Investigators introduced ARG-PASS v2.0 to expand computational detection of divergent antibiotic resistance genes. The study team targeted a lake metagenome from Kolkata, India as a tractable environmental sample in which transfer routes into humans might be observed. The authors reported that they selected 11 candidate ARGs from that metagenome for functional testing after prediction by the pipeline. Ten of those eleven candidates were functionally confirmed following expression in Escherichia coli, according to the source. This high confirmation rate, the authors argued, indicates the pipeline can recognise functional resistance determinants that share limited similarity with characterised enzymes.
The study team expressed predicted genes in Escherichia coli and assayed function. Ten of 11 candidates across diverse ARG classes showed activity in that heterologous host, the authors reported. Short, direct tests yielded functional readouts. The source did not provide step-by-step experimental protocols or detailed assay parameters for every validation in the text of the abstract; those methodological specifics are not reported there. The authors used these experimental confirmations as a primary accuracy check for ARG-PASS v2.0 before scaling predictions against large public repositories.
After validating the approach, the team expanded surveillance to public sequence databases. Data indicated 918 predicted class B1 metallo-beta-lactamases (MBLs) were recovered from MGnify and PLSDB. Most of these predicted MBLs diverged substantially from characterised enzymes, the authors reported, yet many were widely distributed across ecological samples. The source links the predictions to broad environmental sampling rather than exclusively clinical isolates, implying underappreciated diversity in nonclinical reservoirs.
Phylogenetic inference placed clinically relevant class B1 MBL sequences within diversity sampled from marine sources, the authors noted. The study team suggested that a marine reservoir may contribute to the pool of MBL diversity that eventually includes clinically relevant variants. Sequences with apparent clinical relevance co-occur on phylogenetic branches alongside marine-derived diversity, according to the reported analyses.
The authors cross-referenced predictions with the Unified Human Gastrointestinal Protein catalogue and found specific matches. A carbapenemase, an ANT (aminoglycoside nucleotidyltransferase), and class A beta-lactamases traced to Bacillus sonorensis and Bacillus licheniformis were reported. Those Bacillus-derived ARGs were detected in South Korean fermented soybean paste (doenjang) as well as in gut isolates from populations that consume those foods. The data indicate a plausible dietary exposure route by which antibiotic resistance genes reach the human gut via food-associated bacterial hosts.
The authors interpret the combined findings as evidence for a largely chromosomal reservoir of undescribed ARG diversity across environments. That reservoir, they reported, can reach the human gut through both resident hosts and generalist hosts that span habitats. Once such ARGs enter the human microbiome, the source states they could compromise treatment directly if an ARG-host is an opportunistic pathogen or indirectly after transfer to a clinical pathogen.
The authors framed the problem by noting that surveillance historically focused on ARGs resembling already characterised clinical genes, leaving the environmental reservoir poorly described and routes into human microbiomes inferred rather than directly observed. The source emphasises that describing that reservoir requires computational methods that balance detection of divergent sequences with experimental accuracy; ARG-PASS v2.0 was presented in that context. The article is a preprint and has not been peer reviewed, as stated in the source. The abstract does not provide full experimental protocols, detailed prevalence metrics across populations, or clinical outcome data tied to the newly predicted genes. It also does not specify regulatory follow-up, timelines for broader validation, or whether any of the newly predicted ARGs have been observed to mobilise on plasmids in clinical settings. These unanswered questions represent clear next steps for research reported by the authors.