A machine learning approach to weighting schemes in the data fusion of similarity coefficients.

The application of data fusion techniques for combining the results of similarity searches of chemical databases has been shown to improve search performance. When used to combine the results of searches using different similarity coefficients, the optimum combination is dependent on the size, in terms of substructural fragments present, of the molecules being compared. This paper describes preliminary simulation tests which aim to automatically deduce, using machine learning techniques, the optimum combination of similarity coefficient which may be combined using data fusion for a given class of active compounds.