Differentiable Search of Evolutionary Trees from Leaves

Our work introduces a differentiable approach to phylogenetic tree construction, optimizing both tree and ancestral sequences.

Pre-print - https://www.biorxiv.org/content/10.1101/2023.07.23.550206v1

• ICML 2023 Workshop (SODS/DiffAE) Poster (PDF)
• Eric J. Ma has written a very detailed article on our paper's key contribution : making the trees and sequences differentiable. You can read it here. It does a great job at explaining our method.

To run examples in colab, click the below link

Checklist

• Make sure to select GPU (or remove the -g 0 flag when running)
• You can specify your wandb account if you intend to log statistics/tree illustrations

Example : running for trees with 16 leaves

• To run for different number of leaves change the -l to the desired value

Other params :

• sequence length : -sl
• mutations per bifurcation : -m
• alphabet size : -nl
• epochs/steps : -e
• initialization count to run in parallel : -ic

During running, every 200 steps it will print the soft_parsimony_score and parsimony_score (last two values in each line)

!python train_batch_implicit_diff.py -l 16 -nl 20 -m 50 -sl 256 -tLs [0,0.005,10,50] -lr 0.1 -lr_seq 0.01 -t float64-multi-init-run -p Batch-Run-Maximum-Parsimony -alt -n "Final Run" -g 0 -e 5000 -ai 1 -ic 50 -s 42

Current Limitations :

• [ ] Groundtruth trees we evaluate against (optimal solutions) are perfect binary trees. We need to evaluate on diverse grountruth trees of uneven leaf levels
  • [ ] Full binary trees (ramithuh/differentiable-trees#30)
  • [ ] Then, binary trees in general
• [ ] Get rid of site-wise independence assumption

We are working on these aspects in another repo : https://github.com/ramithuh/differentiable-trees. Once those are tested and verified, this repo will be updated. If you have any suggestions/comments/feedback feel free to reach us.