Fig. 3

a Comparison of annotated exon junctions among the transcriptomes sourced from RefSeq (hg38), Ensembl (GRCh38.p12), GENCODE (v29), and CHESS (v2.1). Horizontal bars: total junctions in each reference set; vertical bars: intersections of indicated reference sets. Bar graphs created with UpSetR [27]. b Sequence motifs for 5′ (left) and 3′ (right) splice sites used in annotated junctions observed in at least one analyzed library type (top) and for previously unannotated splice sites in indicated library type (bottom). Sequence logos were generated in R using ggseqlogo [28]; letter height signifies the relative abundance of that nucleotide at each position. N: number of splice sites contributing to each logo. Note that the number of unannotated junctions (6363) is greater than the total number of unannotated splice sites because many unannotated junctions combine an annotated and unannotated splice site (i.e., alternative 5′ or 3′ splice sites). c Venn diagram of annotated and previously unannotated junctions (numbers indicated) shared between library types. Venn diagrams made with eulerr [29]. d Cumulative histogram of exon junction reads (RPM) at annotated (solid line) and previously unannotated (dashed line) junctions in each library type (colors as in Fig. 1a). e Schematic of unannotated splicing events separated by event type: skipped exon (red); alternative 3′ (orange) or 5′ (yellow) splice site; new intron (light blue); new cassette exon (dark blue). N: number of observed events; for new cassette exons, both the number of observed unannotated junctions and number of new exons are shown