COSMO f rag :  A Novel Tool for High-Throughput ADME Property Prediction and Similarity Screening Based on Quantum Chemistry

The COSMO−RS (Continuum Solvation Model for Real Solvents) method has proven its broad applicability for the accurate prediction of thermodynamic, environmental, or physiological properties. On the basis of quantum chemical calculations with COSMO, COSMO−RS calculations were unavoidably restricted to small- to medium-sized compound sets, because of the time demand of the COSMO calculations. The COSMOfrag method, presented here, overcomes this restriction by replacing the costly quantum chemistry step with a selection of suitable fragments from a database of, presently, 40 000 DFT/COSMO precalculated molecules. Since, in the COSMO−RS picture, any molecular information is gathered in the so-called σ profiles, COSMOfrag replaces the single σ profile with a composition of partial σ profiles, selected by the use of extensive similarity searching algorithms. On five representative datasets, the accuracy loss of COSMOfrag versus full COSMO−RS calculations has been shown to be only in the range of 0.05 log units. From the performance point of view, it is now possible to carry out COSMO−RS property calculations for more than 100 000 compounds a day per standard PC CPU