Antimicrobial Resistance¶
BugSeq predicts antimicrobial resistance (AMR) for each organism in a sample, whether the sample is WGS from a bacterial isolate or metagenomic sequencing direct from specimen. BugSeq’s AMR prediction is applied for all bacteria and some viruses. See the Supported Analyses page for up to date details on which organisms are supported.
AMR Prediction Requests
Don’t see your organism? Get in touch.
Bacterial Prediction of Resistance¶
BugSeq uses a curated database of genes, mutations and interactions thereof to predict phenotypic AMR. Prediction is based on the scientific literature and AI/machine learning applied to large datasets. We attempt to predict the phenotype of drugs with a CLSI breakpoint, although not all bacteria will have an established “phenotype panel”.
Note
Phenotypic AMR prediction is undergoing rapid development and validation is recommended.
Confidence of Reporting¶
Bacteria¶
Predicted Non-Susceptibility¶
When a gene sequence is used in the prediction of AMR, BugSeq leverages the coverage and percent identity of that gene to score a prediction confidence.
| Confidence | Percent Identity to Reference Sequence | Coverage of Reference Sequence |
|---|---|---|
| Very High | 100% | 100% |
| High | 100% | < 100% |
| Moderate | < 100% | < 100% |
When multiple genes are present or a multifactor model is used to predict resistance, BugSeq integrates the confidence of each independent factor, along with its importance in predicting phenotype, to ascertain an overall confidence.
Predicted Susceptibility¶
The confidence of predicted susceptibility for a certain drug is based on the completeness of the genome recovered. If a genome is only partially recovered, it is possible that predictors of AMR were missed. These metrics follow the MIMAG standards.
| Confidence | Genome Completeness |
|---|---|
| High | > 90% |
| Moderate | 50% to 90% |
| Low | < 50% |
Future versions of BugSeq will improve confidence prediction to factor in what regions of the genome were recovered and the presence of plasmids in the sample.
Bacterial Exceptions¶
Mycobacterium tuberculosis complex¶
Predicted Non-Susceptibility¶
Confidence for non-susceptibility is based on database annotation of a specific variant, unlike typical bacterial prediction described above. For variants deriving from the WHO database, this mirrors the WHO confidence. Further detail on the WHO mutations and associated confidence is available in the WHO Catalogue of mutations in Mycobacterium tuberculosis complex and their association with drug resistance. For variants not found in the WHO database, BugSeq attempts to derive confidence following WHO criteria.
Predicted Susceptibility¶
See bacterial confidence for predicted susceptibility.