared to RNA-sequencing transcript differential analysis, gene-level differential expression analysis is more robust and experimentally actionable. However, the use of gene counts for statistical analysis can mask transcript-level dynamics. We demonstrate that ‘analysis first, aggregation second,’ where the p values derived from transcript analysis are aggregated to obtain gene-level results, increase sensitivity and accuracy. | Yi et al. Genome Biology 2018 19 53 https s13059-018-1419-z METHOD Open Access Gene-level differential analysis at transcript-level resolution Lynn Yi1 2 Harold Pimentel3 Nicolas L. Bray4 and Lior Pachter2 5 Abstract Compared to RNA-sequencing transcript differential analysis gene-level differential expression analysis is more robust and experimentally actionable. However the use of gene counts for statistical analysis can mask transcript-level dynamics. We demonstrate that analysis first aggregation second where the p values derived from transcript analysis are aggregated to obtain gene-level results increase sensitivity and accuracy. The method we propose can also be applied to transcript compatibility counts obtained from pseudoalignment of reads which circumvents the need for quantification and is fast accurate and model-free. The method generalizes to various levels of biology and we showcase an application to gene ontologies. Keywords RNA-sequencing Differential expression Meta-analysis P value aggregation Lancaster method Fisher s method Šidák correction RNA-seq quantification RNA-seq alignment Pseudoalignment Transcript compatibility counts Gene ontology Background A remedy to this problem is to estimate gene abun- Direct analysis of RNA abundance by sequencing comple- dances . in transcript-per-million units by summing mentary DNAs cDNAs using RNA-sequencing RNA-seq transcript abundances 7 but regularization methods for offers the possibility of analyzing expression at the resolution variance estimation of gene counts 8 cannot be directly of individual transcripts 1 . Nevertheless RNA-seq con- applied to abundances. For this reason recent workflows tinues to be mostly studied at the gene level partly because for gene-level differential analysis rely on converting gene such analyses appear to be more robust 2 and also because abundance estimates to gene counts 2 9 . Such methods gene-level discoveries are more experimentally actionable .
