Global identification of human circRNAs. (A) Schematic illustration of the alternative-splicing isoforms generated from linear splicing (left) and back splicing (right). Two-part alignments identified junction-spanning reads indicative of circRNAs (bottom left). Exons are colored, and donor (GU) and acceptor (AG) signals at splice sites are indicated. (B) The computational pipeline developed to identify and quantify circRNAs from long-read RNA-seq data. (C) Enrichment of donor GT and acceptor AG splicing signals in genomic windows flanking candidate circular junctions supported by ≥5 junction-spanning reads in the CD34 sample. Similar results were obtained from all other cell types. (D) Distribution of circular fractions for circRNA candidates in (C), grouped based on whether their circular junctions were flanked by splicing signals of the major or minor spliceosome (GT-AG- and AT-AC-flanking, respectively). (E) Distributions of exon numbers for circRNAs, mRNAs, and other annotated ncRNAs. (F) Annotations of genomic regions mapping to inferred circRNA exons. CDS, coding sequence; lincRNA, long intervening ncRNA; UTR, untranslated region. (G) Splicing within circRNAs of the CD34 sample. Mapped locations of the mates of junction-spanning reads were compared to the genomic annotations 200 nucleotides downstream and upstream of back-spliced acceptors and donors, respectively. Because the fragment size for the paired-end sequencing averaged 200 nucleotides, these genomic annotations resembled those expected if the introns within the circRNAs were retained.