I have two sets of PacBio reads from genomic DNA of an Aspergillus species that were made from separate preps of the culture. One of them has two additional peaks at 38% and 60% in the percent GC histogram produced by FastQC. Do these additional peaks indicate that there is contamination in that sample?

enter image description here

  • 1
    $\begingroup$ Yeah the prep looks odd, but you cannot really tell its contamination. You have to try and map the reads first, and see how many of them are unmapped, and see where the unmapped reads come from. You can just try with 1000 reads, should be obvious if it is contamination $\endgroup$
    – StupidWolf
    Feb 27, 2020 at 16:32
  • $\begingroup$ Remake the graphs using only mapped reads. Then you can decide on your next steps. $\endgroup$
    – Supertech
    Dec 28, 2023 at 22:10

3 Answers 3


The logic I would us is that if they come from the same genome, coverage peaks will go in multiplications of 2 - 1n, 2n, 4n, ..., while if it's a contamination, the coverage will go wild. So, I would recommend to look at the coverage vs GC to decide where it comes from.

You could either assemble the genome and use something like Blobtools to check coverage, GC and putative origin of sequences (and you should do that anyway).

Perhaps a bit simpler would be to look at it using k-mers, but that assumes you have reads with low error rate (In your case PacBio CCS or HiFi). If you do, you can use KAT that can do GC vs coverage plot - they intended it to look at GC bias in sequencing, but the vary same plot should show you if the three GC peaks have the same coverage or not.


I would suggest plugging your reads into a fast k-mer matching tool like mash screen to get a basic idea of where your reads come from, if you suspect contamination.

An issue with just looking at GC content is that some organelles/plasmids/etc. such as mitochondria have quite divergent GC content, and the different populations you observe might just indicate random fluctuations of copy number of these DNAs due to different sample preps. Not sure that would explain your plot but it does give me pause.

As suggested above, assembly is obviously a good idea if you are ok with the compute cost, and then you can use something like SPAdes or blobtools to pull out contigs from different origins.

KAT and similar tools as suggested are handy for hypothesis free evaluation of different populations in your reads, but they don't directly address the issue of contamination.

And obviously measure the quality of the reads to establish that they are ok and nothing weird happened there, they aren't all Gs or something.


Yes, it looks odd, but you might need to assemble the sequences to tell what is happening. The fact that the peaks are symmetrical around the presumably Aspergillus peak around 50% GC is strange. Usually contamination appears as one peak or shoulder on either side of the host DNA peak rather than symmetrical peaks.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.