KARGVA

KARGVA detects and classifies antibiotic resistance genes (ARGs) and antibiotic resistance gene variants (ARGVs) from high-throughput sequencing data to enable characterization of resistance in metagenomic samples and cultured bacterial isolates.

Key Features:

• Three-way hash-based k-mer search: Links k-mers (strings of fixed length k) to ARGVs and point mutations to identify variants without pre-assembly or a posteriori mutation verification.
• Ad hoc ARGV database: Uses a large database of ARGVs derived from multiple sources to support variant identification.
• Statistical filtering: Applies a statistical filter on sequence classification to reduce type I and II errors under sequencing errors and genomic rearrangements.
• Supports diverse sequencing inputs: Processes high-throughput sequencing data from metagenomic samples and cultured bacterial isolates.
• Empirical performance: Demonstrated 99.2% accuracy with up to 1% base change rate, 86.6% accuracy at 5% base change rate, and 98.2% accuracy in the presence of genome rearrangements on semi-synthetic data.
• Benchmark comparisons: Identified more ARGVs than Resistance Gene Identifier and PointFinder by factors of 4.8× and 6.8× respectively while maintaining expected false positive rates.
• Large-scale testing: Evaluated on MetaSUB consortium data comprising over 3,700 metagenomic experiments worldwide.

Scientific Applications:

• ARG/ARGV detection in metagenomes: Identification and classification of ARGs and chromosomal ARGVs in environmental and urban microbiome metagenomic datasets.
• Analysis of cultured isolates: Detection of acquired and chromosomally encoded resistance variants in sequencing data from bacterial isolates.
• Surveillance and epidemiology: Large-scale screening for ARGV prevalence across global datasets such as MetaSUB.
• Method benchmarking: Comparative evaluation against tools like Resistance Gene Identifier and PointFinder using semi-synthetic and real-world datasets.

Methodology:

KARGVA performs a three-way hash-based k-mer search linking k-mers to ARGVs and point mutations, queries a large ad hoc ARGV database derived from multiple sources, and applies a statistical filter on sequence classification; empirical evaluation used semi-synthetic data and MetaSUB metagenomes.