This repository contains an implementation of Phylo2Vec which includes:

• cfg/: Example configuration files
• data/: Placeholder folder to contain sequence files in FASTA format.
• examples/: Example notebooks for different datasets
• hc/: Phylogenetic tree optimisation via hill-climbing optimisation
  • Branch length and nucleotide subsitution model optimisation relies on RAxML-NG
• tests/: Placeholder folder for unit tests
• trees/: Placeholder folder to contain tree files as Newick strings.
• utils/: Utility functions including definitions of Phylo2Vec and transforms from commonly used tree formats to Phylo2Vec (and vice versa).

A quick demo detailing hill-climbing optimisation with Phylo2Vec is available on the demo.ipynb notebook.

A more minimalistic demo with an updated defiition of Phylo2Vec is available on Colab:

To reproduce the environment, run:

conda env create -f env.yml

To run hill climbing-based optimisation using Phylo2Vec, run:

conda activate phylo

python -m hc.main

• Download a binary of RAxML-NG at: https://github.com/amkozlov/raxml-ng. For Windows, consider using the Windows Subsystem for Linux.

The following datasets were used:

• primates: https://evolution.gs.washington.edu/book/datasets.html
• fluA: https://github.com/4ment/phylostan/tree/master/examples
• M501: DS2 dataset in https://github.com/zcrabbit/vbpi-gnn/tree/main/data/hohna_datasets_fasta
• h3n2_na_20, zika: https://github.com/neherlab/treetime_examples
• yeast: https://cran.r-project.org/web/packages/phangorn/index.html (comes with pre-loaded datasets including yeast)

As mentioned in the submission, we plan to add more optimiation schemes using Phylo2Vec, e.g., MCTS or gradient descent.

See https://github.com/Neclow/gradme