Measuring protein structures through NMR spectroscopy is a complicated task. Several problems arise on the way from NMR experiments to the full determination of the 3D coordinates of a structure. Different types of data have to be measured and analyzed. The interpretation of these data is mostly done by human experts.

The general structure solving process may be divided into three steps:

• Step 1: Build a first model. (several measurements)
• Step 2: Refine this model. (expert knowledge, molecular modeling, possibly additional measurements)
• Step 3: Validate or invalidate the model. (specific experiments)

One goal of this project is to achieve a considerable speed-up in the NMR structure solving process. Therefore computational methods are developed which allow the creation of models for a protein structure at an experimental stage in which this is not possible by hand. These models may then be refined and tested for correctness using the "standard" way. The time saving results from the need for fewer experiments in Step 1.

The construction of a consistent protein chemical shift database is an important step toward making more extensive use of this data in structural studies. Unfortunately, progress in this direction has been hampered by the quality of the available data, particularly with respect to chemical shift referencing, which is often either inaccurate or inconsistently annotated. Preprocessing of the data is therefore required to detect and correct referencing errors.

We have developed a program for performing this task, based on the comparison of reported and expected chemical shift distributions. This program, named CheckShift, does not require additional data and is therefore applicable to data sets where structures are not available. Therefore CheckShift provides the possibility to re-reference chemical shifts prior to their use as structural constraints.