Tertiary protein structure distance constraints

The earliest neural network attempt for protein tertiary structure prediction was done by Bohr et al. (1990). They predicted the binary distance constraints for the C-a atoms in protein backbone using a standard three-layer back-propagation network and BIN20 sequence encoding method for 61-amino acid windows. The output layer had 33 units, three for the 3-state secondary structure prediction, and the remaining to measure the distance constraints between the central amino acid and the 30 preceding residues. 

Besides the NOE, the NMR technique offers other information that can also be used in structure determination. Spin-spin coupling constants are a traditional source of such information, and it is becoming increasingly clear that the chemical shifts in proteins and nucleic acids can often give direct information on the secondary and tertiary structure since they depend upon the local environment. This additional NMR information can be used at several levels. Coupling constant information is often converted to dihedral angle constraints and used in a manner similar to the distance constraints derived from NOE information. Because of the complexity of the interactions that influence the chemical shift, this information is more useful at the final structure refinement stage, where the overall structure has already been determined. 

The good agreement between the folded and the experimental structure is also evident from Figure (l)(center), which shows the secondary structure alignment of the native and the folded conformations. The good physical alignment of the helices illustrates the importance of hydrophobic contacts to correctly fold this protein. An independent measure to assess the quality of these contacts is to compare the C -C distances (which correspond to the NOE constraints of the NMR experiments that determine tertiary structure) in the folded structure to those of the native structure. We found that 66 % (80 %) of the C/3-C/3 distance distances agree to within one (1.5) standard deviations of the experimental resolution. 

Concepts with which to construct three-dimensional structure from sohd-state NMR data are currently being developed in many laboratories. All of these approaches aim at determining both the secondary structure (the backbone conformation) and tertiary structure (the overall fold) of proteins in an efficient manner. As in the solution state, solid-state NMR structure determination involves the calculation of families of molecular conformations that are consistent with the experimentally derived distance and/or angle constraints. The number and precision of these parameters determine the accuracy of the resulting three-dimensional structure. 

---