Find-a-Drug has been set up by Treweren Consultants, the developers of the THINK software. Our objectives are to find potential new drug molecules for a wide range of protein targets including Cancer and HIV. It is intended that this will operate on a not for profit basis with some funding coming from commercial projects, sponsors and licensing arrangements for the results.

Unlike the United Devices and other general purpose distributed computing architectures, this project uses server and client software that has been designed exclusively for the application. Consequently, the architecture is more suitable for THINK, easier to maintain and more lightweight, requiring fewer resources on the members' PC clients as well as on the servers. In addition, the implementation allows multiple servers to be used, eliminating a potential bottleneck.

Over the last 12 months significant progress has been made to increase the speed of the software and improve the quality of the results. We are very grateful to Intel who helped to identify some algorithms which, when re-written, delivered a 20- to 40-fold increase in speed. There is always a need for balance between the time taken per job and the accuracy of the predictions. The latest version of THINK is used in this project and refines the hits to eliminate molecules that are not predicted to have strong interactions.

This project has now been running for one year and we have almost finished searching 3.5 billion molecules for 12 proteins. During this time we have demonstrated the approach of using 35 million starting molecules from which 100 analogues or derivatives are generated on the members' PCs, giving the total of 3.5 billion. At one time it was expected that 16 protein targets would be evaluated on the United Devices platform, but currently this seems very unlikely without further sponsorship. The Oxford University web site includes 3D histograms for the results like the following: These show the distribution of hits across a range of libraries and the quality of hit. Some initial analysis has revealed that the hit rate in the E8 library (not shown in this histogram) is consistently one of the lowest. This type of library is known as a UGI library and is very large - containing 28 million molecules and representing about 80% of the starting molecules. However, the quality and quantity of the resulting hits probably does not warrant continued use of this type of library.

The alpha test started on 18^th April and the number of members has grown steadily during the past few weeks. The purpose of an alpha test is to find and resolve problems, and we are pleased that few problems have been found. We have been concentrating on testing the communications and improving the robustness of the software. We have also taken the opportunity to make some simplifications to the registration procedure and to test the automatic software upgrade. Perhaps inevitably, some members experienced problems which were best addressed by re-installing the most recent revision of application, and the message board has proved a useful forum for discussion of problems and solutions.

We hope to start the beta test late this month, and to achieve production quality software before the end of June. During the beta test we plan to increase the number of molecules and concurrent projects as well as provide web pages to monitor progress. If you are not already participating or have been discouraged by problems, then please accept our assurances that your contribution and feedback would be greatly appreciated.