Periodic Reporting for period 2 - CoMMBi (Computational Microscope on Molecular Binding: from atom to cell membrane scale)
Periodo di rendicontazione: 2023-03-01 al 2024-08-31
The present project aims to develop computational protocols for studying molecular binding interactions under near-real conditions. Currently, most in silico protein-ligand and protein-protein binding studies are hampered by approximations that oversimplify the molecular binding process. Our goal is to overcome the limitations of state-of-the-art techniques by employing atomistic and coarse-grained simulations, free-energy calculations, and machine learning. Our research focuses on the functional mechanisms of G protein-coupled receptors (GPCRs), beginning with adenosine receptors. The information extracted from our studies contributes to elucidating the complex functionality of these receptors and provides a structural basis for drug design. In addition to clarifying the thermodynamic properties of ligands, we also integrate binding kinetic data into the drug design protocol, enhancing its accuracy and success rate. Given the pharmacological relevance of GPCRs, which are targeted by approximately 30% of marketed drugs, our study could have a significant social and economic impact on the development of future therapeutics.
We have developed computational protocols to study the thermodynamics and kinetics of ligand binding to GPCRs. Additionally, we have investigated the long timescale functional dynamics underpinning receptor activation using a dimensionality reduction approach. This method accounts for all the major conformational microswitches that characterize the transition of the receptor from the inactive to the active form. We have also established multiscale simulation protocols to elucidate the dimerization mechanism of GPCRs within the membrane and to investigate how receptor assembly affects its activity.
These studies are integrated with machine learning techniques using graph neural networks to develop more accurate models that better reproduce receptor functionality, thereby improving the drug design success rate.
Four major results have been achieved so far:
1. Identification of the dimer structures of chemokine receptors CCR5 and CXCR4;
2. Discovery of a pre-active state of the adenosine receptor A2A;
3. Characterization of the ligand binding modes and kinetics for adenosine A2A agonists, antagonists, and inverse agonists;
4. Development of graph neural networks that process chemical structures and system properties to enhance sampling capability and accuracy.
These studies are integrated with machine learning techniques using graph neural networks to develop more accurate models that better reproduce receptor functionality, thereby improving the drug design success rate.
Four major results have been achieved so far:
1. Identification of the dimer structures of chemokine receptors CCR5 and CXCR4;
2. Discovery of a pre-active state of the adenosine receptor A2A;
3. Characterization of the ligand binding modes and kinetics for adenosine A2A agonists, antagonists, and inverse agonists;
4. Development of graph neural networks that process chemical structures and system properties to enhance sampling capability and accuracy.
The methods developed in our project address the limitations of state-of-the-art techniques. Our simulations enable the investigation of drug targets under the most realistic conditions, passing from detailed atomistic descriptions to multiscale simulations while preserving the system's correct energetics. Our methods are accurate, reliable, and user-friendly, allowing researchers to achieve millisecond-to-second timescales within an affordable computing time. This capability enables the generation of in silico models that can be faithfully compared with experimental data.
Our objective is to further increase the complexity of our models, pushing the boundaries of molecular simulations toward more realistic yet accurate representations. In this framework, which we refer to as a "computational microscope," drug binding to molecular targets can be investigated, including effects such as molecular crowding, protein dimerization, and interactions with environmental components like membrane phospholipids and cholesterol.
Our simulations could be instrumental in understanding the pathogenesis of certain diseases and clarifying molecular-level aspects that are challenging to capture using standard experimental techniques. By employing our computational microscope, we can test drug candidates before reaching the experimental stage, significantly reducing the time and costs associated with research.
Our objective is to further increase the complexity of our models, pushing the boundaries of molecular simulations toward more realistic yet accurate representations. In this framework, which we refer to as a "computational microscope," drug binding to molecular targets can be investigated, including effects such as molecular crowding, protein dimerization, and interactions with environmental components like membrane phospholipids and cholesterol.
Our simulations could be instrumental in understanding the pathogenesis of certain diseases and clarifying molecular-level aspects that are challenging to capture using standard experimental techniques. By employing our computational microscope, we can test drug candidates before reaching the experimental stage, significantly reducing the time and costs associated with research.