Fig. 8

Schematic diagram illustrating CIDANE’s work flow. In the first phase, a linear model is fitted (black line in c) to a compact representation of the observed read mappings (a) using an initial set T _init of candidate transcripts (b). Second, transcripts not in T _init that can help to improve the prediction are iteratively identified as optimal solutions to the heaviest isoform problem (d). The newly constructed isoform (e) is used to adjust the fitting (orange line in c)