There is a need for improved and generally applicable scoring functions for fragment-based approaches to ligand design. Here, we evaluate the performance of a computationally efficient model for inhibitory activity estimation, which is composed only of multipole electrostatic energy and dispersion energy terms that approximate long-range ab initio quantum mechanical interaction energies. We find that computed energies correlate well with inhibitory activity for a compound series with varying substituents targeting two subpockets of the binding site of Trypanosoma brucei pteridine reductase 1. For one subpocket, we find that the model is more predictive for inhibitory activity than the ab initio interaction energy calculated at the MP2 level. Furthermore, the model is found to outperform a commonly used empirical scoring method. Finally, we show that the results for the two subpockets can be combined, which suggests that this simple nonempirical scoring function could be applied in fragment-based drug design.
PubMed ID: 28688090
Publication type: Journal
Journal: J Comput Aided Mol Des
Citation: J Comput Aided Mol Des. 2017 Aug;31(8):715-728. doi: 10.1007/s10822-017-0035-4. Epub 2017 Jul 7.
Date Published: 9th Jul 2017
Registered Mode: by PubMed ID
Created: 19th Aug 2020 at 13:57