momi3

momi (short for MOran Models for Inference) is a Python package for computing the expected sample frequency spectrum (SFS)—a key summary statistic in population genetics—and using it to infer demographic history.

This third version is a complete rewrite of the original momi and momi2 packages. It introduces several major improvements, including greater flexibility in model specification and enhanced performance and scalability.

Additionally, momi3 is now being built as a core component of the broader demestats package, providing a more comprehensive framework for demographic inference.

To navigate the momi3 documentation, please refer to the Notation section first to understand momi3's representation of parameters within a demographic model.

Given any demes formatted demographic model:

• The Tutorial goes over all of the core functions of momi3, including how to output and modify model constraints, compute the expected SFS, and compute the likelihood and its gradient.
• The Random Projection section teaches users how to use random projections to compute an approximation of the full expected SFS and discusses its benefits.
• The Optimization section demonstrates how to construct custom inference pipelines using scipy.minimize, highlighting momi3's modular design where each component — from objective functions to model constraints — can be tailored to specific research requirements. This flexibility enables researchers to implement specialized optimization strategies.

momi3 is described in the following preprint:

Dilber, E., & Terhorst, J. (2024, March 29). Faster inference of complex demographic models from large allele frequency spectra [Preprint]. bioRxiv. https://doi.org/10.1101/2024.03.26.586844