Virtual screening using reduced graphs


Valerie J. Gillet, Information Studies, Information Studies, University of Sheffield, Regent Court, 211 Portobello Street, Sheffield, S1 4DP, United Kingdom
Many different ligand-based virtual screening methods have been developed using both 2D and 3D descriptors. Both types of descriptors have their limitations. The 2D methods have a tendency to select structural analogues and thus do not easily permit the identification of new lead series. The 3D methods, on the other hand, have been shown to result in greater diversity in the hitlists, however, they are limited by the need to handle conformational flexibility. We have developed virtual screening methods in which the molecules are characterised by reduced graphs which summarise the features of the molecules while retaining the topology between the features. Thus reduced graphs can be thought of as topological pharmacophores. Two different approaches have been investigated for quantifying the similarity of reduced graphs. In one approach, the reduced graphs are mapped to fingerprints before calculating the similarity, in the other, graph-matching methods are applied directly to the reduced graphs. Here, the performance of the reduced graphs is compared with conventional descriptors in simulated screening experiments.

Herman Skolnik Award Symposium
2:00 PM-4:20 PM, Tuesday, August 24, 2004 Pennsylvania Convention Center -- 110A&B, Oral

Division of Chemical Information

The 228th ACS National Meeting, in Philadelphia, PA, August 22-26, 2004