W0199

Beyond the Sparse Matrix Screen: A Web Service for Randomly Generating Crystallization Experiments. Brent W. Segelke and Bernhard Rupp, Biomolecular Crystallography Facility at Lawrence Livermore National Laboratory, Livermore, CA 94551

Random (Shieh et. al. 1995) and pseudo-random (Carter et. al. 1979, Jancarick et. al. 1991, Cudney et. al. 1994) crystallization screens have been previously described and are increasingly popular techniques for initiating de novo crystallization of macromolecules. We have previously demonstrated the inherent efficiency of such screening techniques (Segelke, manuscript in preparation) and are now in the process of developing a Web based service that will generate any number of random combinations of crystallization conditions. The foreseeable benefit is manifold but our rational is mostly statistical. A cursory review of available literature gives some insight into the probability of finding crystallization conditions, at least for well behaved proteins. From sixty six crystallization efforts found in five literature sources, the average proportion of successful experiments for crystallization was approximately 10% and ranged from 2% to 68%. For a protein that has a 2% probability of crystallizing in any one experiment, there is approximately a 36% chance of not finding a useful condition in the 50 trials of the "Sparse Matrix" screen. If one were to employ random sampling and perform 200 randomly generated experiments, the chance of simply failing to find a useful condition is greatly reduced (approximately 2% chance). Having performed these 200 experiments, one can be confident that they are dealing with a stubborn molecule and proceed with more creative techniques to generate cooperative material (such as mutating a large number of hydrophobic sidechains (Dyda et. al. 1994)). Perhaps an even greater benefit will be the flexibility to customize a screen by use of variable sets of precipitants, concentration ranges, buffers, or other components thereby generating a large array of possible experiments rather than the fixed combination in prefabricated screens. Current versions of the service handle e-mail submission of job requests and return mail of the randomly generated experiments, but a fully interactive Web interface and tutorial are under construction.