Enable protein query with tblastn about abricate

[tseemann]

It's a good question @crarlus - tblastn could just be a drop in replacement, but there is lots of business logic that assumes DNA coordinates etc.

What database do you use that is protein only?

[crarlus]

My collaborator has a curated list of proteins collected from various resources. Of course I tried (hard) to map them back to the original gene sequences, e.g. via a blast to uniprot and uniparc databases. However I could recover only some but not all of the sequences. So it might be an odd case but at the same time the data reality we live in.

[tseemann]

I have started adding tblastn support but it is extremely slow due to the way I am using the genome as they query... it's not committed yet.

[felipelira]

pipe it with prokka and you can use blastp after prediction. I mean, a brand new version,
ABRICATE+ (including aminoacids databases)

[tseemann]

Prokka relies on Prodigal to detect genes/ORFs, and often misses broken genes, or false frameshifted genes due to bad homopolymer issues with 454/ION/Pacbio/Minion assemblies.

[thsyd]

Hi, thank you so much for sharing ABRicate, @tseemann.
I just want to second the suggestion to improve the protein database option.
I am currently using abricate with a protein database of Pfam families (ca. 13000 protein sequences) to screen putative plasmids for replication protein-sequences. It does take a very long time :)
I'm sure you are aware of e.g. Diamond https://github.com/bbuchfink/diamond, that should work much faster than blast.

[tseemann]

@thysd Diamond only provides blastp (Prot:Prot) and blastx (Prot:DNA)
Unfortunately the design of abricate needs the query to be the contigs, i need tblastn (DNA:Prot)

I don't think Abricate is the best tool for what you want to do. Just running BLAST or MMSeqs2 or DIAMOND directly would make more sense.