adamixture

Fast Biobank-Scale Population Genetics Clustering

---

ADAMIXTURE is a fast CPU/GPU implementation of ADMIXTURE for biobank-scale genetic clustering. .P and .Q outputs remain compatible with ADMIXTURE.

System requirements

Hardware & Platform compatibility

ADAMIXTURE runs cross-platform on Linux, macOS, and Windows, supporting CPU computation as well as GPU acceleration on NVIDIA GPUs (CUDA) and Apple Silicon (MPS).

Software requirements

We recommend creating a fresh Python 3.10+ virtual environment. For a faster installation experience, we highly recommend using uv.

[!IMPORTANT]
If you plan to use GPU acceleration, ensure that the CUDA toolkit is correctly loaded (e.g., module load cuda) before starting the installation. This ensures that the dependencies and internal components are correctly configured for your hardware.

As an example, using uv (recommended):

$ uv venv --python 3.10
$ source .venv/bin/activate
$ uv pip install adamixture

Installation Guide

The package can be installed using pip:

Basic Installation (CPU default)

Uses the pre-compiled C/Cython wheel and does not install PyTorch:

$ pip install adamixture

GPU Installation (NVIDIA CUDA / Apple Silicon MPS)

Includes PyTorch and Ninja for hardware-accelerated execution:

$ pip install "adamixture[gpu]"

The commands above use the binary wheel when one is available for the Python version and platform. Installing directly from a source checkout (for example, with pip install -e .) requires a local C/Cython build toolchain.

Running ADAMIXTURE

To train a model, simply invoke the following commands from the root directory of the project. For more info about all the arguments, please run adamixture --help. Note that BED, VCF, PGEN and BCF are supported.

Supported input files include:

• PLINK BED: .bed, .bed.gz, .bed.zst, with .bim/.fam sidecars that may also be plain, .gz or .zst.
• PLINK PGEN: .pgen or .pgen.zst, with .pvar/.psam sidecars that may be plain, .gz or .zst.
• VCF: .vcf, .vcf.gz and .vcf.zst.
• BCF: .bcf, .bcf.gz and .bcf.zst.

As an example, the following ADMIXTURE call

$ ./admixture snps_data.bed 8 -s 42

would be equivalent in ADAMIXTURE by running

$ adamixture -k 8 --data_path snps_data.bed --save_dir SAVE_PATH --name snps_data -s 42

By default, the following files will be output to the SAVE_PATH directory (the name parameter will be used to create the full filenames):

• A .P file, similar to ADMIXTURE.
• A .Q file, similar to ADMIXTURE.
• A .png plot file containing the visualization of the inferred ancestry proportions (Q matrix).

Logs are printed to the stdout channel by default. If you want to save them to a file, you can use the command tee along with a pipe:

$ adamixture -k 8 ... | tee run.log

Running with multi-threading

To run ADAMIXTURE using multiple CPU threads, use the -t flag:

$ adamixture -k 8 --data_path data.bed --save_dir out/ --name test -t 8

Running with GPU acceleration

To leverage GPU acceleration (highly recommended for large datasets), use the --device flag:

• NVIDIA GPU (CUDA):

  $ adamixture -k 8 --data_path data.bed --save_dir out/ --name test --device gpu
  

• macOS Apple Silicon (MPS):

  $ adamixture -k 8 --data_path data.bed --save_dir out/ --name test --device mps
  

[!TIP] GPU Acceleration: Using GPUs greatly speeds up processing and is highly recommended for large datasets. You can specify the hardware to use with the --device parameter:

• For NVIDIA GPUs, use --device gpu (requires CUDA).
• For macOS users with Apple Silicon (M1/M2/M3/M4/M5), use --device mps to enable Metal Performance Shaders (MPS) acceleration.
• Note that biobank-scale datasets are best handled on dedicated CUDA-capable GPUs due to high RAM requirements.

For a step-by-step guide on how to run a complete analysis workflow, see the 1000 Genomes Tutorial.

Multi-K Sweep

Instead of running ADAMIXTURE for a single K, you can automatically sweep over a range of K values using --min_k and --max_k. The data is loaded once, and each K is trained sequentially. K=1 is the no-structure null model and is a valid start for a sweep or a CV baseline:

$ adamixture --min_k 1 --max_k 10 --data_path snps_data.bed --save_dir SAVE_PATH --name snps_sweep

Cross-validation

Use --cv to estimate the optimal K by masking a fraction of genotype entries and measuring prediction error. → Full documentation

$ adamixture -k 8 --cv --data_path data.bed --save_dir out/ --name test

Plotting

By default, ADAMIXTURE automatically generates a png plot at 300 DPI without needing any additional flags. Following the method of Clumppling, it automatically aligns clusters across runs and K values. → Full documentation

Plots can include hierarchical population labels if you provide the arguments (--labels, --labels2, --labels3).

If you want to customize the format and resolution (e.g., to generate a PDF), use --plot (or --plot_single for individual per-K plots in multi-K sweeps):

• Single K runs (-k): Use --plot.

  $ adamixture -k 8 --data_path data.bed --save_dir out/ --name test --plot pdf 300
  

• Multi-K sweeps (--min_k and --max_k): Use --plot to configure the combined sweep plot (or --plot_single for individual plots per K).

  $ adamixture --min_k 2 --max_k 10 --data_path data.bed --save_dir out/ --name test --plot pdf 300
  

Projection Mode

Estimate ancestry proportions for new samples using a pre-trained, fixed P matrix (Q-only optimisation). K is detected automatically from P. → Full documentation

$ adamixture-project \
    --data_path new_samples.bed \
    --p_path trained_model/results.8.P \
    --save_dir projection_out/ \
    --name projected

Supervised Mode

Anchor the model with known population labels for a subset of samples while estimating Q freely for unlabeled ones. Labels use the same format as --labels (population name or -). → Full documentation

$ adamixture-supervised \
    --data_path all_samples.bed \
    --labels labels.txt \
    --save_dir supervised_out/ \
    --name supervised_run \
    -k 8

Other options

All hyperparameters and flags can be explored with:

$ adamixture --help

Key arguments:

Algorithm note

The ADAMIXTURE paper introduced Adam-EM as an adaptive first-order optimizer for admixture inference. The package still includes this solver via --algorithm adamem.

In the current implementation, the default is --algorithm brqn. Empirical benchmarking showed that block relaxation with ZAL quasi-Newton acceleration, when paired with our improved SVD+ALS initialization, reaches high-quality solutions in fewer iterations and better wall-clock time. For that reason, BR-QN is the default solver, while Adam-EM remains available for experimentation and reproducibility. Adam-EM tuning parameters are documented in Troubleshooting and Tips.

Troubleshooting and Tips

Common issues and platform notes:

• macOS: Install libomp via Homebrew (brew install libomp) before compiling.
• Windows: Pre-compiled wheels are provided on PyPI (pip install adamixture). If compiling from source (pip install -e .), install Build Tools for Visual Studio with Desktop development with C++.
• CUDA: Set CUDA_HOME or install nvcc via Conda (conda install -c nvidia nvcc).

→ Full documentation

License

This project is licensed under the BSD 3-Clause License - see the LICENSE file for details.

Cite

When using this software, please cite the following paper:

@article{saurina2026adamixture,
  title={ADAMIXTURE: adaptive first-order optimization for biobank-scale genetic clustering},
  author={Saurina-i-Ricos, Joan and Mas Montserrat, Daniel and Ioannidis, Alexander G.},
  journal={Bioinformatics},
  volume={42},
  number={Supplement\_1},
  pages={btag236},
  year={2026},
  doi={10.1093/bioinformatics/btag236},
  url={https://doi.org/10.1093/bioinformatics/btag236}
}