Current MD forcefields fail to capture key features of protein structure and fluctuations: A case study of cyclophilin A and T4 lysozyme

Globular proteins undergo thermal fluctuations in solution, while maintaining an overall well-defined folded structure. In particular, studies have shown that the core structure of globular proteins differs in small, but significant ways when they are solved by x-ray crystallography versus solution-based NMR spectroscopy. Given these discrepancies, it is unclear whether molecular dynamics (MD) simulations can accurately recapitulate protein conformations. We therefore perform extensive MD simulations across multiple force fields and sampling techniques to investigate the degree to which computer simulations can capture the ensemble of conformations observed in experiments. By analyzing fluctuations in the atomic coordinates and core packing, we show that conformations sampled in MD simulations both move away from and sample a larger conformational space than the ensemble of structures observed in NMR experiments. However, we find that adding inter-residue distance restraints that match those obtained via Nuclear Overhauser Effect measurements enables the MD simulations to sample more NMR-like conformations, though significant differences between the core packing features in restrained MD and the NMR ensemble remain. Given that the protein structures obtained from the MD simulations possess smaller and less dense protein cores compared to those solved by NMR, we suggest that future improvements to MD forcefields should aim to increase the packing of hydrophobic residues in protein cores.