The scope of TNT is to optimize atomic coordinates with respect to a series of observations, usually diffraction data and ideal stereochemistry. It also will provide information to aid in the examination and correction of the refined model. These include the difference maps to be displayed along with the model, the automatic location of peaks in these maps, and indicators of the problem spots in the refined model that should be examined directly by the crystallographer. The principal advantages of TNT are listed in Table i.
While TNT is an excellent package for everyday refinement, its flexibility allows it to be used as a tool for the development of new refinement techniques. One can reorder TNT's components, add new components to implement the refinement of the model against a novel set of observations, or make other kinds of changes without having to modify the existing programs. Many of the current features of TNT were first tested in this fashion, which reduced the time invested in an idea before our determining its usefulness.
The package has been used in this fashion a number of times. Chapman (1995) has implemented novel methods for low-resolution real-space refinement, and Bricogne & Irwin (1996) have used very sophisticated protocols to interface TNT and their maximum-likelihood refinement program BUSTER. Neither of the projects, which implement ideas far from the thoughts of TNT's creators, required significant modifications to TNT's code.
Other investigators have implemented their ideas by modifying parts of TNT itself [Abrahams, 1996, Pannu & Read, 1996]. In these cases it was decided that the computational efficiency of adding new code directly to TNT justified the extra time required to learn the internal workings of TNT programs.