# How to filter ribosomal RNA from scRNA-seq data

I want to filter out ribosomal RNA from scRNA-seq data (downloaded from here). Is there a list of known ribosomal RNA?

The only solution I found is SortMeRNA, however it works with raw sequencing data afaik, while I already have a matrix with transcript counts for each cell. I searched for a comprehensive list of rRNAs but I didn't find any.

• What GTF file did you use to generate the counts? For some species only 5S (and maybe MT) rRNA is left, for others you might have one or more copies of the various 48S components. – Devon Ryan Jan 2 '18 at 22:23
• I actually don't have GTF files. I downloaded the data provided by the Tabula Muris Consortium (figshare.com/projects/…) – gc5 Jan 2 '18 at 22:27

The rRNA genes in that dataset are Rn45s and Rn4.5s.

BTW, you have gene counts, not transcript counts.

• Thanks. What do you mean by gene counts instead of transcript counts? That the counts are already collapsed by gene? – gc5 Jan 3 '18 at 15:25
• Also, how did you find the list of rRNA genes in such dataset? – gc5 Jan 3 '18 at 15:32
• @gc5 The counts were likely only ever done by genes to begin with. rRNA genes typically have names starting with "Rn" in mouse. – Devon Ryan Jan 3 '18 at 15:34
• Ok. Should I need to find ribosomal genes in human do you know any database which I can use? – gc5 Jan 3 '18 at 15:42
• I think they have similar names in humans. – Devon Ryan Jan 3 '18 at 15:43

In the paper mentioned, we used the ScaleData function in Seurat to regress out the number of reads, Rn45s abundance, and percent ribosomal gene transcripts. Ribosomal genes were found with the regular expression ^Rp[sl][[:digit:]].

tiss <- ScaleData(object = tiss, vars.to.regress = c("nReads", "percent.ribo","Rn45s"))


Here's a fuller notebook, and we'll have a better organized repository soon.

• That's great that the workflow is available. I am just curious why did you decide to use ribosomal genes instead of the mitochondrial genes, which is suggested by Seurat? – burger Dec 18 '18 at 1:46
• Hello! Thanks for asking. In the final version of the paper (this question cites the preprint), we ultimately decided not to normalize on either ribosomal or mitochondrial genes. We found that this normalization added differences where there shouldn't be any, e.g. between cells of the same type in the Bladder. – Olga Botvinnik Dec 18 '18 at 21:51
• Thanks for clarifying! I was wondering why I could not find any mention of ribosomal genes anywhere in the methods (in the final paper). – burger Dec 18 '18 at 22:41