Optimization of the druglikeness of chemical libraries

JENS SADOWSKI

ZHF/G -A 30, BASF AG, D-67056 Ludwigshafen, Germany

Summary. A scoring scheme for the classification of molecules into drugs and non-drugs was established. It was set up by using atom type descriptors for encoding the molecular structures and by training a feed-forward neural network for classifying the molecules. The approach was parameterized by using large databases of drugs and non-drugs - the Available Chemicals Directory (ACD) with 169 331 molecules and the World Drug Index (WDI) with 38 416 molecules. It was able to reveal features in the molecular descriptors that either qualify or disqualify a molecule for being a drug. The method classified about 80% of the ACD and the WDI correctly. It was extended to the application for crop protection compounds and can be used to prioritize compounds for synthesis, purchase, or biological testing. An enhancement allows to optimize the drug character of combinatorial libraries.

Key words: drug-likeness, fingerprints, genetic algorithms, neural networks Introduction

With the upcoming experimental methods for handling large numbers of compounds in drug design - high-throughput screening and combinatorial chemistry - the focus of molecular modelling in this area was on diversity [1]. The questions to computational chemistry were most often like: Which is the most diverse subset of a given set of compounds? However, more recently a need for methods that can handle additional criteria like drug-likeness [24] or bioavailability [5] was discovered. The questions are now manifold, like: Which compounds are drug-like, toxic, or bioavailable? How can these multiple criteria including diversity be optimized simultaneously in library design? In the following, some methods for answering these questions will be demonstrated.

0 0

Post a comment