Modeling local splice variants from RNA-Seq gives a new view of transcriptome complexity and substructures

Alternative splicing of genes is a major contributor to transcriptome variations in metazoans and humans in particular. RNA-Seq, the main technology used nowadays to probe the transcriptome, provides a wealth of infomation yet accurate mapping of splicing variations has been a challenge plaguing research efforts in diverse fields. Taking a holistic approach to this challenge we developed a method (MAJIQ) to formulate local splicing variations (LSVs), quantify LSVs, and visualize these variations using a new interactive visualization package (VOILA). Applying MAJIQ/VOILA to several human and mouse datasets and over 200 RT-PCR experiments we demonstrate our new approach significantly improves LSV quantification accuracy and gives new insights into the transcriptome's substructures and complexity.