rRNA contamination in RNAseq datasets

Question

[schelarina]

Hello,

i searched in the forum but did not found a clear answer.

I have a certain number of bacteria RNA samples that were sequenced using an illumina paired end approach. The library construction illumina total RNA + ribodepletion were performed by the sequencing provider. By using FastQC and sortMeRNA I found that all datasets are contaminated with rRNA to different %. Some datasets have 1%, while others have 10%, 20% and some even 50%! I understand that there was an obvious problem with the rRNA depletion. Still, I have more than 20M reads mapping to transcripts, in principle 5M reads are sufficient for DEG analysis in bacteria. My questions are:

These datasets could be still be used for DEG analysis if i remove the rRNA reads ? In general how much rRNA contamination is tolerated ? Is there something published in the litterature about rRNA contamination issue for DEG analysis?

thank you very much

[andres.firrincieli]

there is no need to remove the rRNA reads.

1. Map your reads against the reference genome and get a gene count
   table

2. Remove the rRNA genes from the gene count table

3. Normalize the data
4. Run a PCA to see if you have any outlier

Answer 1

[Hyper_Odin]

Your question has been answered before: What % of rRNA contamination is acceptable?

[schelarina]

10% of contamination is acceptable according to what discussed, but there is no article or study really reporting this threshold, and this is what I would like to find.

[GenoMax]

This is not simply about the rRNA % but what fraction of data remains, if you discount rRNA fraction. 10% of rRNA in a 1M reads is not the same as 10% of 5M aligned reads.Follow the recommendation by @andres and make an informed decision based on your data.

[schelarina]

the number of reads that do not align to rRNA go from a minimum of 6M to a maximum of 45M and about 95-97% of them align to mRNA.