Fig. 4

a Left panel: Recall of unique SIRV introns and intron-chains obtained by mapping reads to the SIRV genome before (Raw) and after correction with RATTLE, CONSENT, Canu, isONcorrect, and TranscriptClean (Clean) for the dRNA-seq (RNA 1) and the cDNA (cDNA 1, 2, 3) samples. Recall was calculated as the fraction of unique annotated introns or intron-chains exactly found by each method. Middle panel: precision values for introns and intron-chains for the same methods and datasets. Precision was calculated as the fraction of unique introns or intron-chains predicted by reads that matched exactly the SIRV annotation. Right panel: Read-precision for introns and intron-chains for the same methods and datasets. Read-precision was calculated as the fraction of all introns or intron-chains predicted in reads that corresponded to SIRV introns or intron chains. Similar plots for internal and external exons are given in Additional file 1: Fig. S5. Only cases with > 5 reads support were considered. b Same accuracy plots as in a for the same methods and datasets but using expressed Gencode transcripts as reference. c We plot the recall (green), precision (red) and read-precision (blue) of the SIRV introns (y-axis), as a function of the number of minimum reads supporting the predictions (x-axis). We indicate for each case the threshold at which a precision (red) of approximately 0.95 is achieved. For that threshold, we indicate the corresponding recall (green) and read-precision (blue). The plot corresponds to the dataset cDNA2. Results for other samples are available in Additional file 1: Fig. S7. d Same accuracy plots as in c but using expressed Gencode transcripts as reference. Results for other samples are available in Additional file 1: Fig. S9