Description: Converging a Knowledge-Based Scoring Function: DrugScore2018

This title appears in the Scientific Report : 2019

Converging a Knowledge-Based Scoring Function: DrugScore2018

We present DrugScore2018, a new version of the knowledge-based scoring functionDrugScore, which builds upon the same formalism used to derive DrugScore, but exploits atraining data set of nearly 40,000 X-ray complex structures, a highly diverse and the by farlargest dataset ever used for such an end...

Personal Name(s):	Dittrich, Jonas
	Schmidt, Denis / Pfleger, Christopher / Gohlke, Holger (Corresponding author)
Contributing Institute:	Strukturbiochemie; ICS-6 Jülich Supercomputing Center; JSC John von Neumann - Institut für Computing; NIC
Published in:	Journal of chemical information and modeling, 59 (2019) 1, S. 509–521
Imprint:	Washington, DC American Chemical Society64160 2019
PubMed ID:	30513206
DOI:	10.1021/acs.jcim.8b00582
Document Type:	Journal Article
Research Program:	Forschergruppe Gohlke Physical Basis of Diseases Computational Science and Mathematical Methods
Link:	Restricted Published on 2018-12-04. Available in OpenAccess from 2019-12-04. Restricted
	Publikationsportal JuSER

Please use the identifier: http://hdl.handle.net/2128/21489 in citations.
Please use the identifier: http://dx.doi.org/10.1021/acs.jcim.8b00582 in citations.

We present DrugScore2018, a new version of the knowledge-based scoring functionDrugScore, which builds upon the same formalism used to derive DrugScore, but exploits atraining data set of nearly 40,000 X-ray complex structures, a highly diverse and the by farlargest dataset ever used for such an endeavour. About 2.5 times as many pair potentials thanbefore now have a data basis required to yield smooth potentials, and pair potentials could nowbe derived for eight more atom types, including interactions involving halogen atoms and metalions highly relevant for medicinal chemistry. Probing for dependence on training data set sizeand quality, we show that DrugScore2018 potentials are converged. We evaluated DrugScore2018in comprehensive scoring, ranking, docking, and screening tests on the CASF-2013 dataset,allowing for a comparison with >30 other scoring functions. There, DrugScore2018 showedsimilar or improved performance in all aspects when compared to either DrugScore,DrugScoreCSD, or DSX and was, overall, the scoring function showing a most consistently goodperformance in scoring, ranking, and docking tests. Applying DrugScore2018 as objectivefunction in AutoDock3 in a large-scale docking trial, using 4,056 protein-ligand complexesfrom PDBbind 2016, reproduced a docked pose to within 2 Å RMSD to the crystal structure in>75% of all dockings. These results are remarkable as the DrugScore2018 potentials werederived from crystallographic information only, without any further adaptation using bindingaffinity or docking decoy data. DrugScore2018 should thus be a competitive scoring andobjective function for structure-based ligand design purposes.