hAMRoaster

hAMRoaster benchmarks AMR gene detection tools by evaluating their accuracy in predicting antibiotic resistance phenotypes from shotgun metagenomic data.

Key Features:

• Benchmarking capability: Calculates sensitivity, specificity, precision, and accuracy to compare AMR detection software.
• Simulated data evaluation: Uses simulated datasets including highly resistant mock communities and low-resistance Illumina short-read metagenomic samples derived from strains with known phenotypes.
• Comprehensive tool assessment: Benchmarks nine open-source tools that are conda- or Docker-installable, actively maintained, and compatible with FASTA or FASTQ input files.
• Performance metrics across coverage: Compares detected AMR genes against susceptibility testing results across genome coverage levels (5x, 50x, 100x).
• Comparative analysis across drug classes: Simulates high- and low-resistance scenarios to assess variability in AMR gene identification across different drug classes.
• Data simulation with defined phenotypes: Generates synthetic communities with predefined taxa and phenotypes for controlled evaluation.

Scientific Applications:

• Tool selection for metagenomics: Guides selection of AMR detection software for research and clinical shotgun metagenomic studies.
• Public health and epidemiology: Informs metagenomic AMR surveillance and study design by comparing tool performance under varied resistance scenarios and coverage depths.

Methodology:

Generates synthetic communities with predefined taxa and phenotypes, simulates Illumina short-read datasets for high- and low-resistance scenarios, benchmarks nine conda/Docker-installable open-source tools compatible with FASTA/FASTQ by calculating sensitivity, specificity, precision, and accuracy, and compares detected AMR genes against susceptibility testing results across coverage levels (5x, 50x, 100x).