The rational design of medicinal compounds has been an essential component of drug discovery for many years. Computational scientists contribute to this process in many different ways, e.g. through data-driven approaches exploiting the ‘knowledge base’ methods in compound design. Another avenue is to use physics-based approaches to gain deeper understanding of phenomena driving properties of compounds, and ultimately predict those based on theoretical calculations. Even though such predictions remain extremely challenging, the conceptual understanding of underlying driving forces focuses one’s attention in the right direction, helps design new experiments and reveals fundamental rules about a system’s behavior. Molecular dynamics simulations strive to provide better understanding of dynamic behavior of proteins, in particular the way they interact with the ligands and with its aqueous environment. The solvation structure analysis resulting from simulations has been used in our group extensively to rationalize small molecule activity, selectivity, and binding kinetics against a variety of soluble protein targets. A large class of protein targets that are embedded in membranes (e.g., GPCRs, ion channels) has not yet been satisfactorily described due to the lack of high-quality membrane models. Lipid membranes by themselves pose many challenges to our comprehension, and even seemingly simple phenomena such as partitioning of small molecules in lipid bilayers or small molecule transfer across the membranes remain poorly understood. It would be of great interest to develop a better description of membranes, and their interactions with small molecules and membrane proteins, in order to advance our understanding of binding and distribution of medicinal compounds.