Metalloproteins

[peastman]

A possible addition for a future version is to provide data for simulating metalloproteins. This would require conformations involving a metal atom in an appropriate protein-like environment. If we decide to do that, there are some questions we need to answer.

1. What metals should we include? There are a lot that can appear in proteins. Trying to include all of them would be very expensive. It's probably best to include only a few of the most common ones. This page has statistics on occurrences of metals in PDB structures. Mg is the most common by a large margin. It's followed in decreasing order by Zn, Ca, Fe, Na, Mn, and K. There's a very long tail of other elements. We already have some data on Mg, Ca, Na, and K since they appear in DES370K
2. Is our existing level of theory good enough for transition metals? This review by Mardirossian and Head-Gordon found that ωB97M-V was one of the very best functionals for transition metals. It's very similar to the ωB97M-D3BJ functional we use. They did their calculations with the def2-TZVP basis set, which is similar to but smaller than the def2-TZVPPD basis set we use. So I believe our existing level of theory should be fine.
3. How should we generate structures and conformations? The obvious choice would be to take them from PDB files. Each one would include a single metal atom and its surrounding environment. We would need to figure out how to generate that environment. It should be large enough to capture the important interactions, but small enough for the calculations to be fast, and everything needs to be capped correctly.

There are some existing databases of metal containing structures, but I haven't found one that looks usable. tmQM is a database of transition metals complexed to organic molecules. It has a few problems: it's distributed under the restrictive CC-NC license, it only has energy minimized structures, and while the environments are organic, they aren't specifically proteins. They based their structures on the Cambridge Structural Database, which itself has similar problems. It's a proprietary database that requires you to get a license, and the molecules aren't necessarily protein-like. I found several databases that index metals in PDB files: MetalPDB, InterMetalDB, MbPA, and MetalMine. But it's not clear that any of them provides much we couldn't get by just going to the PDB directly.