Alternative splicing (AS) of precursor mRNA provides an important means of genetic control and is a crucial
step in the expression of most genes. AS gives rise to multiple isoforms that can exhibit differential stabilities,
molecular binding capabilities, and phenotypic effects, thereby greatly expanding the functional capacity of
genes. These functions are frequently deregulated in human disease, leading to aberrantly expressed isoforms
that can act as drivers and “rewirers” of cellular pathways. Given the intrinsic role of AS in nearly every aspect
of biology, tools and technologies for the functional understanding of isoforms are desperately needed.
Unfortunately, prediction of isoform functions are notoriously difficult, due to our only crude understanding of
the molecular determinants of isoform activities as well as a paucity of experimental datasets annotated at
isoform resolution. The experiments, in turn, are challenging in their own right, due to a lack of robust methods
for the detection, mapping, and phenotyping at isoform-resolution in vivo. As a consequence, one of the
biggest gaps in the genomics and proteomics fields is an understanding the functional and evolutionary
implications of the astonishing complexity of the human proteome.
To make progress towards this gap, we must reformulate existing approaches. Here, we propose to build
computational and experimental methodologies that are intertwined across the entire development and
evaluation lifecycle. Computational approaches, capitalizing on ever-improving machine learning algorithms,
can predict the effect of AS at high coverage. Experimental validation, on the other hand, is critical to
benchmark the predictions as well as shed light on heretofore uncharacterized features of isoform functionality.
The goals of this project are to develop (i) a predictor of protein isoform stability, a “first line of evidence” and
prerequisite of isoform functionality, (ii) a novel bioinformatics approach to study the “rewiring” effects of AS
isoforms on protein interactions, and (iii) a novel measure, the alternative splicing impact factor, that predicts
the functional role of an isoform based on metrics such as the loss of interaction and expression patterns, and
apply this concept to determine AS-induced phenotypes in-silico and in vitro/vivo (cell-based assay). Each
computational stage will be closely complemented with a highly customized and novel experimental approach–
large-scale isoform proteogenomics, interactomics, and functional assays experiments–to validate and
benchmark the predictors, as well as feed into an iterative computation-experimental “virtuous cycle”.
