K-mer usage in metagenome assemblies

Question

[v.berriosfarias]

Hello just want to ask what should be the considerations at choosing a specific k-mer size on metagenome assemblies pipelines (particularly based on de Bruijn graphs)? I have read an issue here https://github.com/ablab/spades/issues/251 where coverage seems to be a great determinant on the k-mer choose, but i want to know more, what are your recomendations?

Some researches have told me that a good practice is to dont modify the k-mer size of the fragments unless one have idea of what the result is going to be.

so, what are your considerations?

Answer 1

[Rayan Chikhi]

The key sentence in the link you point out is: "use default" :) Short-read metagenomics assemblers are multi-k and their default choices are often very good. Never assemble with a single k with those assemblers.

Answer 2

[onestop_data]

This tutorial shows how to estimate the best k-mer size for assembly – Step-by-Step

[v.berriosfarias]

oh, thanks so much :)