Crystallized proteins retain the ability to move around, thus making interpretations of diffraction data an imprecise process. Crystallographers fit their data to models that pass quality controls, but they usually report only one such model. The Cambridge group generated alternate models that fit the data for several proteins. “We found a reasonable number of structures that are surprisingly different in their finer details,” says DePristo. And as the diffraction resolution decreased, the differences increased.
Most variability was found at the protein surface rather than its core, suggesting that a good idea of protein fold can be gleaned even at low resolution. But detailed conclusions that depend on precise atomic location, such as catalytic mechanism, may be misinterpretations. “We need a change in thinking of structures as less of a static, perfect model, but rather as models that have uncertainties,” says Tom Terwilliger (Los Alamos National Laboratory, Los Alamos, NM). “Crystallographers need to develop a means for communicating the uncertainty in their atomic model.” ▪