Skip to content

Release Notes

BASALT-Air v1.0.0 (2026-04-16)

BASALT-Air is a lightweight, modular reimplementation of BASALT using Pixi for dependency management.

Key Improvements

  • Absolute path support: No need to cd into data directories. Run BASALT from anywhere.
  • Separated work/output directories: --workdir for intermediate files, --outdir for final results.
  • Pixi-based installation: Faster, more reproducible than Conda. No environment activation headaches.
  • Hugging Face model hosting: Download weights via huggingface-cli or pixi run download-weights.
  • Built-in dependency check: BASALT --check-deps and BASALT --version.
  • Configurable CUDA version: Set in pixi.toml instead of manual CUDA toolkit management.
  • Semicolon separator: Multiple datasets separated by ; instead of / for cleaner CLI.

Breaking Changes from BASALT (Conda)

BASALT (Conda) BASALT-Air
Package manager Conda Pixi
Activation conda activate basalt_env pixi shell
Dataset separator / (slash) ; (semicolon)

BASALT-Air shares the same core pipeline and produces identical output. The change is in packaging and CLI ergonomics only.

v1.2.0 (2025-12-16)

Major Changes

  • Python 3.12 support: Upgraded from Python 3.8 to 3.12 with a streamlined installation script.
  • LorBin integration: Added LorBin (Nat. Commun., 2025) as an extra binner (-e l), the state-of-the-art long-read binning tool.
  • CheckM2 as default: Changed the default quality assessment from CheckM to CheckM2 v1.1.0.
  • GPU acceleration: Added GPU support for Semibin2 and deep learning model inference.
  • Flexible model weights: Model weights path is now configurable via BASALT_WEIGHT environment variable instead of being fixed to ~/.cache/BASALT.
  • VAMB 5.0 support: Updated VAMB integration for compatibility with VAMB 5.0 (Python 3.12).

Breaking Changes

  • Python 3.8 is no longer supported. Python 3.12 is required.
  • The default QC backend is now CheckM2 (previously CheckM). Use -q checkm to revert to CheckM.
  • --autopara argument has been renamed to --sensitive.

v1.1.0 (2024-06-12)

Changes

  • Improved installation workflow with BASALT_setup.py.
  • Added China mainland mirror support (BASALT_setup_China_mainland.py).
  • Added Singularity image support.
  • Various bug fixes and stability improvements.

v1.0.0 (2023-08-18)

Initial public release.

Features

  • Multi-assembly binning with four binners (MetaBAT2, Maxbin2, CONCOCT, Semibin2).
  • Deep-learning-based contamination detection and removal (S5 outlier remover).
  • Contig retrieval via paired-end tracking (S6) and long-read connections (S7).
  • Reassembly with SPAdes (S9) and Unicycler hybrid (S9p).
  • Data Feeding module for importing external bins.
  • CheckM-based quality assessment.
  • Checkpoint-based resumption.

Publication

Qiu, Z., Yuan, L., Lian, CA. et al. BASALT refines binning from metagenomic data and increases resolution of genome-resolved metagenomic analysis. Nature Communications 15, 2179 (2024).