DRIMSeq

The software tool "DRIMSeq" introduces a statistical framework based on the Dirichlet-multinomial distribution to analyze multivariate responses in genomics data. Designed explicitly for scenarios involving alternative splicing and single nucleotide polymorphisms (sQTL), the method enables the detection of changes in isoform usage between conditions and identification of SNPs influencing relative transcript expression. Leveraging RNA sequencing and accurate transcript quantification methods, DRIMSeq employs a Dirichlet-multinomial model that naturally accommodates differential gene expression while preserving information about overall gene abundance. By jointly modeling isoform expression, the tool addresses correlated nature, and its approach to handling limited replicates improves existing methods. DRIMSeq applies not only to alternative splicing scenarios but also to other multivariate scenarios, such as Poly-A-seq or applications involving beta-binomial models (e.g., differential DNA methylation).