United Nations Industrial Development Organization

in aplhabetical order QSAR MODELING AND LIBRARY DESIGN STRATEGIES Dr. Wolfram Altenhofen Chemical Computing Group AG, Lörrach, Germany [email protected] The session will be devided into an introduction to basic concepts of QSAR Modeling and Library Design and a hands-on tutorial which will allow participants to experience the basic steps from deriving a QSAR model to designing a focused library themselves. In the theory section, a general overview on • representation of chemical structures in the context of computer applications, • deriving physico-chemical properties • the theory of ligand-protein interactions • building QSAR models • strategies for library design • will be presented. During the tutorial, a methodology is presented that guides through the drug design cycle starting from the analysis of experimental HTS data, constructing a QSAR model and using the model to design a virtual focused combinatorial library for cyclic GMP Phospho-diesterase V inhibitors in an almost fully automated way. The analysis of the experimental dataset is based on 2.5D descriptors. These descriptors are fast and easy to calculate since they rely on 2D information and still reflect 90 % of the information inherent in 3D structures. They were specifically designed to provide a tool for a rapid though stable initial approach to large datasets of unknown SAR. The descriptor values correspond to binned van-der-Waals surface areas. The binning procedure was based on logP, MR and partial charge (PEOE), supposed to be fundamental physico-chemical properties that cover most of the relevant property space in an intuitive and interpretable manner. The QSAR model applies a non-linear probabilistic binary method rather than a linear regression based technique. The focused library design uses virtual enumeration with a binary QSAR model as product-based scoring agent for reagent selection. The dataset consists of about 400 known cGMP Phosphodiesterase V inhibitors with activity data selected from the literature and a total of 1800 molecules. The initial QSAR model is about 20 times more potent in selecting active compounds over random picking. The building blocks (2 x 10 x 12 x 27 = 6500 potential products) used in the combinatorial design of a focused quinazoline library (1 x 3 x 3 x 5 = 45 products) reflect chemical intuition and input from the literature. Using the binary QSAR model as focusing agent the percentage of predicted active compounds increases from 5 % in the unfocused library to 75 % in the focused library. The resulting focused library preserves the essential SAR known from the literature. Role of Combinatorial Chemistry in Original Drug Discovery Péter Arányi CHINOIN Co. Ltd., Budapest, Hungary [email protected] Combinatorial synthetic methods became a routine in drug discovery during the nineties. Use of combinatorial libraries find two well discernible applications. In order to identify random hits, a diverse combinatorial library can be added to in-house existing compounds and tested in first screen assays. Later in the discovery process a focussed library is more useful to optimize the structure in order to get a lead. Several different technical solutions exist today. The most straightforward approach apparently is parallel synthesis of individual compounds. An aspect that should be considered while designing the basic scaffold (and set of substituents) is drug-likeness of the resulting compounds. Known toxicophores, mutagenic cores, alkylating, acylating or other highly reactive side chains should be avoided. Molecular weight of the compounds should remain below or in the vicinity of 500. Many published libraries are built around core structures of known drugs on the market or in development. Structures that are not stable in the biological milieu, or otherwise have poor bioavailability, such as peptides or alkyl esters are defavorized even if their chemistry is easy