Fig. 1.

Overview of the HBA-DEALS model. The log-transformed expression of a gene with three isoforms (green, orange, and blue) is shown. The gene expression is the sum of the expression of the isoforms. Differential gene expression is modeled as two Normal distributions whose means differ by the parameter β. The proportions of the corresponding isoforms have a Dirichlet prior, and the difference in proportions between controls and cases is modeled by α (symbolized by the two triangles). An MCMC procedure is used to solve for the posterior distribution of the parameters of the model for all genes and isoforms at once. The marginal posterior distributions are in turn interpreted to classify each gene as DAST, DGE, both, or neither (the “Methods” section). In this example, the gene shows both downregulation at gene level and a change in the dominant isoform in sample B