The Utilization of Reduced Dimensional Representations of Molecular Structure for Rapid Molecular Similarity Calculations

The availability of two-dimensional representations of molecules which retain structural information permits the application of techniques from digital image processing to be applied to molecular similarity. Here three such approaches, invariant moments, radial integration and radial scanning, are investigated. They overcome the time consuming adjustment of the relative orientations of the molecules to be compared so as to optimize the similarity and can speed up the overall calculations by many orders of magnitude, hence offering a technique for dealing with the enormous numbers of compounds being generated by high throughput synthesis and combinatorial chemistry. These techniques are applied to the well-known set of steroids which are frequently used as a benchmark set for QSAR studies.