In Molecular Engineering one tries to rationally design and synthesise new molecules and materials that have certain desirable properties. So rather than start from first-principles, put molecules together and then observe the properties, in Molecular Engineering one works the other way around.

In general, this is a very compute and data intense problem. There can easily be as much as 10**9 molecules involved in a single design task and a large number of temporary data (10**12 descriptors) that could be available in geographically distributed databases.

This typically lends itself to Grid computing. The OpenMolGrid project, started on September 1st 2002, is addressing this issue, taking the popular package OpenMol as a starting point.

The OpenMolGRID analysis process requires the location, extraction, and integration of globally distributed databases describing environmental, toxicological, chemical, biological, and pharmacological information. OpenMolGrid will be using the Unicore Grid software to integrate these resources, systems, and tools.

The project will also create software modules for the calculation of molecular descriptors for statistical analysis and the prediction of biological activity from the molecular structure.

Using molecular engineering, OpenMolGrid will generate candidates for chemical compounds with predetermined target properties from appropriate fragments according to established rules. All constructed candidate structures with the target properties are validated by a standard procedure. This involves massive number crunching, which makes it a candidate for Grid computing. As a result, a small subset of candidate structures that match the predefined target properties are selected for testing in a physical laboratory.

After creating the software and setting up the infrastructure, the OpenMolGrid performance will be tested in real life chemical and pharmaceutical applications. In an initial test, the structure-activity model for 20,000 structures will be created and tested in the laboratory. The accuracy of the models will be validated. In the second step, the predictive capability of the models constructed in the previous step will be tested against the biological activity of 10,000 unknown structures. When these tests are successful, the OpenMolGrid system will be taken into production. But that will only be in two years from now, as the project is anticipated to be finished in November 2004.

Apart from CSC, partners in OpenMolGrid are the Istituto di Ricerche Farmacologiche "Mario Negri" in Italy, Comgenex in Hungary, the University of Ulster in the United Kingdom, and the Forschungszentrum Jülich Gmbh in Germany.

More information on the OpenMolGrid can be found at: http://www.openmolgrid.org/. For more information on the other products and projects mentioned, consult: http://EnterTheGrid.com.