Figure 4

Variation in Percent Spliced-In (PSI) and error rates. Volcano plots where ΔPSI is plotted against the -log10 p-value of the Fisher’s Exact Test, to assay variation due to sampling and sequencing error in simulations (A), technical RNA-Seq replicates (B), and biological replicates (C). (A) Plot comparing two simulated read sets of equal reads per kilobase (300 Reads Per Kilobase). Since transcript abundances are equal, expected ΔPSI is zero. (B) Variation in ΔPSI between replicate RNA-Seq runs of the same libraries. (C) Variation in ΔPSI between independent biological samples, each with a distinct RNA-Seq library. Results within each sex were similar, results for female samples shown. (D) False positive differential splicing calls in a null dataset. Cufflinks results are shown for both splicing analysis (Jensen-Shannon Divergence, JSD), and isoform abundance comparison. MISO results shown are based on isoform-centric analysis. (E) Volcano plot of ΔPSI calculated by Spanki for false negative analysis, comparing simulated datasets with PSI = 0.25 and PSI = 0.75 respectively, for 1644 events (expected ΔPSI = -0.50, red line). (F) Counts of false negatives for differential splicing calls in simulated data with an input ΔPSI = -0.50.