BusyBee

BusyBee performs automated deconvolution of metagenomic sequences into population-level genomic bins to enable culture-independent recovery and analysis of individual or closely related microbial genomes.

Key Features:

• Reference-Independent Binning: Performs binning without requiring pre-characterized reference genomes.
• Data Compatibility: Supports assembled contigs from Illumina and long reads from Pacific Biosciences and Oxford Nanopore Technologies.
• Automated Deconvolution Process: Employs a reversible compression step combined with bootstrapped supervised binning for rapid and accurate separation of genomic bins.
• Visualization: Produces interactive 2D scatterplots for inspection of binning results.
• Quality and Annotation Metrics: Reports quality estimates per bin, provides taxonomic annotations, and identifies antibiotic resistance genes.
• Performance Benchmarking: Uses ground truth-based benchmarks reporting median F1 scores ranging from 70.02% to 95.21%, with strong performance on long-read data.
• Real-World Validation: Validated on real-world metagenomic datasets.
• Input Format: Accepts FASTA-formatted input data.

Scientific Applications:

• Metagenome binning and genome recovery: Deconvolves mixed-community sequencing data to recover population-level genomes from individual or closely related microorganisms.
• Taxonomic profiling and annotation: Produces taxonomic assignments for recovered bins to support microbial community characterization.
• Antibiotic resistance surveillance: Identifies antibiotic resistance genes within genomic bins for resistance monitoring.
• Cross-technology analysis and benchmarking: Enables analysis and performance evaluation across Illumina, Pacific Biosciences, and Oxford Nanopore Technologies datasets using ground-truth benchmarks.

Methodology:

Implements reference-independent binning using a reversible compression step combined with bootstrapped supervised binning, generates interactive 2D scatterplots, provides per-bin quality estimates, taxonomic annotations, antibiotic resistance gene identification, and reports ground-truth-based benchmarking metrics (median F1 scores 70.02%–95.21%).