THINK (To Have Information aNd Knowledge) is a new innovative computer program produced by the founders of Chemical Design to assist with Drug Discovery Lead Generation and Optimisation process and reduce time-scales. It is a modular product designed to be used with other commercial and proprietary software programs. THINK is supported on Windows and LINUX platforms.

Core Module

The Core module provides:

• Chemical structure file import and export of MACCS SD, SMILES and PDB formats (including calculated properties and stereochemisty). CSV (comma separated variable) files may also be exported to spreadsheets such as Excel. Combinatorial chemistry libraries may also be enumerated from their R-groups or component parts.
• Various listing output of molecular data including atom types, connectivity (including bond orders) and geometry.
• Command script for batch and server jobs.

There is no charge for the Core module which is only available with other modules.

2D Module

The 2D module provides some innovative data analysis and de novo derivative generation capabilities as well as more traditional 2D exact, substructure and similarity searching. Functional group keys are used to create fingerprints which are used in the similarity searching. The R-group search is a variant of the substructure search which generates R-groups or components from reagents for Combinatorial Chemistry libraries.

The de novo derivative generation is intended to explore the drug-like synthetic accessibility from a "hit" in order to evaluate its potential as a good "lead". It uses a genetic algorithm to generate hundreds or thousands of derivatives of a known active starting molecule. A series of rejection tests are performed using ranges for calculated properties and undesirable substructures (such as those which are associated with metabolism or toxicity) prior to an annealing step using functional group similarity to active and inactive molecules. The default rejection criteria include drug-like criteria but these can be extended to increase the probability of the derivatives being active.

A learning or training option which uses Tertiary SAR analysis may be used to create the rejection criteria from experimental data. The analysis identifies properties, functional groups and/or pharmacophores (licenced separately) to distinguish active molecules from inactive molecules. Such criteria may be useful to select molecules to test as well as providing a measure of the quality of a hit. Property diversity measurement and selection are also included

3D Module

The 3D module also includes the ability to generate 3D coordinates (including fused rings etc) from connection tables when these are not imported and to search molecules using 3D queries.

At the heart of the 3D module is the conformational analysis engine. This provides the ability to sample low energy conformations systematically or randomly and to generate a representative subset. The implementation allows ring conformations to be searched in unfused and fused rings and includes the use of VdW or CPK contact accelerators. Conformations generated may also be displayed using the Graphics and GUI module, listed to the screen or saved to a MACCS SD file.

Pharmacophore Module

This module requires the 3D module to generate conformers from which 2, 3 and 4 centre pharmacophores are identified. The pharmacophores for all low energy conformations of a molecule can be saved in a file for importing into a MySQL or ORACLE databases. This approach permits the use of standard SQL queries to extract common pharmacophores within sets of molecules or pharmacophores which discriminate between actives and inactives. A bin model is used for the pharmacophore distances with fuzzy boundaries.

The 'Active Volume Constraint' technology is included in this module which allows volume constraints to be constructed for each common pharmacophore for a set of active molecules. Lead explosion and optimisation can utilise pharmacophore searches using the volume constraint to eliminate most of the false positives.

The Receptor Site search or Virtual Screening capabilities of THINK use pharmacophores to quickly eliminate conformations of molecules which cannot binding in the receptor site. The initial position of the molecule in the receptor site is defined by a match of atoms in the molecule to a pharmacophore. This is then refined by adjusting torsion angles, x,y,z position and orientation of the molecule to increase the binding energy predicted by the ChemScore function. The implementation is very robust, fast and accurate. In the Find-a-Drug project which uses thousands of PCs on the Internet up to 250,000 molecules per 2GHz PC day are processed with about 20% of the hits (molecules predicted to be active) showing the desired experimental activity in National Cancer Institute tests. The implementation includes side-chain relaxation and automatic elimination of water when appropriate.

Graphics and GUI Module

The Graphics and GUI are currently only available for the Windows version of THINK and uses OpenGL. For structures that are generated by THINK or read from files which do not contain 2D coordinates (such as SMILES), 2D representations are generated in order that structures may be displayed. 3D stick representations may also be displayed and manipulated (the 3D Module is required to generate 3D coordinates). The graphical output includes the ability to generate 2D diversity plots and functional group distribution plots and histograms which may be coloured by activity.

The GUI provides a series of dialogs which may be used in place of command scripts or entering commands from the keyboard. A simple spreadsheet for reviewing numerical molecular properties is also included.

Screening Database

The THINK software was used for a very large scale virtual screening project in collaboration with Oxford University and is being used for the Find-a-Drug project. Such projects use PC time donated by individuals and connected over the Internet. This provides a massive computing capacity capable of screening billions of drug-like molecules. This module consists of our current collection of drug-like molecules. Further details of this collection and the procedures used to generate it are described here. To make effective use of this collection a Site Licence of the software is required as many PCs are required in order to screen a billion molecules.