Delphy is a fast, scalable, accurate and accessible tool for Bayesian phylogenetics based on Explicit Mutation-Annotated Trees (EMATs). EMATs are an extension of the Mutation-Annotated Trees (MATs) introduced by UShER where nodes have explicit times, mutations are represented as explicit timed events along branches, and missing data is explicitly represented. EMATs open substantial simplifications and scaling opportunities in the calculations powering Bayesian phylogenetics, at the cost of some statistical efficiency with respect to the traditional calculations based on Felsenstein pruning. For genomic epidemiology datasets, where the total number of mutations on a tree is comparable to the number of samples, this is a very favorable trade-off.

These sources comprise the "core" computational engine of Delphy. The web application that allows users to immediately and intuitively use Delphy (currently at https://delphy.fathom.info) is developed in collaboration with Fathom Information Design, and its separately licensed sources are hosted here.

• Whitepaper with overview of key ideas and accuracy+speed benchmarks

• TO COME: Preprint with full details and benchmarking (expected: late July 2024)

Delphy can be compiled either as a native, standalone command-line program (delphy and its graphical cousin, delphy_ui), or as a WebAssembly bundle that lies at the core of delphy-web. Build instructions for both can be found in INSTALL.md. Delphy does not require special hardware, such as a GPU.

Delphy was developed and is primarily tested under Linux (Ubuntu 22.04.4 LTS, x86-64). Delphy-web, deployed at https://delphy.fathom.info, should work in any modern web browser; see the Delphy-web README.md for specific browsers and operating systems on which it has been tested.

Delphy is developed in the Sabeti Lab at the Broad Institute.

Delphy draws a lot of inspiration from:

• UShER and matOptimize
• BEAST
• BEAST2
• MAPLE
• IQ-TREE2

Copyright (c) 2022-2024 Broad Institute, Inc. See LICENSE for details.