Abstract
G protein-coupled receptors (GPCR), a member of 7TM, are the important targets of drugs. Amongst them, about 150 are called “orphan receptors” for unknown ligands, which are regarded as the important candidates for new drug design. Here we attempted to predict their ligands by use of machine learning. We used quantified and vecterized properties of amino-acids surrounding known ligand-binding sites of class A GPCR for the classification of receptors. As the result, we found eight orphan receptors classified with other receptors of common ligands and verified by use of docking simulation.