riboswitch-epistasis

Calculations and analysis files for "Disentangling contact and ensemble epistasis in a riboswitch" by Wonderlick, Widom and Harms.

Data files

• Our experimental data is in 2AP_corrected.csv
• All of our ML fitting results and MCMC samples are in the all-samples directory.

Models

The binding models are implemented in a set of python scripts:

• 4.5 (four_state_five_param.py)
• 4.4 (four_state_five_param.py)
• 3.4 (three_state_four_param.py)
• 3.3 (three_state_three_param.py)
• 2.3 (two_state_three_param.py)
• apparent binding constants, not shown in manuscript (kapp_one_param.py)

Setting up environment

To set up the environment for reproducing and extending the analyses in the paper, create a python environment (>=3.8) with standard scientific computing libraries: (matplotlib scipy tqdm numpy pandas jupyter-lab). We recommend conda.

The only non-standard dependency is the "likelihood" library. This can be installed using the following commands:

git clone git@github.com:harmslab/likelihood.git
cd likelihood
python setup.py install

Reproducing fits

All ML and MCMC fits can be reproduced by running:

bash run-all-fits.sh

This could take a long time. (For the manuscript, we ran the MCMC samples in parallel on a computing cluster). You might want to run each fit with its own script.

Reproducing figures and tables

The analyses shown in the figures and tables can be reproduced using two jupyter notebooks:

• Fig2B-S2-S3_TableS1-S2.ipynb
• Fig-3-4-5-6.ipynb

Notes are in each notebook.