Wonder if anyone knows a good (fast, ideally) read...
# meta-omics
j
Wonder if anyone knows a good (fast, ideally) read profiling tool that would be able to assign low-level eukaryotic taxonomy to reads - broad categories like "insects", "annelids", "mammals", etc.?
m
You can use any fast one, like sourmash, and then just go up the taxonomy, can't you?
j
That would definitely work - I will give that a try. However, I was curious if there was one with some nice pre-built database that would aid the process - I suspect I'll need to build my own DB as most read profiling tools seem geared towards prokaryotes
m
That's likely true, and your coverage of genomes will vary a lot.
j
The goal is to help determine the true coverage of a host genome in a given read set. I can very quickly paint on very broad categories with Tiara, but it would be very nice to be able to get a bit more specific in the case of eukatyotic co-biosis. (in this case, the target organism is a green alga at a coverage of about 50X - but there look to be other eukaryotic things in the sample)
g
Can't you use kraken/bracken for this with a premade database? It is probably overkill, but if you have the RAM for the
core_nt
database, it would do this
j
Definitely another option to try, but I think running this against core_nt would be too resource intensive for what I'm aiming for here. What I'm trying to do is devise a fast and resource-cheap way to quickly ascertain coverage - currently I can subsample 200k reads, get an estimate of their coverage with kmers + a loose domain classification in about half an hour. Ideally any additional steps would not take this beyond a couple hours total (this may be too big an ask!)
g
Hm, it would involve making a database, but I've found sylph to be pretty fast for this.
j
I don't think Sylph can give read-level classifications?
g
No, but if you are already doing a kmer based solution... I guess I don't know why you want a "true coverage" vs good estimate of true
j
Ah, to be clear - the goal is to get an estimate of the (homozygous) coverage of the target organism, as this is an important parameter to tune for genome assembly. This is easy in a clean sample, as you can just read this off a kmer plot - but for a contaminated sample, the contamination makes finding this value more complicated. Our samples often contain multiple diploid eukaryotes as well as a whole host of bacteria. Blobplots like the one I posted above can help, though, if you can annotate them well enough. So the goal is to find a way to add some basic taxonomic annotations that go beyond the domain level.