Skip to content

FAQ

General

What is BASALT?

BASALT (Binning Across a Series of Assemblies Toolkit) is a tool for binning and post-binning refinement of metagenomic assemblies. It generates high-quality metagenome-assembled genomes (MAGs) from short-read, long-read, and hybrid datasets.

What makes BASALT different from metaWRAP or DASTool?

Three main differences:

  1. Multi-assembly support: BASALT accepts multiple single assemblies and co-assemblies in a single run with automatic dereplication.
  2. Deep learning refinement: Neural network-based contamination detection and removal at the individual contig level.
  3. Long-read integration: Efficient utilisation of long reads for contig retrieval and polishing.

What data types does BASALT support?

Mode Short Reads Long Reads (ONT/PacBio) PacBio HiFi
SRS only — —
SRS + LRS —
SRS + HiFi —
HiFi only — —

Long-read-only (ONT/CLR) without short reads is not yet supported.

Performance

How long does BASALT take to run?

This depends on dataset size, complexity, and chosen parameters. A typical run with --sensitive sensitive on a 32-core workstation with 256 GB RAM may take 12–24 hours for a moderate-complexity metagenome (e.g., human gut). The demo dataset completes in ~6 hours on a 32-core machine.

How can I speed up BASALT?

  • Use --sensitive quick --refinepara quick for the fastest preset
  • Run only the module you need: --module autobinning, --module refinement, or --module reassembly
  • Increase thread count (-t)
  • Ensure sufficient RAM to avoid swap overhead
  • For multiple assemblies, BASALT is already more efficient than running multiple single-assembly jobs

Does BASALT support GPU?

Yes. BASALT v1.2.0 supports GPU acceleration for Semibin2 and deep learning model inference. Ensure CUDA-compatible PyTorch is installed.

Workflow

Can I run only part of the pipeline?

Yes. Use --module:

  • --module autobinning: Run only binning and bin selection
  • --module refinement: Run only contamination removal and contig retrieval (on existing bins)
  • --module reassembly: Run only reassembly (on existing refined bins)
  • --module all: Run the full pipeline (default)

Can I use my own bins with BASALT refinement?

Yes. There are two approaches:

  1. Standalone refinement (-r): Point BASALT to a single binset folder.
  2. Data Feeding (-d): Import multiple external binsets with automatic coverage recalculation.

Can the same contig appear in multiple bins?

Under SA + CA mode, redundant bins can be generated. BASALT's Bin Selection module identifies and removes redundancies automatically through dereplication. The final best binset is non-redundant.

What if my run is interrupted?

BASALT records progress in Basalt_checkpoint.txt. Resume with:

BASALT --mode continue

For a completely fresh restart:

BASALT --mode new

Input & Output

Does BASALT support absolute file paths?

No. Place all input files in the current working directory, or create soft links using ln -s.

Can I specify an output directory?

BASALT generates all output in the current working directory. The -o flag sets the output folder name prefix, not the path.

Where are the final MAGs?

In <output_prefix>_final_binset/ (default: Final_binset_final_binset/). Each bin is a single multi-FASTA file.

Troubleshooting

CheckM2 database not found

checkm2 database --download

Refer to CheckM2 documentation for details.

BASALT: command not found

Ensure the conda environment is activated and scripts have correct permissions:

conda activate basalt_env
chmod -R 755 $(dirname $(which BASALT))/*

IndexError: list index out of range

This typically means BASALT cannot parse file paths. Ensure all input files are in the working directory (no absolute paths) and the -s argument uses the correct delimiter format (/ between samples, , between read pairs).

quality_report.tsv not found

This occurs when too few bins pass quality thresholds, so CheckM2 produces no output. Try:

  • Lowering --min-cpn (e.g., --min-cpn 20)
  • Using --sensitive more-sensitive for more thorough binning
  • Ensuring sufficient sequencing coverage

Installation

What's the difference between BASALT and BASALT-Air?

BASALT-Air (v1.0.0, 2026) is a new lightweight edition that replaces Conda with Pixi for dependency management. Key improvements include absolute path support, --workdir/--outdir separation, and built-in dependency checks. The core pipeline is identical — both versions produce the same results. See the home page for a comparison table.

Can I install BASALT without Conda?

Yes. BASALT-Air uses Pixi instead of Conda. Or use the Singularity image for either version.

How do I install on macOS?

BASALT is designed for Linux x64 systems. macOS is not officially supported. Use the Singularity image in a Linux VM or Docker container on macOS.