Fig. 1

Recall and precision for transcripts with different exon structure and unequal abundance rates. Violin plots showing the recall (a) and precision (b) of IsoCon and ICE. In each panel, the rows correspond to different families and, hence, different error rates. The shortest gene family (TSPY, labeled by the name of the reference gene copy used to generate it, TSPY13P), with correspondingly lower read error rates, is shown in the top panel rows while the longest gene family (DAZ, labeled as DAZ2), with correspondingly higher read error rates, is shown in the bottom rows. The columns correspond to different mutation rates (μ) used in simulating the gene copies (see Supplementary Note 2). A lower mutation rate implies more similar gene copies. Each plot shows results for a total of 30 isoforms with abundances randomly assigned and ranging from 0.1% to 15%. Within each plot, the x-axis corresponds to the number of simulated reads, while the y-axis shows the recall/precision of the methods. Each violin is generated using 10 simulated sequencing replicates. The white dot shows median, the thick black line is the interquartile range (middle 50%), the thinner black line is the 95% confidence interval, and the colored area is the density plot. We note that the density plot is cut at the most extreme data points