Improving the Alternative Splicing Annotation of Plant Genomes

Next generation sequencing has significantly accelerated the pace at which new genomes are being produced and annotated. Deep sequencing of cellular RNA (RNA-seq) is making it possible to characterize the transcriptome of an organism in great detail, under a variety of conditions. However, the repertoire of splicing variation remains largely uncharacterized in most species. The main reasons are challenges in data analysis that are not adequately addressed by existing tools, and the ensuing lack of a systematic effort to catalog the variation. Current bioinformatics methods can identify most major isoforms of a gene but have difficulty capturing finer splicing variations, or certain types of variations such as intron retentions, which are hard to distinguish from mapping and sequencing artifacts. We present a resource, SpliceBox (http://sourceforge.net/projects/splicebox/), to facilitate the annotation of both full-length splice variants and local events from RNA-seq data. The SpliceBox suite includes CLASS, a tool for assembling RNA-seq reads aligned to a genome, and ASprofile, a package for identifying, quantifying and comparing events across samples. CLASS captures significantly more splicing variation than existing methods at comparable precision. It is also efficient, using a splice graph model to compactly represent a gene coupled with a dynamic programming algorithm to select informative transcripts, and can be used on very large data sets without sacrificing accuracy. We illustrate by applying the tools to publicly available RNA-seq resources to improve the annotations and create a first repertoire of splicing variations in the sequenced Rosaceae genomes. Acknowledgment: NSF awards ABI-1159078 and IOS-1339134.