ABSTRACT
High resolution liquid-state NMR emerged as a structure determination technique for biological macromolecules in 1985. From the beginning, molecular modeling has had a central place in the derivation of NMR solution structures [1-4]. There are several reasons for this. First, the energy parameters, typically derived from a molecular dynamics or molecular mechanics force field, play a central role in calculating and refining the structure. This is because experimental data are scarce, being available for only a fraction of the atoms (mostly the hydrogens). An additional difficulty is that most of the data describe relative positions of atoms and do not directly correspond to the global structure of the molecule. Second, models are not built manually but are automatically calculated by appropriate algorithms. In this way the conformational space consistent with the data is sampled randomly to test whether the data determine the structure uniquely. Consequently, a lot of effort has gone into the development of algorithms to fit the experimental data. The methods used for NMR structure calculations are usually adapted from algorithms originally developed for different purposes in molecular modeling. Third, the wealth of dynamic information obtained by NMR and the difficulties in interpreting it in structural terms have led to a close interaction with MD simulation [5-9].
