Using `--subassemblies` to merge longitudinal short-read assemblies

[fconstancias]

Hi Flye team,

I’m working with longitudinal fecal samples (same individuals, multiple timepoints). I am interested in AMR genes spread so MGE (plasmids, phages) and MAGs.

When I perform single-sample assemblies, I overall detect more plasmids (either by mapping against PLSDB or using geNomad) than in a co-assembly of all samples of the same individual. Therefore, I am tinking about keeping single-sample assemblies and dereplicate / merge contigs across timepoints to reduce redundancy and obtain the most complete contigs/plasmids.

Is it appropriate to use Flye --subassemblies for this? Any known limitations or parameter recommendations for this use? Or alternatives?

Thanks for your guidance!

[mikolmogorov]

Sorry for the late response! I think this might work, but I would try alternative dereplication strategies as well. Or maybe construct a pooled assembly first, then map individual assemblies to it and only keep contigs that have no mappings against the pooled assembly. Note that --subassmeblies does not get much testing these days, but it is essentially very similar to other assembly modes with tweaked --read-error parameters.