Contribution of 2D,3D structural features of drug molecules in the prediction of Drug Profile Matching
We recently introduced Drug Profile Matching (DPM), a novel affinity fingerprinting method capable of predicting the complete effect profiles of small molecules based on their interaction patterns which are generated by flexible docking to a series of rigidly handled non-target protein active sites. DPM was found to classify molecules excellently and the question naturally came up: What is the contribution of 2D and 3D structural features of the small molecules to the surprisingly high prediction power of the docking-based DPM? To answer this question, the performance of DPM and 2D and 3D similarity fingerprinting approaches using ChemAxon JChem Base and Calculator Plugins were compared. Drug classification was carried out for a two-level hierarchical effect database using a set of cca. 1200 FDA-approved small molecule drugs. To get a more realistic view about the feasibility domain of DPM, its predictive power was also tested on external data. We found that 2D and 3D similarity fingerprinting of rigid structural categories produced as similar high prediction power as DPM, however, for many effects DPM was able to overcome the common screening problems of 2D and 3D similarity searches arising from the presence of structurally diverse molecules.