Semi-supervised Learning Predicts Approximately One Third of the Alternative Splicing Isoforms as Functional Proteins

Abstract

Alternative splicing acts on transcripts from almost all human multi-exon genes. Notwithstanding its ubiquity, fundamental ramifications of splicing on protein expression remain unresolved. The number and identity of spliced transcripts that form stably folded proteins remain the sources of considerable debate, due largely to low coverage of experimental methods and the resulting absence of negative data. We circumvent this issue by developing a semi-supervised learning algorithm, positive unlabeled learning for splicing elucidation (PULSE; http://www.kimlab.org/software/pulse), which uses 48 features spanning various categories. We validated its accuracy on sets of bona fide protein isoforms and directly on mass spectrometry (MS) spectra for an overall AU-ROC of 0.85. We predict that around 32% of "exon skipping" alternative splicing events produce stable proteins, suggesting that the process engenders a significant number of previously uncharacterized proteins. We also provide insights into the distribution of positive isoforms in various functional classes and into the structural effects of alternative splicing.

FFPred 3: feature-based function prediction for all Gene Ontology domains

“…In turn, those efforts will definitely stimulate further bio-curation work to interpret this information and make it available in machine-readable format. Initial computational studies have been carried out to advance this area of functional genomics using gene expression profile data2728; their integration with other complementary sources of biological information that are tissue and condition-specific will undoubtedly be the focus of many more investigations in the near future.…”

Splice junction centric approach to identify translated noncanonical isoforms in the human proteome

“…A clear picture on how alternative splicing impacts proteomes has yet to emerge but recent reports suggest that splicing can rewire interactomes (Yang et al 2016) and also serve as a mechanism to decrease the abundance of the major canonical isoform of a gene (Liu et al 2017). It is appreciated that only a minor subset of expressed transcripts have the potential to be translated, with about one-third of skipped exon events having been estimated to preserve protein structure and hence have higher translational potential (Hao et al, 2015), whereas the rest may be removed by nonsensemediated decay or co-translational proteasomal degradation (Weatheritt et al, 2016). The ability to empirically discriminate which isoform transcripts exist at the protein level at a particular tissue will aid in elucidating the biochemical and signaling functions of alternative isoforms.…”

Systematic evaluation of isoform function in literature reports of alternative splicing

“…Hao et al (Hao et al, 2015) developed a machine learning algorithm (PULSE) that predicted 1/3 of human genes have more than one functional isoform. We hypothesized that our curated genes with FDSIs would be enriched among those predictions, because even though Hao and colleagues were not attempting to predict functional distinctness, genes with FDSIs by definition have more than one functional isoform.…”

“…Beyond identifying knowledge gaps, establishing a set of genes with FDSIs provides potential avenues for improving computational approaches to analyzing alternative splicing. For example, classifiers, such as PULSE, attempt to predict genes with multiple functional splice isoforms (Hao et al, 2015 Here we present a literature-based analysis of experimental evidence for functionally distinct splice isoforms (FDSIs) for over 700 human and mouse genes. Despite a gene selection strategy that was highly biased towards genes suggested to have multiple functional isoforms, we found good experimental evidence for FDSIs for fewer than 10% of genes.…”