In order to determine transcript expression levels, mapped RNA-seq reads were distributed to transcripts using a modified version of HTSeq v0.5.3p that allows for reads that are mapped to shared portions of overlapping transcripts to be counted as a full read for each overlapping transcript. This was necessary to properly assign reads to each of multiple redundant annotations of transcripts present in the combined set from all public databases and de novo assemblies prior to the merging of overlapping lincRNA annotations (described below). These redundant annotations are the result of the repeated de novo assembly of the same transcript in multiple different datasets or redundant existing annotations in public databases, each of which have slightly different genomic coordinates yet may represent the same transcript. As such, all reads were distributed fully to each redundant annotation rather than proportioned between them. Read counts were converted to FPKMs using total mapped reads for each dataset calculated by the SAMTools flagstat function and custom scripts. Transcripts not expressed at FPKM>1 in at least one dataset were removed. As a result of this FPKM>1
