AMR Surveillance with WGS: From Reads to Resistance Profiles

Why AMR surveillance needs genomics

The most comprehensive estimate of the global burden found 4.95 million deaths associated with bacterial antimicrobial resistance in 2019, of which 1.27 million were directly attributable to it ^[1]. Six pathogens led the deaths associated with resistance: Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii and Pseudomonas aeruginosa ^[1]. The same study highlighted serious data gaps in low-resource settings and called for more microbiology laboratory capacity.

Phenotypic susceptibility testing tells you whether an isolate grows in the presence of a drug. Whole genome sequencing (WGS) adds why: which gene or mutation is responsible, whether it sits on a mobile plasmid, and whether the same resistance element appears in isolates from different hosts, farms, foods or hospitals. That context is what turns individual results into surveillance.

What a robust pipeline looks like

Several curated databases are in common use. ResFinder detects acquired genes, and its companion PointFinder detects chromosomal mutations; version 4.0 also translates genotypes into a predicted antibiogram ^[2]. The Comprehensive Antibiotic Resistance Database (CARD) organises resistance determinants into an ontology and supports resistome prediction ^[3]. NCBI's AMRFinderPlus and its Reference Gene Catalog also cover stress-response and virulence genes ^[4]. PlasmidFinder identifies plasmid replicons, which helps show whether a resistance gene is likely to spread horizontally ^[5].

How well does genotype predict phenotype?

For well-studied pathogens, very well. ResFinder 4.0 was validated on more than 2,200 isolates of E. coli, Salmonella, Campylobacter jejuni, Enterococcus faecium, E. faecalis and S. aureus. Genotype–phenotype concordance was 95% or higher for 46 of 51 drug–species combinations in Gram-negative bacteria and 25 of 32 in Gram-positive bacteria ^[2]. Where agreement was lower, the discrepancies were mainly linked to phenotypic test interpretation and sequence quality rather than to the tool ^[2].

The EUCAST subcommittee on WGS is more cautious about clinical use. Its review concluded that for most bacterial species there is not yet enough evidence to guide treatment decisions with WGS-predicted susceptibility. It recommended epidemiological cut-off values (ECOFFs) as the main comparator, agreed quality-control metrics, and a single curated public database of resistance loci ^[6]. For research and surveillance, which is where BacSeq Analytics works, WGS is already a powerful tool.

Pitfalls to watch for

• Gene presence is not expression. A detected gene may be silent, truncated or poorly expressed, so the isolate can still test susceptible.
• Absence is not proof of susceptibility. Novel genes, efflux pump regulation and unknown mutations are missed by databases built from known determinants.
• Database and threshold choices change results. Identity and coverage cut-offs and the database version should always be reported alongside the gene list.
• Fragmented assemblies can split genes across contigs, which lowers coverage and can hide a hit. They also make it harder to tell whether a gene sits on a plasmid or the chromosome.
• Low sequencing depth or contamination produces both false positives and false negatives, which is why QC comes first.

What an AMR section of a WGS report should contain

A useful AMR report goes beyond a list of gene names. In the BacSeq Analytics Standard and Premium packages, the AMR section lists each detected gene or mutation with its drug class, percentage identity and coverage, and the contig it was found on. It flags genes on plasmid-associated contigs and summarises the predicted resistance profile in plain language. It also states the database and thresholds used, so the results can be reproduced and compared with other studies.

BacSeq Analytics reports are for research use only. They are not intended for clinical diagnosis or to guide patient treatment.