CASIM: RNA-Seq Problems with rRNA contamination

[casim]

I would like to know if people think that ribosomal alignments (rRNA) are a problem for standard RNA-Seq (especially for normalisation) and whether there are any standardised procedures to get rid of them, e.g. align all reads against rRNA sequences and remove them from the genome wide alignment results.

[Blahah404]

Ideally, you eliminate the rRNA at the mRNA isolation stage of the wet procedure. Otherwise you're wasting money on sequencing rRNA you aren't interested in. However, assuming you've got reads with rRNA contamination...

I use the SILVA rRNA database and bowtie2 to filter rRNA reads after read QC and trimming, but before assembly/quantification. Specifically, I use the `--un-conc` option of bowtie2 to output only the reads that don't map concordantly to the rRNA database.

Anecdata follows:

If you're using RNAseq for transcriptome assembly, rRNA massively complicates the assembly graph and can make assemblies take many times longer and be less contiguous.

If just quantifying expression, the accuracy of expression is not greatly affected by rRNA contamination. But it takes longer to map the reads.

[buthercup_ch]

Hello,
I hope this post is still alive…
I'm having a similar problem after sequencing reaction of my cDNA libraries for RNA-Seq analysis to study differential expression.
The total RNA was treated with RiboZero kit for gram-negative bacteria, and a low amount of 5S rRNA was still remaining in the sample that were used to prepare the library.
Logically, I observe this contamination in my RNA-Seq data… in the form of a huge amount of reads mapping in a small region (116 bp) of the 5S subunit of one of the 5 rRNA operons that exist in the bacteria strain I'm studying.

Blahah404, what do you mean by "the accuracy of expression is not greatly affected by rRNA contamination"?? After quality control and trimming, about 18% reads mapped to the former small region of the 5S rRNA. This high % is not affecting the expression quantification???

Thank you very much

[Marianna85]

Hi buthercup_ch,

actually I have the same problem as you.
I think that if different samples, belonging to the same experiment, have more or less the same percentage of rRNAs (even if high) the expression is not greatly affected. You will have a minor coverage and probably you will lose the rare transcripts.

I think that the problem arises when you have to compare samples with quite different rRNA percentages. In this case rRNA reads removal, mapping to reference transcriptome and normalization allow a reliable DE analysis??
Does anybody have an idea?

Thanks
Marianna