Sulfonamide to it's own "drug_class"?

[evolarjun]

Hi,

I am wondering how the "drug_classes" are decided upon. I just had a case where I wanted to check the phenotypic effect of a dfr gene (trimethoprim resistance), however most of the AST test results we have are in the context of cotrimoxazole. So I wanted to take a look at 'trimethoprim-sulfamethoxazole' resistance and combinations of dfr, with sul genes and folP mutations.

It looks like "SULFONAMIDE" is in the "Other antibacterials" drug class (according to https://github.com/AMRverse/AMRgen/blob/main/data-raw/amrfp_drug_classes_agents.tsv, which is where I think they're defined). That makes the combination graphs a bit busy with non-relevant stuff and hard to read (trimethoprim-sulfamethoxazole.E.coli_and_Shigella.Trimethoprims.Other antibacterials.pdf). I'm not casting aspersions here, drug class definition decisions are definitely judgement calls.

I suggest including or pointing to the drug_class list from the documentation since, like our classes and subclasses, they're sometimes not-obvious. To get what I want, I think I should be able to just filter the input data to only include the markers I'm interested in, but I'm wondering what the logic is / process is for defining drug_class.

And thanks again for a great tool. I wouldn't be writing if I wasn't very happy. The visualizations are far better than any I've come up with for similar problems, and I've never spent the time to figure out how to re-call SIR values from the MICs. It's really really nice.

Arjun

[katholt]

Hi Arjun, thanks for posting this issue, and for the kind comments.
The mapping of drugs/classes is something we put on the backburner for a bit because we are trying to come up with a sustainable solution to harmonise across AMRfp/refgene, CARD, and AMR package that suits the needs of AMRgen package and AMRrules! In short, the goal is to have drug/drug class fields readable by AMR package as 'ab' class, so in AMRgen functions you could specify a drug or class once (using AMR package terms) and extract both geno & pheno data... but this required some work.
It's time we sorted it out properly!
The AMR package has been updated, now we just need to update AMRgen and the import function to utilise this properly.
Will let you know when it's ready, hopefully next week.

[evolarjun]

Hi Kat,

Filtering the AMRFinderPlus results for only genes and point mutations of interest before running the analysis worked great and I got the helpful graphs I was looking for, so there's a pretty easy workaround.

Thanks again for the great tool, and please do not hurry to change things for my use cases, especially since I have an easy fix for my issue.

Arjun