Computationally efficient approach for the identification of ice-binding surfaces and how they bind ice.

Recognition and binding of ice by proteins, crystals, and other surfaces is key for their control of the nucleation and growth of ice. Docking is the state-of-the-art computational method to identify ice-binding surfaces (IBS). However, docking methods require a priori knowledge of the ice plane to which the molecules bind and either neglect the competition of ice and water for the IBS or are computationally expensive. Here we present and validate a robust methodology for the identification of the IBS of molecules and crystals that is easy to implement and a hundred times computationally more efficient than the most advanced ice-docking approaches. The methodology is based on biased sampling with an order parameter that drives the formation of ice. We validate the method using all-atom and coarse-grained models of organic crystals and proteins. To our knowledge, this approach is the first to simultaneously identify the ice-binding surface as well as the plane of ice to which it binds, without the use of structure search algorithms. We show that biased simulations even identify surfaces that are too small or too weak to heterogeneously nucleate ice. The biasing simulations can be used to identify of IBS of antifreeze and ice nucleating proteins and to equilibrate ice seeds bound to an IBS for the calculation of heterogeneous ice nucleation rates using classical nucleation theory.