Rrna Removal In Rna-Seq Data

Question

[Nicolas Rosewick]

Hi,

I want to know how many reads are coming from rRNA in my data. My librairies are done with illumina TruSeq (Stranded) and RiboZero. So here my idea.

In UCSC Tables :

Select "All Tables" from the group drop-down list Select the "rmsk" table from the table drop-down list Choose "GTF" as the output format Type a filename in "output file" so your browser downloads the result Click "create" next to filter Next to "repClass," type rRNA Next to free-form query, select "OR" and type repClass = "tRNA" Click submit on that page, then get output on the main page

Now I've a gtf file with the rRNA and the tRNA

After that, I use htseq-count to extract the number of reads per rRNA and tRNA gene.

Is that ok ?

Thanks,

N.

[Sean Davis]

Looks OK to me without checking the details.

Answer 1

[Ashutosh Pandey]

You can also download RepBase (database of repetitive elements) and Rfam (database for different RNA species, http://rfam.sanger.ac.uk/) and use it as a filter database.

Answer 2

[Ryan Dale]

Your post-alignment filtering strategy should work. Another strategy is to do a separate alignment of unaligned reads to rRNA sequence. Since rRNA genes tend to be duplicated in eukaryotes, it's possible that highly multi-mapping reads are discarded (depending on the aligner and parameters you use) such that those reads don't make it into the final alignments you would use with htseq-count.

Also, if your goal is to remove rRNA reads from downstream analysis, and you use the post-alignment filtering strategy, you may want to go back and remove other alignments of that read that multi-mapped to non-rRNA regions.

I don't have a feel for how different these strategies are though -- it's possible they get you roughly the same answer in the end.