In silico guided identification of molecular determinants for modulation of GPCRs

Drug discovery aims to identify molecules, also called ligands, that can be used to manipulate the behaviour of proteins for a beneficial therapeutic effect. If this process is successful, the result is a drug or medicine. Due to the inherent complexity of biology underlying many diseases, drug development is lengthy, challenging, and significantly expensive, and, from the outset, has very low chances of success (on average, <10% of drugs tested in humans will become approved drugs). For the greatest chance of success, drug discovery requires input from multiple scientific (and non-scientific) disciplines. The field of computational chemistry aims to facilitate the drug discovery process by means of modelling which tries to to predict, among others, how a potential medicine should look and behave to artificially modify the behavior of the malfunctioning protein. Like with any forms of prediction, the degree of confidence in the outcome highly depends on how informed that prediction was. In other words, the more information we have, for example in the form of previous known ligands, the higher the chances of identifying patterns in it, and therefore the more accurate the estimation can be. In the particular field that concerns this manuscript, such information includes: how medicines look, what properties they have, how they interact with such proteins, and how they trigger a response. This way, ligands are identified and can be modified to transform them into molecules that can be used as medicines. In particular, this manuscript will emphasize and focus on those cases where preceeding information to inform predictive models is scarce. The human body contains over 20,000 unique proteins. Each have specific functions that, when impaired, may lead to the development of diseases. Importantly, individual cells need to act in coordination with other (often neighbouring) cells to carry out vital functions, and in order to do that they need to be stimulated to do so from the outside. The transmission of these key messages is often done through G-protein coupled receptors. This is set of over 500 different proteins that are found in the membrane of cells and act as communication channels. They have a wide variety of functions depending on the messages they transmit: from enabling many of our senses, to messaging between neurons, as well as playing a key role in immune system. The primary aim of this thesis is to combine different computational chemistry techniques for GPCRs, focusing on those where key protein and ligand information are scarce, with the following objectives: Develop three-dimensional models to predict how these proteins and their ligands interact. Identify the mechanisms that trigger the modulation of these proteins by their ligands. Computationally discover and design new ligands that interact with the proteins of interest. Chapter 2 highlights different ways models can be constructed and applied to predict three dimensional structures of chemokine receptors that have not been obtained experimentally. Chapter 2 also deals with how different inputs of information can be combined to inform the design of novel ligands. Chapter 3 focuses on a particularly challenging chemokine receptor, known as Atypical Chemokine Receptor 3, for which ligand data is scarce and no 3D structure has been reported. Chapter 4 explores the application of molecular modelling to guide pharmacology experiments to understand and provide detailed insights into the molecular interaction networks responsible for the activity of an even more challenging family of GPCRs, adhesion GPCRs. In the final chapter, Chapter 5, a thorough discussion of the research described in this thesis is provided, alongside an overview of the future perspectives and potential pitfalls regarding the use of computational methods on challenging computer-aided drug discovery projects.