투고논문

Conventional structure determination of proteins using NMR data is an iterative process where NOE assignments and structure calculation are tightly coupled.¹ Steady advances in algorithm development now permit automatic calculation of 3D structures, provided chemical shifts are assigned for most atoms and the NOESY data quality is sufficient to obtain structural restraints.² For performing a search of conformational spaces using a suitable algorithmic or software method, these methods must meet the experimental restraints as well as force field that defines the physical energies between atoms. Traditional software for NMR structure calculation use simplified force fields compared to atomistic molecular dynamics (MD) simulation to enhance the search efficiency of conformational space.^3-5 For instance, Lennard-Jones potential and solvation energies are not included. The simplified force fields permit fast calculations by higher temperature annealing. However, the geometries of the regions lacking structural restraints often diverge and are inaccurate. On the other hand, atomistic MD calculations are suitable for characterizing the regions where experimental restraints are insufficient. Most atomistic MD-driven calculations for NMR structure refinement approximate the solvation effects by using generalized Born implicit solvent (GBIS) model⁶ because of the very long computational times required for calculating interaction energies with explicit solvents. Thus, GBIS attains the right balance between the computational times and structural accuracy. We have reported extensively on the advantages of using GBIS in refining protein, protein–protein complex, and membrane protein structures.