Molecular properties of psychopharmacological drugs determining non-competitive inhibition of 5-HT3A receptors.

We developed a structure-property-activity relationship (SPAR)-model for psychopharmacological drugs acting as non-competitive 5-HT(3A) receptor antagonists by using a decision-tree learner provided by the RapidMiner machine learning tool. A single molecular descriptor, namely the molecular dipole moment per molecular weight (mu/MW), predicts whether or not a substance non-competitively antagonizes 5-HT-induced Na(+) currents. A low mu/MW is compatible with drug-cumulation in apolar lipid rafts. This study confirms that size-intensive descriptors allow the development of compact SPAR models.