Molecular binding is a major research topic that has undoubtedly benefited of recent technological innovation, especially in the area of the so-called computational sciences. However, the predictive power of computations remains low mainly due to the poor correlation of the in-silico models with the real world. A clear example is drug discovery where the drug in vivo efficacy is seen correlated to the ligand residence time, which is hardly predictable by current computational methods. The present proposal tackles the challenge aiming at reshaping the border of the state-of-the-art simulations in molecular binding. I outline a research program that realises a vision where drug design is entrusted to ligand binding affinity and kinetics prediction, and molecular binding interactions are simulated in a realistic plasma membrane model. To achieve the ambitious goals of the research, my team will develop and apply cutting-edge computational techniques based on free-energy calculations, machine learning and multiscale molecular dynamics simulations. Evidence of the innovative nature of the developed approaches will be given by elucidating fundamental aspects of the functional mechanism of the G-protein coupled receptors (GPCRs), a pharmacologically prominent membrane protein family targeted by ~ 40% of marketed drugs. We will achieve a thorough characterization of the binding thermodynamics and kinetics of signal molecules (antagonists and agonists) that will be used by an original machine learning model to identify novel receptor antagonists with prescribed binding affinity and residence time. We will then investigate the receptor conformational transition from the inactive to the active state and develop an ad hoc multiscale approach to characterize the formation of GPCRs dimers, oligomers and clusters in cell membrane and their interaction with the G-protein that activates the signal transduction. Experiments will be performed to validate all the in-silico results.