Quantitative structure-activity relations based on quantum theory and wavelet transformations

We describe the use of wavelet-transformed quantum-chemical property fields for three-dimensional quantitative structure-activity relations (3D-QSARs) and compare this new ansatz with standard approaches like comparative molecular field approach (CoMFA) or comparative molecular similarity approach (CoMSIA) for datasets from the literature. The gain in CPU-time and memory through using the wavelet transformation permits evaluation of many QSAR models to find the optimum one, the inclusion of nonlinear terms in the PLS is trivial. The limiting factor is still the quantum-chemical calculation of property fields on dense grids, preferentially by density functional theory, for the set of molecules.