An introduction is provided to three papers which compare corresponding protein crystal and NMR solution structures determined by the Joint Center for Structural Genomics (JCSG). Special mention is made of the JCSG strategy for combined use of the two techniques, and of potential applications of the concept of 'reference crystal structures', which is introduced in the following three papers.The NMR Core of the Joint Center for Structural Genomics (JCSG) has devoted a large part of its work to efficient high-quality NMR structure determination of small soluble proteins based on the recording of extensive networks of nuclear Overhauser effect (NOE) upper limit distance constraints. This effort is an alternative to other projects pursued under the auspices of the Protein Structure Initiative (PSI; see, for example, Cornilescu et al., 2007;Liu et al., 2005) and is complementary to PSI projects that are focused on obtaining NMR structures of proteins from a minimal amount of experimental data In the context of validating the results of this new approach against high-quality crystal data (Brown & Ramaswamy, 2007), a series of NMR structure determinations were performed for proteins for which a high-resolution crystal structure had previously been determined by the JCSG. In the following three papers, we present comparisons of the crystal and NMR structures for a selection of five of these proteins. Thereby, once it had been established that the two methods yielded near-identical global molecular architectures, we further investigated possible complementarities of the results from the two techniques.Over the years, much effort by many different groups has been devoted to deriving the behavior of protein molecules in solution or other physiological environments from crystallographic data. Examples include the representation of crystal structures by a bundle of conformers (DePristo et al., 2004), computational prediction based on comparison of NMR and X-ray data (Yang et al., 2007), combination of multiple crystallographic data sets collected at ambient temperature with and without bound ligands (Fraser et al., 2009) and supplementing crystal structures with NMR measurements of the frequencies of dynamic processes (Boehr et al., 2010).Here, the individual crystal structures were solved by the JCSG at 100 K to about 1.8 Å resolution, whereas the corresponding NMR structures were determined in solution at ambient temperature. Despite the large differences in experimental conditions, the NMR structures could be superimposed with the crystal structures with r.m.s.d. values of <1.0 Å for the backbone heavy atoms. This provided