Suggestion : add checkv step beforehand to remove complete genomes

[michoug]

Hi,
I compared the results that I obtained using vibrant contigs before and after running phamb using checkv

Before

checkv_quality	n	mean	sum	max
Complete	557	46179.5	25721993	373392
High-quality	413	44008.8	18175622	275626

After

checkv_quality	n	mean	sum	max
Complete	351	54600	19164596	197996
High-quality	975	72413.1	70602775	622104

Where mean is the mean of contig length, sum is the total length of all contigs and max is the maximum size of the biggest “virus”.

This suggests to me that phamb wants to combine contigs even if they are considered "complete".
Maybe adding a checkv step beforehand to remove the complete ones from the phamb analysis could be useful?

[joacjo]

Hi Greg

Yes, I have seen cases where VAMB clustering results in complete virus contigs (DTR presence) are binned with unrelated contigs (not very frequent though), i.e. bin contamination. Based on your results, it looks like it is a matter of 557-351=206 complete virus contigs that ends up in a HQ bin, is my interpretation correct?

• I think you're absolutely right that an option would be to run a prior search on the individual contigs for the presence of complete virus contigs (i.e. a DTR search) then remove them from the set going into phamb. However, that would mean you end up running CheckV twice, which turns into a lot of computation time + money.
• An alternative idea that I would appreciate as a user, would be to scan viral bins for the presence of DTR's (complete virus contigs) then extract the complete virus and discard the "contaminating" contigs. Howver, if the discarded contigs frequently represents virus sequences this solution is not ideal.

I would appreciate your thoughts on this, thanks!

Best,
Joachim

[kellifeeser]

Regarding the above issue, I have found a few hundred contigs representing complete viral genomes (based off CheckV - DTR presence and high confidence only). I am using both VAMB and PHAMB. My questions are:

1. When would be the best time to remove these contigs from the dataset?
2. When would be the best time to add them back in (and how)?