Systematic Errors and Biases

This style of refinement also reflected a change of emphasis with respect to the computational balance between initial structure generation and restrained MD. Previously, most workers had spent much time refining structures with respect to distance restraints before applying MD. With this modified force field, it was practical to start MD refinement with crude metric matrix structures after only substructure embedding, or even simply from extended strand initial structures. 

A last feature of this kind of refinement is that even if it is technically using an MD algorithm, a structure refined with it will not have the energetic properties of a structure refined in a full force field. For this reason, it is probably advisable to take structures from the simplified force field and subject them to at least brief simulation in a complete force field. 

Although procedures such as the conversion from NMR spectra to distances might be a likely cause of systematic errors, there is also the possibility that certain computational methods might introduce systematic errors. 

Some investigators would argue that the results of distance geometry are only initial structures and that biases should disappear after thorough MD or other energetic refinement. In this case, one should bear in mind that one may then suffer from biases in the molecular mechanics force field. The most serious of these is the tendency for structures to become too compact if simulations are conducted in vacuo. The obvious solution here is to perform simulations in the presence of explicit solvent, although this is computationally expensive. 

---

The third chapter maintains the theme of determining conformations by describing how to generate initial structures of organic and bioorganic molecules and how to model experimental NMR data. Andrew Torda and Wilfred van Gunsteren also discuss refinement methods, force fields, systematic errors and biases, and the quality of predicted structures. 

---