This repository contains code to reproduce some of the plots and analysis in the paper "Leveraging DAGs to improve context-sensitive and abundance-aware tree estimation."

Environment Setup:

Running the scripts in this repo requires a python environment with gctree and other dependencies installed. The dependencies are listed in requirements.txt and are all installable with pip. However, the conda environment file environment.yml is also provided.

git clone git@github.com:matsengrp/gctree2-validation-plots.git
cd gctree2-validation-plots

conda env create -f environment.yml
conda activate gctree2-validation-plots

Producing plots:

To produce the plots, you must first edit gctree_benchmark_direct_run.sh to point it toward the simulated data and inference outputs. This may be done by editing the variables sim_prefix and inference_prefix at the top of the script. The values for sim_prefix and inference_prefix specify where to find the outputs from running the simulation and inference code, which is provided with the paper.

To produce the plots, run gctree_benchmark_direct_run.sh in the environment set up as described above. This script will send jobs to a cluster using sbatch. If you do not have access to a cluster using sbatch, the use_cluster variable at the top of the script will need to be set to 0. The script will not finish until all sbatch jobs for your user are finished running, so be sure all unrelated cluster jobs are finished before running this script.

Plot outputs:

Figure 4 from the paper is the file hdag_comparison_faceted.pdf. A variety of plots similar to Figure 5 will be placed in the directory example_sim_scatters. For the paper, we selected one which showed the following:

• improvement in RF distance in history sDAG trees compared to dnapars trees,
• many more history sDAG trees than dnapars trees, and
• qualitative correlation between RF distance improvement and the two likelihoods.