VelvetK long Kmers

Question

[samuel.lipworth]

I have been using velvetK to estimate optimal Kmer lengths and then plugging these into velvet optimiser. When I then annotate these assemblies (using Prokka) and stick them through Roary, they seem to group by assembly length rather than true phylogeny. I know the reads are not nonsense because when I assemble them using a reference genome the phylogeny makes sense so I think the problem is my denovo assembly. Any ideas as to how I could improve this?

[Brian Bushnell]

It would be helpful to describe what you are trying to assemble; the sequencing platform; library prep; run mode / read length; preprocessing you have attempted; etc.