Question: How do I know if I have DNA contamination in RNA-seq data?
gravatar for blur
3 months ago by
European Union
blur110 wrote:


I want to know whether I have DNA contamination in my RNA-seq data. I have a library with DNAse and one without DNAse and another pair with additional treatment [DNAse(+) vs. DNAse(-), DNAse(+)+X vs. DNAse(-)+X]

I tried:

1) Checking overall alignment rate of DNAse(+) vs. DNAse(-) and DNAse(+)+X vs. DNAse(-)+X] It is slightly higher on DNAse free samples

2) Checking alignment vs. transcripts It's all over the place - DNAse(+)+X has lowest rate, then DNAse(-) (without treatment), then DNAse(+) (without treatment) and finally the least background was in DNAse(-)+X... But I figured that might be due to rRNA? If it had not been sufficiently cleaned in some libraries it would not map to transcripts.

3) Looking at the reads in IGV after alignment to see if there is background/baseline in non-exon areas.

I am out of ideas, anyone know what else I can do?


rna-seq contamination • 263 views
ADD COMMENTlink written 3 months ago by blur110

Hi blur,

Try RSeQC's With DNA-contamination, you'll observe more introns and more intergenic 'tags'.



ADD REPLYlink written 3 months ago by michael.ante3.3k

Basically, you have tried almost everything. I would suggest a massive effort on step 3.

For example you might count the number of reads that align over introns in all your samples. The higher the number of such reads the higher the DNA contamination (assuming that the proportion of pre-mRNA in your samples is the same. I guess it was total RNA (not poly-A selection).

I also guess you didn't remove rRNA.

I am not sure, but maybe filtering rRNA reads might give you better estimates.

ADD REPLYlink written 3 months ago by Fabio Marroni2.2k

Thanks, I'll try it all (the library is after experimental rRNA depletion, but I guess it never hurts to check again)

ADD REPLYlink written 3 months ago by blur110
  1. Build intronic sequences fasta
  2. Index it
  3. Use fastqscreen
  4. You should see only small % mapping to the test reads. blur
ADD REPLYlink written 3 months ago by cpad011211k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1911 users visited in the last hour