Revisiting the identification of canonical splice isoforms through integration of functional genomics and proteomics evidence

Abstract

Canonical isoforms in different databases have been defined as the most prevalent, most conserved, most expressed, longest, or the one with the clearest description of domains or post-translational modifications. In this article, we revisit these definitions of canonical isoforms based on functional genomics and proteomics evidence, focusing on mouse data. We report a novel functional relationship network-based approach for identifying the Highest Connected Isoforms (HCIs). We show that 46% of these HCIs are not the longest transcripts. In addition, this approach revealed many genes that have more than one highly connected isoforms. Averaged across 175 RNA-seq datasets covering diverse tissues and conditions, 65% of the HCIs show higher expression levels than non-highest connected isoforms (NCIs) at the transcript level. At the protein level, these HCIs highly overlap with the expressed splice variants, based on proteomic data from eight different normal tissues. These results suggest that a more confident definition of canonical isoforms can be made through integration of multiple lines of evidence, including highest connected isoforms defined by biological processes and pathways, expression prevalence at the transcript level, and relative or absolute abundance at the protein level. This integrative proteogenomics approach can successfully identify principal isoforms that are responsible for the canonical functions of genes.

Functional Networks of Highest-Connected Splice Isoforms: From The Chromosome 17 Human Proteome Project

“…27 A recent study suggested that a multi-isoform gene expresses only one dominant isoform at the protein level in a given tissue, providing evidence for functional differences between isoforms. 28 A brief overview of a variety of methods for identifying dominant/major/principal isoforms is in our previous work 29,30 …”

Genome-wide peripheral blood transcriptome analysis of Arab female lupus and lupus nephritis

“…Despite being the hotspot area, not much is known regarding the expression spectrum of alternatively spliced genes and it role in diseases progression. Very limited studies have characterized the functional role of splice variants with majority being centered on algorithm based computational methods and RNA-seq technology (Li et al, 2014a(Li et al, , 2014bEksi et al, 2013;Wang et al, 2008). In the present study, we aimed to investigate the genome-wide peripheral blood transcriptome of Arab female lupus and LN cases compared to healthy control.…”

Computational Inferences of the Functions of Alternative/Noncanonical Splice Isoforms Specific to HER2+/ER–/PR– Breast Cancers, a Chromosome 17 C-HPP Study

“…Hence, we used an integrated computational approach with tools for sequence alignment, 24 motif prediction, 25 structure prediction, 26 structure comparison, 27,17 and splice variant function prediction 28–30 to study the potential functions of the noncanonical splice proteins compared to those of the canonical proteins. GeneCards was used for gene level annotations (http://www.genecards.org/).…”

“…28–30 Briefly, by treating a gene as a “bag” of its isoforms of potentially different functions, MIL can build a model that is able to predict functions for isoforms using heterogeneous data and gene-level function annotation data. Essentially, instead of considering all isoforms of the genes annotated to a specific function, MIL selects a subset of these isoforms that show similar patterns to establish the classification model.…”