taxvamb

name	taxvamb
description	TaxVAMB — semisupervised bimodal VAE binning that integrates TNF, coabundance, and contig taxonomy for improved MAG recovery (incl. incomplete genomes and long-read sets). Use when: (1) Taxonomy-informed binning, (2) Comparing to VAMB/AAMB/SemiBin2, (3) Single-sample setups needing more HQ bins. Upstream: https://github.com/RasmussenLab/vamb. Taxonomy labels are an input feature — garbage taxonomy in, garbage bins out.
license	MIT
category	analysis-tools
tags	["binning","TaxVAMB","VAMB","taxonomy","deep-learning"]
upstream	https://github.com/RasmussenLab/vamb
stage	binning

TaxVAMB

Upstream: RasmussenLab/vamb

Citation

Kutuzova, S. et al. Improving metagenome binning by integrating intrinsic features and taxonomy. Nat. Biotechnol. (2026). https://doi.org/10.1038/s41587-026-03098-0

See also docs/references.md.

Analytical thinking

TaxVAMB adds contig-level taxonomy into a VAMB-style VAE. Strong reported gains on CAMI2 human sets and long-read data, and on incomplete genomes — but performance inherits taxonomy DB bias. Prefer high-quality taxonomic inputs; do not circularly claim novelty from taxonomy-informed bins without independent QC (checkm2, gunc, gtdbtk).

How to run

# taxonomy table + contigs + abundances — follow TaxVAMB section of VAMB docs

Related skills

vamb · aamb · metabuli · mmseqs2 · basalt · checkm2 · tool-selection