Focused library design in GPCR projects on the example of 5‐HT_2c agonists: Comparison of structure‐based virtual screening with ligand‐based search methods

Abstract

The aim of this study was to investigate the usefulness of structure-based virtual screening (VS) for focused library design in G protein-coupled receptors (GPCR) projects on the example of 5-HT(2c) agonists. We compared the performance of structure-based VS against two different homology models using FRED for docking and ScreenScore, FlexX, and PMF for rescoring with the results of 12 ligand-based similarity searches using four different query compounds and three different similarity metrics (Daylight, FTree, Phacir). The result of the similarity search showed much variation, from an enrichment factor up to 3.2 to worse than random, whereas the structure-based VS gave a more stable result with a constant enrichment factor around 2. Additionally, actives retrieved by the structure-based approach were more diverse than the actives among the top scorers of the similarity searches. Based on these results, we suggest basing a focused library design for a GPCR project on a combination of a ligand-based similarity search and structure-based docking.

Ligand and structure-based methodologies for the prediction of the activity of G protein-coupled receptor ligands

“…In 2003, Cavasotto and coworkers demonstrated the applicability of this scoring function to the identification of the binding site of retinal in bovine rhodopsin and to the discrimination between binders and nonbinders of rhodopsin [67]. Similarly, using molecular databases spiked with known binders, a number of studies based on combinations of different scoring functions (such as Gold [68], Dock [69], CScore (Tripos), Fresno [70], Score [71], FlexX [72], and PMF [66]) have demonstrated the applicability of molecular docking at GPCR models to virtual screening [9,10,73]. In agreement with the expectations set by these studies, a variety of docking programs and scoring functions have been successfully applied to the discovery of novel ligands for a plethora of GPCRs, including neurorokinin [74,75], adrenergic [76], chemokine [75,77], dopamine [75], serotonin [75,78], cannabinoid [79], and free fatty acid receptors [80], among others.…”

“…Thus, structure-function studies of GPCRs relied heavily on sequence and phylogenetic analyses, rhodopsin-based homology modeling, and docking studies supported by site-directed mutagenesis and ligand structure-activity relationships (SAR) data [5,6]. In this context, experimentally validated GPCR homology models have proven to be valuable tools for lead identification, and the scientific literature flourished with successful rational drug design and virtual screening examples [7][8][9][10][11]. In 2007 Kobilka and coworkers unveiled the 3D crystal structure of the human β 2 -adrenergic receptor (β 2 -AR), finally providing the long awaited proof that the structure of GPCRs generally resembles that of rhodopsin [12][13][14][15].…”

GPCRHomology Model Development and Application

“…To address this, a number of groups have sought to develop docking-based QSAR models [98,109,[117][118][119][120][121][122][123][124]. A combination of ligand-supported homology modeling and 3D-QSAR models has been applied to the discovery of A2a adenosine receptor antagonists, for example [124].…”

GPCR Structure‐Based Virtual Screening Approach for CB2 Antagonist Search.

“…[6][7][8][9][10][11][12][13][14][15] Among these studies, it has been demonstrated that the homology models of dopamine D3, muscarinic M1, vasopressin V1a receptors, and 5HT 2c were reliable enough to retrieve known antagonists via structurebased virtual screening from several compound databases. 7,14 Rhodopsin-based homology models of the R 1A receptor could be used as the structural basis for the lead finding and optimization through the application of a hierarchical virtual screening procedure. 6 In addition, virtual screening has been successfully performed to identify a submicromolar antagonist of the neurokinin-1 receptor based on a ligand-supported homology model.…”

“…[6][7][8][9][10][11][12][13][14][15] However, the relatively low global sequence identity between the structure-unknown GPCRs and bovine rhodopsin is generally considered to be insufficient for reliable homology modeling. 6 The different conformation states of GPCRs could also be induced by various types of bound ligand.…”