Using Expression Data for lincRNA Annotation

In recent years, RNAseq technology has allowed for the in depth study of the transcriptomes a wealth of species. Transcriptomic studies revealed the complexity of eukaryotic transcriptomes. In Ensembl, we use RNAseq data in order to demonstrate an extensive expression-based annotation. In the past we have focused on using these data to annotate protein-coding transcript models, using our RNAseq pipeline.

More recently we have created a pipeline for annotating long intergenic non-coding RNA of vertebrate genomes, the lincRNA pipeline. In the first stage, we collect the models created by the RNAseq pipeline that do not overlap known protein coding genes. Then, we filter these models by looking for Pfam domains in all candidate protein sequences to determine the coding potential of the model. In addition, if need be, we BLAST sequences from the UniProt database against the candidate models in all six frames and filter out any models that have hits with significant e-values. Finally, if the RNAseq model is long enough and doesn’t have coding potential, it is labelled as a lincRNA.

Here, we demonstrate the lincRNAs found for sheep. Full annotation of the sheep genome was made public in Ensembl for release 76. It is comprised of 20921 protein-coding models, 3985 short noncoding RNAs and 291 pseudogenes. Due to an extensive transcriptomic data set provided by the International Sheep Genome Consortium, more than 800 GB data from 89 tissue samples which was used for the annotation, we identified 1858 potential lincRNA models.