We also assembled 18,498 (Dataset S1b) experimentally supported (with full-length cDNA, Broad Institute RNAseq, or Gencode-curated cDNA and/or expressed sequence tag, i.e. EST, evidence), non-redundant (with respect to genomic position and orientation) lncRNA genes from six published sources: 1) our manually annotated list of human lncRNAs that are supported by full-length cDNA clones from 5′cap-trapped, dT-primed libraries [5]; 2) the Broad Institute lincRNA resource consisting of transcript assemblies inferred from exhaustive RNAseq of a human tissue collection [67]; 3) the ENCODE Consortium's official Gencode catalog [70] of human genes (www.gencodegenes.org), a manually curated list of coding and non-coding genes that are supported by full-length cDNA, EST, RNAseq, and targeted-RTPCR evidence from public sources as well as from Gencode's ongoing validation efforts. We enhanced this collection with non-redundant lncRNAs from three additional sources: 4) NCBI RefSeq (NR identifier) non-coding transcripts that do not host any known small RNAs according to the UCSC Genome Database sno/miRNA repository, 5) human ESTs from the dbEST division of Genbank (NCBI) [71] that were submitted by RIKEN (Japan) and that mapped beyond 10 kb from any
