Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Retained introns in long RNA-seq reads are not reliably detected in sample-matched short reads

Fig. 3

a Short-read tool performance across different thresholds of intron persistence. Each panel displays tool performance along the y-axis (measured by one of precision, recall, or F1-score as labeled) for a set of introns defined by the indicated threshold for intron persistence along the x-axis. Data for HX1 and iPSC are shown at left and right, respectively, with each tool’s per-sample performance depicted in a different color (IRFinder-S [red], superintronic [yellow], iREAD [green], IntEREst [purple], and KMA [blue], MAJIQ [light pink], rMATS [light green], and SUPPA2 [light blue]). b Variation in short-read tool performance across intron persistence thresholds for potential vs. called RIs. Each panel displays tool performance as measured by precision (left), recall (middle), and F1-score (right) for HX1 (top) and iPSC (bottom) samples. The performances for each tool’s potential RIs and called RIs are shown along the x- and y-axes, respectively, with centroid and whiskers denoting, respectively, the median and interquartile range of tool performance across intron persistence thresholds. Each tool’s performance is depicted in a different color (color labels same as 3a.). Reference lines are shown with slope of 1. c Varying degrees of consensus of retained intron calls among short-read tools. Bar plots depict the number of true positive (green), false positive (pink), and false negative (blue) intron calls (y-axis) consistent across a specified number of short-read (SR) tools (x-axis). Upper and lower panels depict HX1 and iPSC data, respectively. LR denotes long-read data

Back to article page