Interpreting inconsistencies with VFDB annotations between Pharokka and Phold

[Terrachomp]

Thank you, George, for your contributions to making genome annotation faster, more effective and easier.

I've been using Phold for genome annotation in addition to Pharokka but encountered two circumstances where Pharokka v 1.7.0 predicted virulence factors with VFDB but Phold v0.2.0 did not predict virulence factors with VFDB.

In both circumstances (Vibrio phage CTXphi GCA_00893195.1 and Salmonella phage ST160 NC_014900.1) Phold matched with a protein "like" or similar to that which was detected by Pharokka.

There were also cases where proteins expected to be in VFDB were not, including enterotoxins ctxA and ctxB in GCA_00893195.1 both detected by Pharokka and Phold, as well as PHROG 201 (Panton-Valentine Leukocidin) classifying this toxin as DNA RNA nucleotide metabolism and not VFDB in AB045978.2

I'm keen to hear your thoughts on how this inconsistency between Pharokka and Phold (VFDB) can be interpreted, if it might be fixed or changed, and if Phold or Pharokka annotations are to be trusted more.

Files from Pharokka and Phold output with information about PHROGs and VFDB:

Files for George.zip

Thanks again.

[gbouras13]

Hi @Terrachomp - sorry for missing this. I'll have a closer look now and get back to you soon. My feeling is in the way phold handles the annotation outputs.

George