This is a Python implementation of the molecular dynamics fingerprints (MDFP) methodology for building predictive models for phys-chem properties as delineated in our publication.
This toolkit is described and applied to the SAMPL6 logP prediction challenge, with the results published here.
Visit our documentation to learn details about installation, example workflow and API references.
The openeye entry is the basic version. The rdkit handler version allow more customisation, e.g. specifying custom conformer for the molecule
Bibtex citations for the toolkit and the method are as follows:
@article{esposito2020combining,
title={Combining Machine Learning and Molecular Dynamics to Predict P-Glycoprotein Substrates},
author={Esposito, Carmen and Wang, Shuzhe and Lange, Udo EW and Oellien, Frank and Riniker, Sereina},
journal={Journal of Chemical Information and Modeling},
volume={60},
number={10},
pages={4730--4749},
year={2020},
journal = {Journal of Chemical Information and Modeling}
}
@article{Wang2019,
doi = {10.1007/s10822-019-00252-6},
url = {https://doi.org/10.1007/s10822-019-00252-6},
year = {2019},
month = nov,
publisher = {Springer Science and Business Media {LLC}},
author = {Shuzhe Wang and Sereina Riniker},
title = {Use of molecular dynamics fingerprints ({MDFPs}) in {SAMPL}6 octanol{\textendash}water log P blind challenge},
journal = {Journal of Computer-Aided Molecular Design}
}
@article{Riniker2017,
doi = {10.1021/acs.jcim.6b00778},
url = {https://doi.org/10.1021/acs.jcim.6b00778},
year = {2017},
month = apr,
publisher = {American Chemical Society ({ACS})},
volume = {57},
number = {4},
pages = {726--741},
author = {Sereina Riniker},
title = {Molecular Dynamics Fingerprints ({MDFP}): Machine Learning from {MD} Data To Predict Free-Energy Differences},
journal = {Journal of Chemical Information and Modeling}
}
Copyright (c) 2021, Greg Landrum, Shuzhe Wang, Carmen Esposito
Project based on the Computational Molecular Science Python Cookiecutter