Fig. 7

Annotation-aided two-pass alignment rescues missing splice junctions. a ROC scatterplot and b precision/recall scatterplot showing true-positive rate and false-positive rate of novel splice junction classification in simulated Arabidopsis read alignments, at different rates reference annotation missingness. Annotated transcript isoforms were subsampled to simulate incomplete reference annotations, and these were used to inform annotation-aided two-pass alignment. c, d Line plots showing the improvement in the percentage of correctly aligned reads using two-pass alignment compared to reference-guided alignment at different reference annotation missingness rates for c Arabidopsis and d humans, respectively. Blue line shows improvement compared to reads aligned using two-pass method only. Orange line shows improvement compared to reads aligned using reference-annotation in first-pass, followed by annotation-aided junction filtering and second-pass alignment. Shaded regions represent 95% confidence intervals