A Note on Ongoing Revisions of the Antibiotic Resistance Ontology
The Comprehensive Antibiotic Resistance Database gratefully acknowledges recent funding from the Genome Canada & Canadian Institutes of Health Research's Bioinformatics & Computational Biology program, allowing integration of the Antibiotic Resistance Ontology (ARO) with the Genomic Epidemiology Ontology, IRIDA platform, and OBO Foundry (see Genome Canada press release). As such, the next two years will be a time of active development for the ARO. Expect significant changes in the ARO between monthly releases as well as occasional incomplete revisions, which may affect downstream analyses.
February 2017 Changes
Use of the part_of relationship now follows canonical usage and is restricted to association of sub-units with their large multi-unit protein complexes
Extensive revisions to the antimicrobial efflux branch of the ARO
Extensive revisions to the rRNA mutations branch of the ARO
New use of the participates_in and has_part relationships in place of formerly incorrect usage of the part_of relationship for association of resistance determinants with their mechanism of action.
April 2017 Changes
Extensive addition of confers_resistance_to_drug relationships for efflux complexes
Drug and mechanism category updates for the Resistance Gene Identifier
May 2017 Changes
Addition of bitscores to detection models, curation of chloramphenicol exporter proteins, ontology changes, JSON file format changes
August 2017 Changes
Removal of redundant intermediate terms relating resistance determinant to drug class, with improved overall classification by Drug Class and Resistance Mechanism
January 2018 Changes
Parallel classification system added to the ARO for organization of RGI results: Drug Class, Resistance Mechanism, AMR Gene Family
The protein homolog model is an AMR detection model. Protein homolog models detect a protein sequence based on its similarity to a curated reference sequence. A protein homolog model has only one parameter: a curated BLASTP bitscore cutoff for determining the strength of a match. Protein homolog model matches to reference sequences are categorized on three criteria: perfect, strict and loose. A perfect match is 100% identical to the reference sequence along its entire length; a strict match is not identical but the bitscore of the matched sequence is greater than the curated BLASTP bitscore cutoff. Loose matches are other sequences with a match bitscore less than the curated BLASTP bitscore.