High-Throughput Modeling of Human G-Protein Coupled Receptors:  Amino Acid Sequence Alignment, Three-Dimensional Model Building, and Receptor Library Screening

The current study describes the development of a computer package (GPCRmod) aimed at the high-throughput modeling of the therapeutically important family of human G-protein coupled receptors (GPCRs). GPCRmod first proposes a reliable alignment of the seven transmembrane domains (7 TMs) of most druggable human GPCRs based on pattern/motif recognition for each of the 7 TMs that are considered independently. It then converts the alignment into knowledge-based three-dimensional (3-D) models starting from a set of 3-D backbone templates and two separate rotamer libraries for side chain positioning. The 7 TMs of 277 human GPCRs have been accurately aligned, unambiguously clustered in three different classes (rhodopsin-like, secretin-like, metabotropic glutamate-like), and converted into high-quality 3-D models at a remarkable throughput (ca. 3s/model). A 3-D GPCR target library of 277 receptors has consequently been setup. Its utility for "in silico" inverse screening purpose has been demonstrated by recovering among top scorers the receptor of a selective GPCR antagonist as well as the receptors of a promiscuous antagonist. The current GPCR target library thus constitutes a 3-D database of choice to address as soon as possible the "virtual selectivity" profile of any GPCR antagonist or inverse agonist in an early hit optimization process.