The basis and pipeline of circRNA identification in CIRI. (A-C) CIGAR signals identified in the first scan. (A) For most circRNAs, two segments of junction reads align to the reference sequence separately in reverse orientation. (B) If one segment is longer than the exon flanking junction, the rest of the segment can align to the nearby exon contained in the circRNA. (C) If the circRNA length is shorter than the read length, two terminal segments will possibly align to the termini of the area where the middle segment aligns. (D) To reduce the false positive rate, candidate circRNAs are filtered based on the following information: i) PEM signals - the paired read of a junction read should align within the inferred circRNA area; ii) GT-AG splicing signals should be present in the inferred junctions; iii) mapping statistics - mapping quantity and quality, and mapping read length in the junctions. (E) The CIRI pipeline for detecting circRNAs from transcriptome data. (DP: dynamic programming).