Basic question about lncRNA identification

Question

[ZZOYEA]

Hello, I am currently conducting an analysis to identify lncRNA in a human cell line. I have used STAR-cufflinks-cuffmerge pipeline and successfully identified novel lncRNAs.

What I want to ask is: how do I identify known lncRNAs, not novel ones? In one paper, I've seen that they used blastn to identify known lncRNAs in sheep samples by blasting against a database called ALDB.(A domestic animal long noncoding RNA database) Can the same method be applied to human cell lines or mouse cell lines to identify known lncRNAs? Or are there other methods to find known lncRNAs?

Thanks!

[dsull]

Known lncRNAs (such as Xist and Malat1) already exist in the standard human genome reference's GTF annotation.

If you just align to the human genome reference with STAR, you'll already have the expression levels of those lncRNAs.

Is there something else you're asking about? Edit: Ah, seems like you want to identify it from the output of cufflinks-cuffmerge. The answer below should give you that.

Answer 1

[matt.a.bennett25890]

You could check out GFFcompare, a tool which would allow you to align your de novo GTF with e.g. an Ensembl GTF

https://ccb.jhu.edu/software/stringtie/gffcompare.shtml

It outputs codes for various levels of transcript alignment, e.g. "=" for a complete intron chain overlap

Answer 2

[Darked89]

As pointed above, Gencode or ENSEMBL annotation contains known human lncRNAs. So the simple minded approach would be to i.e. extract lncRNAs exons and intersect the GTF with BAM. Or just extract read counts corresponding to lncRNA genes starting with featureCounts output.

The problem with it is that with not stranded RNA-Seq and lncRNAs overlapping parts of coding genes it may be non trivial to distinguish between say coding gene transcript with a retained intron vs lncRNA gene. Viewing the mappings in IGV may help, but this does not scale well.

[ZZOYEA]

Thank you! I will try it as you suggested!