Fig. 3

Alignment of full-insert SMRT sequences to identify reliable gene structures. Multiple independent SMRT reads derived from the same gene were used to (a) confirm genes previously predicted using AUGUSTUS default parameters and to (b) identify new gene models without prior annotation. Gene predictions were considered as validated if all aligning SMRT sequences indicated the same intron boundaries. For new gene models the most abundant isoform per locus supported by at least two reads was reported. c Prediction artefact through intronic transposable elements and corrected prediction in BeetSet-2. Numbers next to gene names indicate the percentage of predicted gene features supported by expression evidence