Cross-species contamination in NGS data

Question

[T]

Dear all,

Given that in a certain sequencing study, a big majority of the reads are not mapping to the origins of interest (human / mouse & yeast), I am looking for a tool / approach check for cross-species contaminations of the reads.

A quick BLAST of some sequences revealed some bacterial RNA, but I want to classify all of the reads. Can you recommend me a tool / approach / best practice to do this high-throughput.

I have found a few online, but as far as I can see most of them are made for bacterial metagenomic studies. Probably some of you experienced users have a quickhack or a best practice.

Thank you very much.

[Devon Ryan]

Do you have some (ideally) small sent of species you want to check? If you want to check everything then you're largely restricted to blasting a smallish number of reads.

Answer 1

[GenoMax]

In this case I suggest binning the reads you are interested in and separating the "others" into a different bin. Take a look at BBSplit from BBMap which would be perfect for this. If you are interested in finding out what the "other" bin contains you can do that separately later.

[T]

Thanks, this is a very nice approach to get the "unmapped" reads.

However, the question what the "others" do would be the main point of the question.

[GenoMax]

So you are interested in content of "other" bin? Your original question made it sound like that was "contamination" (but not in real meaning of the word then?). Like @Devon said it would depend on if you expect only a few species to be present. Otherwise you would have to blast against refseq/bacterial in order to try and identify what is there.

Answer 2

[User 59]

You could try Kontaminant: https://github.com/TGAC/kontaminant

Answer 3

[h.mon]

I would get the unmapped reads, assemble them with metaSPAdes or MEGAHIT, and use GC content, mapping coverage and Blast searches to examine the assembled contigs - a great tool for this is blobtools.