Reducing the structure bias of RNA-Seq reveals a large number of non-annotated non-coding RNA

The study of RNA expression is the fastest growing area of genomic research. However, despite the dramatic increase in the number of sequenced transcriptomes, we still do not have accurate estimates of the number and expression levels of non-coding RNA genes. Non-coding transcripts are often overlooked due to incomplete genome annotation. In this study, we use annotation-independent detection of RNA reads generated using a reverse transcriptase with low structure bias to identify non-coding RNA. Transcripts between 20 and 500 nucleotides were filtered and crosschecked with non-coding RNA annotations revealing 115 non-annotated non-coding RNAs expressed in different cell lines and tissues. Inspecting the sequence and structural features of these transcripts indicated that 60% of these transcripts correspond to new tRNA and snoRNA genes. The identified genes exhibited features of their respective families in terms of structure, expression, conservation and response to depletion of interacting proteins. Together, our data reveal a new group of RNA that are difficult to detect using standard gene prediction and RNA sequencing techniques, suggesting that reliance on actual gene annotation and sequencing techniques distort the perceived architecture of the human transcriptome. RNA was isolated from SKOV3ip1 human cell lines and various normal human tissues from which fragmented ribo-depleted TGIRT-Seq libraries were made. The resulting libraries were multiplexed and paired-end sequenced using Illumina HiSeq.