This is the repository of the final project for APC 523 of Noah Cape. The project is title: Reconstructing Phylogenies using DNA Typewriter Data.
IMPORTANT Before submitting either executable below make sure to run:
source ./setup_env.sh
This will create a venv, install the required packages, and then in the job scripts the venv will be activated.
The key files are as follows:
-
./py/model.py contains the main data abstraction of DNA Typewriter tapes
-
./py/n_model.py contains the computation of the likelihood over a fixed tree as well as the parameter optimization code (note that ./py/model.py contains stale code unoptimized which I compared against n_model.py)
-
./py/tree_search.py contains code for searching tree space
-
./py/myo.py contains io code
-
Other files contain checks and plotting functions for evaluating and visualizing results
The following executables contain code for submitting jobs to the cluster:
- test_reconstruction.sh will reconstruct a small tree
- process_reconstruction.sh will evaluate the reconstruction accuracy against the true tree of a larger pre-reconstructed instance
Note that these are small test instances as larger instances take on the order of 10s of hours to complete.
./src contains an old implementation of the likelihood computation using Rust as well as some simulation code that was used to generate data for testing.
This codebase requires the following python packages:
| Package | Version |
|---|---|
| diffrax | 0.7.2 |
| equinox | 0.13.6 |
| jax | 0.9.2 |
| jaxlib | 0.9.2 |
| matplotlib | 3.10.9 |
| networkx | 3.6.1 |
| numpy | 2.4.3 |
| optax | 0.2.8 |
| pandas | 3.0.2 |
| scipy | 1.17.1 |