donni (demographic inference from spectra using ML) paper published

[Gutenkunst, Ryan N - (rgutenk)]

Hello dadi users,

The paper describing donni, our new machine-learning approach to demographic history inference from frequency spectra, has been published in MBE: https://doi.org/10.1093/molbev/msae077 . The software is available at https://github.com/lntran26/donni .

If you’re working with standard demographic models, donni will download pre-trained neural networks for the inference, so it will be almost instantaneous. Results aren’t identical to dadi, because neural networks don’t maximize a likelihood, but they can be nevertheless informative. One option is to use the donni results as a starting point for dadi inference, which dadi-cli supports directly https://github.com/xin-huang/dadi-cli .

Let us know if you have questions or run into issues using donni!

Best,
Ryan

--
Ryan Gutenkunst
Professor and Department Head, Molecular and Cellular Biology, University of Arizona
http://gutengroup.arizona.edu, (520) 626-0569, LSS 444A