Fig. 4
From: Benchmarking splice variant prediction algorithms using massively parallel splicing assays
Transcriptome-normalized sensitivity. A Example shown for SpliceAI. Upper panel shows SpliceAI scores for the 500,000 background set variants (teal histogram) and the cumulative fraction (black line) of variants above a given score threshold (10% of background set variants with deltaMax ≥0.06). Below, histograms of SpliceAI scores for BRCA1 SGE benchmark variants, either SDVs (middle) or splicing-neutral variants (bottom), and the resulting transcriptome-normalized sensitivity and specificity at a deltaMax cutoff of 0.06. B Transcriptome-normalized sensitivity (at 10% background set SDV) versus within-benchmark variant set prAUC by benchmarked dataset. C Transcriptome-normalized sensitivity on benchmark exonic and intronic variants plotted as a function of the percent of the background variant set called SDV