Data-driven discovery of linear molecular probes with optimal selective affinity for PFAS in water

This repository demonstrates the data-driven discovery of linear molecular probes with optimal selective affinity for per- and polyfluoroalkyl substances (PFAS) in water. It integrates molecular dynamics (MD) simulations, enhanced sampling methods, deep representational learning via variational autoenecoders, surrogate model training and multi-objective Bayesian optimization using random scalarations. We use perfluorooctanesulfonic acid (PFOS) as the target PFAS and sodium dodecyl sulfate (SDS) as a representative interfernt to demonstrate our approach.

---

Installation

INSTALL_LOCATION=<set path here for installation of conda env>

conda env create -f environment.yml --prefix $INSTALL_LOCATION

---

Usage

1. Codes: Estimation of binding free energies & binding constants. Please see Codes/README.md for getting started.
2. Data: GROMACS simulation files and PLUMED template file (master-pbmd-files-final) for performing enhanced sampling, probes embedding in latent space and helper data (VAE_data), JCED Supplementary Information data (JCED_data_each_cycle), binding free energy and binding constant all cycles combined and in each cycle (GPR_training_data),and probe structures in each cycle (smiles-each-cycle). Please see Data/README.md.
3. Notebooks: Analysis python notebooks for calculating potential of mean force (PMF) profiles, binding free energies & binding constants, GPR (Gaussian process regression) training, multi-objective Bayesian optimization using random scalarizations, and example notebook (Notebooks/example_notebook_to_read_SI_tables_deltaG_Kb.ipynb) to read data reported in Supplementary Information of JCED (Data/JCED_data_each_cycle). Please see Notebooks/README.md.

---

Cite

If you use the codes or notebooks from this repo in your work, please cite:

S. Dasetty, M. Topel, Y. Tang, Y. Wang, E. Jonas, S. Darling, J. Chen, A. L. Ferguson. "Data-driven discovery of linear molecular probes with optimal selective affinity for PFAS in water" XXXXX. DOI: XXXX

@article{ferglab2023PFAS,
  title={Data-driven discovery of linear molecular probes with optimal selective affinity for PFAS in water},
  author={Dasetty, S. and Topel, M. and Tang, Y. and Wang, Y. and Darling, S. and Chen, J. and Ferguson, A.L.},
  journal={XXXX},
  volume={XXXX},
  number={XXXX},
  pages={XXXX-XXXX},
  year={2023},
  publisher={XXXX}
}