Computational treatment of metalloproteins

Metalloproteins present a considerable challenge for modeling, especially when the starting point is far from thermodynamic equilibrium. Examples include formidable problems such as metalloprotein folding and structure prediction upon metal addition, removal, or even just replacement; metalloenzyme design, where stabilization of a transition state of the catalyzed reaction in the specific binding pocket around the metal needs to be achieved; docking to metal-containing sites and design of metalloenzyme inhibitors. Even more conservative computations, such as elucidations of the mechanisms and energetics of the reaction catalyzed by natural metalloenzymes, are often nontrivial. The reason is the vast span of time and length scales over which these proteins operate, and thus the resultant difficulties in estimating their energies and free energies. It is required to perform extensive sampling, properly treat the electronic structure of the bound metal or metals, and seamlessly merge the required techniques to assess energies and entropies, or their changes, for the entire system. Additionally, the machinery needs to be computationally affordable. Although a great advancement has been made over the years, including some of the seminal works resulting in the 2013 Nobel Prize in chemistry, many aforementioned exciting applications remain far from reach. We review the methodology on the forefront of the field, including several promising methods developed in our lab that bring us closer to the desired modern goals. We further highlight their performance by a few examples of applications.