ONN4ARG

ONN4ARG predicts and annotates antibiotic resistance genes (ARGs) across diverse microbial communities using an ontology-aware annotation framework to identify known and novel ARGs.

Key Features:

• Ontology-Aware Layer: Incorporates a layer that enforces annotation predictions to align with established ontology rules and hierarchical structure of biological ontologies.
• Comprehensive ARG Discovery: Evaluates a dataset of 200 million candidate microbial genes from 815 diverse microbial community samples across various environments and hosts to discover ARGs beyond homology-based methods.
• Performance Metrics: Systematic evaluations report improved efficiency, accuracy, and comprehensiveness compared to existing methods such as DeepARG, identifying over 120,726 candidate ARGs with more than 20% novel relative to public databases.
• Environmental and Host-Specific Insights: Reveals environment-specific and host-specific patterns of ARG distribution across ecological niches.
• Validation of Novel ARGs: Supported validation through wet-experimental functional testing and structural analysis of docking sites, including identification and confirmation of a novel streptomycin resistance gene from oral microbiome samples.

Scientific Applications:

• Microbial genomics: Enables detection and annotation of ARGs in microbial community gene catalogs for genomics studies.
• Antibiotic resistance surveillance: Supports comprehensive ARG discovery and mapping to inform surveillance of antibiotic resistance genes globally.
• Evolution and ecology of ARGs: Facilitates studies of the evolution, spread, and ecological impact of ARGs across environments and hosts.
• Public health research: Contributes data for strategies addressing antibiotic-resistant infections and the distribution of ARGs in clinical and environmental contexts.

Methodology:

Computational methods explicitly include an ontology-aware annotation layer enforcing ontology hierarchy, large-scale evaluation of 200 million candidate genes from 815 samples, benchmarking against DeepARG, and structural analysis of docking sites.