Hey Joon, I am running viralmetagenome on default settings/databases with a clinical sample extracted twice (one with and without IC, the only difference). At the kraken step, in both samples, the pipeline identified the suspected virus. But only in one sample (non-IC) did it assembly the genome of the virus. Any idea why? The total reads in kraken2 was plenty for assembly: 12k and 60k.

It’s difficult to say, viralmetagenome heavily relies on getting good denovo contigs from spades or megahit to initiate the refinement from. If they only identify short fragmented contigs then there is little to refine, deduplicate, ... To verify if your followup comment is true (do our reads only cover sporadic parts of the genome), you can pass a multifasta of your suspected references file through

--mapping-constraints

see docs. Setting the selection to

true

so it uses mash to select the most similar reference towards your read dataset. Note, it will select only one reference per multifasta file. So if it’s segmented, you’ll need to make different files for each segment, to check the individual coverages.

Thanks Joon, great advice. viralmetagenome performed very well on all the samples btw. 🔥