Shared and distinct exonic parts in alternative paths of splicing bubbles

Detecting alternative splicing from complex splicing bubbles that contain more than two transcript paths are challenging.

We present GrASE, a framework that distinguishes alternative path groups by the distinct and shared exonic parts they contain.

Using lower-set bipartitioning, which efficiently enumerates downward-closed sets of paths, we find bipartitions of complex bubbles with well-defined distinct exonic parts.

Applied to the human genome, GrASE tests 496,565 path groups from valid bipartitions across 22,299 genes, including alternative TSS and TTS events that are inaccessible to junction-based methods.

GrASE can trade recall for higher precision and still detect substantially more events, due to its expanded test universe.