APAIQ outperforms other methods on PAS identification. (A) Number of the top 10,000 predicted PASs within different distances from the PAS in the ground truth. (B) ROC curves showing the performance of the APAIQ and four other published methods, including Aptardi, DaPars2, mountainClimber, and SANPolyA. (C) Venn diagram showing the number of true positives predicted by APAIQ and the four published methods. The number of overlaps between APAIQ and Aptardi, APAIQ, and SANPolyA were indicated by arrows. (D) Different categories of the PAS identified by APAIQ. (E) An example showing that the IPA event was successfully identified by APAIQ, but not the other three methods. (F) Genome Browser showing predictions of PAS from the genes DKK1 based on the hybrid model integrating DNA sequence and RNA-seq coverage, the model only using DNA sequence (sequence-only), and the model only using RNA-seq coverage (coverage-only).
