Inquiry

[songmj86]

Hi . I have been currently mining prophages from thousands of prokaryotic genomes with complete and high/medium-quality genome. I run Genomad with default mode and have a question on the example of the result shown in the file

Data.txt

Question

1. "provirus_3961739_3974913" is free of integrase. Can we call it prophage ?
2. n_hallmarks of "provirus_3780879_3799432" is zero, but it has high virus score. Am I supposed to deem it as false positive ?
3. Can you recommend any other "reliable" tool or method to cross-validate whether they are prophage as large discrepancy of result obtained from genomad and checkv is observed ?
4. (etc) It seems to be impossible to compute the genome quality of giant virus using genomad and checkv. Do you have any idea how to comupute it?

Thanks :)

[apcamargo]

Hi @songmj86

"provirus_3961739_3974913" is free of integrase. Can we call it prophage?

Not having an integrate detected by geNomad doesn't mean that the sequence is not a provirus. Several things could explain this: (1) the provirus is not complete; (2) the integrase was not detected by the profiles that geNomad uses; (3) the phage uses a different integration mechanism (e.g., a transposase).

n_hallmarks of "provirus_3780879_3799432" is zero, but it has high virus score. Am I supposed to deem it as false positive ?

I'd say you can consider that a provirus, yes. Maybe it is truncated and the hallmark is missing. However, the only way to be confident is by manual inspection. This is true for every provirus detection tool.

Can you recommend any other "reliable" tool or method to cross-validate whether they are prophage as large discrepancy of result obtained from genomad and checkv is observed ?

You can find the benchmarks I did with other tools here. I consider both Phigaro and VIBRANT to be very conservative (in a good way!), so you could use them to see what they tell you.

(etc) It seems to be impossible to compute the genome quality of giant virus using genomad and checkv. Do you have any idea how to comupute it?

What do you mean by that? geNomad can predict giant viruses, but it is not meant to evaluate the genome quality. You can use CheckV can estimate the quality.