Annotation and differential analysis of alternative splicing using de novo assembly of RNAseq data

Genome-wide analyses reveal that more than 90% of multi exonichuman genes produce at least two transcripts through alternative splicing (AS). Various bioinformatics methods are available to analyze ASfrom RNAseq data. Most methods start by mapping the reads to anannotated reference genome, but some start by ade novoassemblyof the reads. In this paper, we present a systematic comparison ofa mapping-first approach (FaRLine) and an assembly-first approach(KisSplice). These two approaches are event-based, as they focuson the regions of the transcripts that vary in their exon content. Weapplied these methods to an RNAseq dataset from a neuroblastomaSK-N-SH cell line (ENCODE) differentiated or not using retinoic acid.We found that the predictions of the two pipelines overlapped (70% ofexon skipping events were common), but with noticeable differences.The assembly-first approach allowed to find more novel variants, including novel unannotated exons and splice sites. It also predicted ASin families of paralog genes. The mapping-first approach allowed tofind more lowly expressed splicing variants, and was better in predicting exons overlapping repeated elements. This work demonstrates thatannotating AS with a single approach leads to missing a large number of candidates. We further show that these candidates cannot beneglected, since many of them are differentially regulated across conditions, and can be validated experimentally. We therefore advocate forthe combine use of both mapping-first and assembly-first approachesfor the annotation and differential analysis of AS from RNAseq data.