The performance of different algorithms on ENCODE RNA-Seq datasets. (A) The number of predicted isoforms by ISP and Cuffmerge using multiple samples. The shaded region at the bottom of the bar shows the number of predicted isoforms that match UCSC human known transcripts. (B-E) The precision of both algorithms on multi-exon isoforms using all samples (B), and the precision of both algorithms on isoforms grouped by the number of exons (C-E). The plots in C, D and E show the corresponding precision for isoforms with 2-4, 5-10 and over 10 exons, respectively. (F-H) The sensitivity, precision and running time of ISP, Cuffmerge and MITIE on 500 randomly selected genes in chromosome 1.