Analysis of Splicing Aberrations in Human Tissue Transcriptomes

Recent advances in genomic studies and high-throughput sequencing have uncovered massive and heterogeneous repertoire of RNAs produced by alternative splicing. Although many alternative splicing isoforms are annotated and present in public databases, a growing number of unannotated isoforms is reported in each study. While some of them can represent natural variation of splicing across tissues and individuals, i.e., be splicing noise, others might be associated with disease. Alternative splicing patterns, even considered as noise, are expected to preserve reading frame in protein-coding RNA sequences and thus can be a valuable indirect source of information about translation of a transcript. In particular, micropeptides are a class of transcripts that are being extensively studied now because they were mistakenly annotated as non-coding transcripts, but in fact they contain short open reading frames. In this work we study splicing aberrations, defined as shifts of splice sites from the major splice site to a nearby unannotated splice site by not more than 30 nucleotides. We analyze a large panel of human RNA-seq data from Genotype-Tissue Expression Project (GTEx) and find distinct patterns of splicing aberrations in protein-coding and in non-coding regions. As expected, shifts by a multiple of three nucleotides preserving the coding frame are more frequent in coding regions compared to non-coding regions. Contrary to expectations, there are fundamental differences in the propensity of reading-frame-preserving aberrations among donor and acceptor splice sites, as well as different preferences in the upstream and the downstream shifts. Overall, the discovered patterns are consistent with previous observations and provide novel insights about splicing process, its output, and error rates.