Recent work in Sheffield on similarity searching has focussed on the use of data fusion and fragment weighting methods to search the MDDR, WOMBAT and MUV databases.  Data fusion involves the combination of multiple similarity searches.  The overlap between multiple searches is shown to follow a Zipf-like, power law distribution, with very few molecules (or active molecules) common to multiple searches; and a comparison of a large number of different group-fusion algorithms shows that one based on molecules' inverse rank positions is the most effective of those tested. Information about the frequencies with which fragments occur in molecules can be used in two ways to increase search effectiveness (when compared with using just the presence or absence of fragments in molecules): using functions of the frequencies of fragment occurrences in individual molecules, and using inverse functions of the frequency of fragment occurrences in the database as a whole.

Weighting and fusion methods for similarity-based virtual screening

240th National Meeting (2010)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.