Fig. 2

Schematic representation of the classification of lncRNA into genome context-based subtypes. Annotations were categorised by genomic context using a decision tree. LncRNAs that overlapped a gene on the same strand were classified as either intronic if contained within the intron or sense if contained within a single exon. No lncRNAs were annotated that spanned multiple exons in a gene. LncRNAs that overlapped a UTR and lncRNAs nearby genes (within 150 bp of an annotated UTR or exon or read from the gene) were flagged as potential UTR-associated lncRNAs. To delineate UTR-associated lncRNAs from UTR transcripts (that could be fragmented due to drops in GC content or alternative start sites) careful examination of collative data was performed. This included an analysis of the level of overlap between reads from the putative lncRNA and gene/UTR, the presence of a unique transcriptional start site (distinct from the gene) and the lack of evidence of a drop in GC content. LncRNAs that were antisense (opposite strand) to genomic features were classified based on the type of antisense genomic feature: antisense-to-gene, antisense-to-intron, antisense-to-UTR and antisense-to-lncRNA. The antisense-to-intron lncRNAs were contained within the intron boundaries (with little to no overlap with the exon). The antisense-to-UTR lncRNAs only overlapped the UTR, not the exons and the level of overlap varied. Some lncRNAs could be classified as multiple subtypes if overlapping multiple features – the classification has a hierarchy starting with: intronic, sense, UTR-associated, antisense-to-intron, antisense-to-gene, antisense-to-UTR and antisense-to-lncRNA. LncRNAs not overlapping, antisense to, or nearby (150 bp) any feature were classified as intergenic