ASMC and FastSMC

This repository provides ASMC and its extension FastSMC, implemented in C++ with Python bindings. Prebuilt CPython wheels are available for Linux (compatible with glibc ≥ 2.28) and macOS (built on macOS 15 for x86_64 and macOS 14 for arm64).

Platform \ CPython	≤3.8	3.9	3.10	3.11	3.12	3.13	3.14
Linux x86_64	❌	✅	✅	✅	✅	✅	✅
Linux aarch64	❌	✅	✅	✅	✅	✅	✅
macOS Intel (x86_64)	❌	✅	✅	✅	✅	✅	✅
macOS Apple Silicon (arm64)	❌	✅	✅	✅	✅	✅	✅

Quickstart

Install the Python module from PyPI

Most functionality is available through a Python module which can be installed with:

pip install asmc-asmc

Documentation

The following pages of documentation contains specific information:

• Quickstart guide for users
• ASMC python docs
• FastSMC python docs

This Python module is currently available on Linux and macOS.

Example Jupyter notebooks showcasing basic functionality can be found here:

• Example notebooks

License

ASMC and FastSMC are distributed under the GNU General Public License v3.0 (GPLv3). For any questions or comments on ASMC, please contact Pier Palamara using <lastname>@stats.ox.ac.uk.

Reference

If you use this software, please cite the appropriate reference(s) below.

The ASMC algorithm and software were developed in

• P. Palamara, J. Terhorst, Y. Song, A. Price. High-throughput inference of pairwise coalescence times identifies signals of selection and enriched disease heritability. Nature Genetics, 2018.

The FastSMC algorithm and software were developed in

• J. Nait Saada, G. Kalantzis, D. Shyr, F. Cooper, M. Robinson, A. Gusev, P. F. Palamara. Identity-by-descent detection across 487,409 British samples reveals fine-scale evolutionary history and trait associations. Nature Communications, 2020.

ASMC Release Notes

v1.4.0 (2025-10-22)

Breaking changes

• Dependencies are no longer managed with vcpkg. Boost and zlib should now be obtained from a system package manager, while other dependencies are fetched using CMake's FetchContent during configuration.

Other changes

• Added support for cross-platform SIMD vectorization using the Google Highway library. This improves performance by dynamically dispatching to the most powerful supported instruction set at runtime (e.g., AVX-512, AVX2, or NEON), ensuring optimal performance on different CPU architectures.

• Python wheels are now available for Linux and macOS on both x86_64 and arm64/AArch64 architectures, for CPython versions 3.9 to 3.14 inclusive.

v1.3.1 (2023-06-30)

Breaking changes

None

Other changes

• The location of a .map or .map.gz file can now be optionally specified explicitly: previously it was assumed to be at the inFileRoot.

v1.3 (2023-03-03)

Breaking changes

None

Other changes

• Decoding a batch can now be done in a selected subregion with from / to indices. A cm_burn_in parameter takes into account additional variants on either side of the subregion for HMM burn-in.
• Allow the user to access selected attributes of the DecodingParams and Data from the ASMC object.
• Python continuous integration now uses Python 3.8 and 3.11 (previously 3.6 and 3.9)
• Update Catch to v2.13.

v1.2 (2021-09-28)

All functionality for ASMC and FastSMC is now in this repository (link).

Breaking changes

• Fixed an issue with demographic models. The CEU.demo demographic model and the decoding quantities for CEU+UKBB previously provided in the repository were mistakenly encoded as diploid rather than haploid. CEU.demo and CEU+UKBB decoding quantities have now been updated and can be found in this repository. Also see the manual for a note on how this affects analyses.

Other changes

• New API for decoding pairs with ASMC. In addition to running full analyses as described in the ASMC paper, users can now decode specific pairs and get back a variety of summary statistics. See the ASMC python documentation for details.
• New, more extensive, documentation is available.

v1.1 (2021-01-20)

Legacy repository

Improvements to documentation and default use. No changes to any core functionality.

Breaking changes

• The hashing functionality, previously named GERMLINE, has been renamed to hashing. This includes the command line flag for turning this behaviour on/off, which is now --hashing.

Other changes

• --hashing is now ON by default when running the FastSMC executable: previously, --GERMLINE was OFF by default.
• Extra output, including the IBD segment length, posterior mean, and MAP, are now on by default. This behaviour can be toggled with the flags --segmentLength, --perPairPosteriorMeans, --perPairMAP.
• An example script has been added to cpp_example/FastSMC_example_multiple_jobs.sh that demonstrates how to run FastSMC with multiple jobs simultaneously.
• The README has been updated to focus on FastSMC functionality.
• More robust checking is now used to verify the decoding quantities file is correct before reading it.
• CMake will now, by default, build in Release mode (giving 03 optimisation on Linux). Previously, Debug was used by default.

v1.0 (2020-09-18)

Legacy repository

First public release of FastSMC, with functionality as described and used in this paper.