Coarse-grained modeling for molecular discovery Applications to cardiolipin-selectivity

The development of novel materials is pivotal for addressing global challenges such as achieving sustainability, technological progress, and advancements in medical technology. Traditionally, developing or designing new molecules was a resource-intensive endeavor, often reliant on serendipity. Given the vast space of chemically feasible drug-like molecules, estimated between 10⁶ - 10¹⁰⁰ compounds, traditional in vitro techniques fall short.
Consequently, in silico tools such as virtual screening and molecular modeling have gained increasing recognition. However, the computational cost and the limited precision of the utilized molecular models still limit computational molecular design.
This thesis aimed to enhance the molecular design process by integrating multiscale modeling and free energy calculations. Employing a coarse-grained model allowed us to efficiently traverse a significant portion of chemical space and reduce the sampling time required by molecular dynamics simulations. The physics-informed nature of the applied Martini force field and its level of retained structural detail make the model a suitable starting point for the focused learning of molecular properties.
We applied our proposed approach to a cardiolipin bilayer, posing a relevant and challenging problem and facilitating reasonable comparison to experimental measurements.
We identified promising molecules with defined properties within the resolution limit of a coarse-grained representation. Furthermore, we were able to bridge the gap from in silico predictions to in vitro and in vivo experiments, supporting the validity of the theoretical concept. The findings underscore the potential of multiscale modeling and free-energy calculations in enhancing molecular discovery and design and offer a promising direction for future research.