MAJIQ V3: Scalable and accurate RNA splicing analysis with splicegraph algebra and improved intron retention modeling

Abstract

RNA splicing plays important roles in cell-type-specific expression of gene isoforms and is causally linked to numerous diseases. Previously, our group developed Modeling Alternative Junction Inclusion Quantification (MAJIQ) V1 and V2 for RNA splicing detection, quantification, and visualization from RNA sequencing. Here we report an update, MAJIQ V3, that offers critical new functionality together with faster and more accurate quantification algorithms. MAJIQ V3 enables users to combine splicegraphs, quantify splicing from an RNA-seq experiment with respect to a splicegraph, and compare one splicegraph to another, enabling new patterns of performant RNA splicing analyses from RNA-seq. MAJIQ V3 utilizes a new probabilistic model for detecting intron retention, correcting a bias that resulted from previous heuristics. MAJIQ V3 benefits from a full re-implementation utilizing more performant I/O patterns, data structures, and algorithms. We demonstrate the benefits of MAJIQ V3 using GTEx and LRGASP data for different analysis scenarios involving different dataset sizes.