Setup

git clone https://github.com/KoslickiLab/branch-lengths-assignment.git
cd branch-lengths-assignment
pip install -r requirements.txt

Experiment 1: recovering branch lengths using compatible distance matrices

The tests in this part correspond to section III.B of the manuscript. There are two tests in this section. The first evaluates the performance of bottom-up, naive NNLS and regularized NNLS methods in restoring branch lengths in the event when the distance matrix is completely compatible with the tree.

mkdir -p data/test_data/test1
cd test_scripts
bash test1_perfect_scenario_method_performance.sh
bash test1_selection_factor_effect.sh
cd ..

Experiment 2: recovering branch lengths using pairwise distances with errors

The test in this part corresponds to section III.C of the manuscript.

mkdir -p data/test_data/test2
cd test_scripts
bash test2_distance_matrix_with_errors.sh
cd ..

Experiment 3: comparing efficiency between NNLS and bottom-up methods

The test in this part corresponds to section III.D of the manuscript.

mkdir -p data/test_data/test3
cd test_scripts
bash test2_efficiency.sh
cd ..

Application: FunUniFrac

Obtaining the KEGG data

The details for KEGG data extraction can be found in the extraction repo. A cleaned version of KEGG tree rooted at ko00001 can be found in data/fununifrac_data/kegg_trees/kegg_ko00001_no_edge_lengths.txt.

Obtaining pairwise distance matrix

To download a pre-built version of the pairwise distance matrix, do the following:

cd data/fununifrac_data/pw_distance_files
wget https://zenodo.org/records/13129003/files/KOs_sketched_scaled_10_k_5