Molecular modelling in pharmacology: selected examples

Computational chemistry and biology are helpful in understanding protein structure and the relationships between structure and biological activity. In particular, to develop a new drug, medicinal chemists and pharmacologists are interested in understanding and predict drug action at a molecular level, especially if the action of the drug is unknown or poorly understood. In these cases, the molecular modelling should reduce some of the work in the development of drug compounds. Here, we present two examples of homology modelling applied to pharmacology that obtained a great success. Phosphatidylcholine-sterol acyltransferase (LCAT) is a glycoprotein of 416 residues, synthesized by the liver and secreted in plasma. It catalyzes the transacylation of the sn 122 fatty acid of lecithin to the free 3 12OH group of cholesterol, generating cholesterol esters and lysolecithin. LCAT shares the Ser/Asp 12Glu/His triad with lipases, esterases and proteases, but the low level of overall sequence homology between LCAT and these enzymes makes standard modelling procedures unsuitable. For this reason, to build an LCAT model, we implemented a combined approach that included folding recognition, secondary structure prediction, and \u2018chimeric\u2019 homology modeling. In detail, the ab initio model was used as scaffold to merge the two best homology templates identified, Paucimonas lemoignei depolymerase for the N-terminus and Candida antarctica lipase A for the C-terminus. In this way, we built an accurate LCAT structure with a well-defined binding site. Then, we performed a high-throughput virtual screening exploration of LCAT pocket with a large database of chemical compounds. The best compounds identified during the HTS were tested in vitro and demonstrated the ability to modulate LCAT activity. G-protein coupled receptors (GPCRs) responding to signalling molecules are key transducers in cell-to-cell communication. Malfunctioning of GPCRs invariably leads to disease conditions; for this reason, they represent the target of more than 70% of currently marketed drugs. The rational design of new (ant)agonists targeting GPCRs strictly depends on the resolution of their atomistic structure and their appropriate in silico molecular modeling. However, the low homology degree of GPCRs with the available crystallized templates still represents a serious limitation. For several years, bovine rhodopsin (bRh) has been the only available high-resolution crystal structure of a GPCR, thus representing the unique possible template. The recent publication of new GPCR structures (human A2A adenosine receptor, human \u3b22-adrenergic, turkey \u3b21-adrenergic receptor, squid rhodopsin, human dopamine D3 receptor, human CXC chemokine type 4 receptor) has allowed the construction of more accurate and predictive models of GPCRs, which has represented a significant advancement toward a more rational drug design. A molecular modeling approach based on multiple templates has been applied to GPR17, a previously orphan GPCR that has recently emerged as a promising therapeutic target to foster recovery in diseases characterized by dysfunction of myelin, the oligodendroglial sheath that, by wrapping nerve terminals, ensures impulse transmission and communication between neurons. The obtained \u201cchimeric\u201d model of GPR17 has been then submitted to an efficient pipeline including ii) a high-throughput virtual screening with more than 130,000 lead-like compounds, and ii) a wet pharmacological validation of the top scoring chemical structures. This integrate strategy allowed us to successfully identify 5 agonists or partial agonists that had never been expected a priori to act on a GPCR, and behaved as extremely more potent ligands than GPR17 endogenous activators. Considering the relevance of the role proposed for GPR17 in demyelinating diseases, this new 5 ligands may represent an advancement toward the design of new pharmacological approaches aimed at restoring the myelin sheath integrity