We evaluated the stringency with which our filtering process removed protein coding transcripts by analyzing ribosomal profiling data from HeLa cells (Figure 2B) [22]. As expected, lincRNAs resemble the 3′ untranslated region exons of protein coding genes, with very few transcripts showing significant engagement with the ribosome. This finding is in agreement with the recent observation that GENCODE long noncoding RNAs (a subset of our catalog) generally lack mass spectrometry based evidence for translation [23]. In contrast, a recent study found that many previously annotated mouse lincRNAs bind the ribosome [24]. While the biological significance of this discrepancy is unknown, it may be the result of differences in the stringency of the filtering approach employed in the generation of the lincRNA annotations under consideration. Further confirming the stringency of our filters, a computational analysis of protein coding potential using the program PhyloCSF revealed that our set of filtered lincRNAs lack predicted protein coding capacity (Figure 2C). From these analyses we conclude that our filtering approach effectively removed protein coding transcripts from the catalog.
