Figure 1

The four types of measures: In this example the evaluation starts with a model (right) and the correct structure of a protein (left). The left part of the model is quite good, but a large loop is inserted which results in a shift in the alignment in the central part of the model. Some residues in the model are not aligned to the template. This creates a shift that results in the right part of the model being correctly aligned again. A global measure (A) would use the complete model and compare it with the complete correct structure and probably not score this model very well. As shown in the (B) an alignment dependent measure would only consider the first and last fragments as correct. Measures that use an alignment independent approach (C) first do a structural alignment and then and the most significant fragment. The shifted residues in the center of the model would be included in the evaluation. In template based measures (D) the template used to build the model is used for comparison. In our example the template extends to the right of the model and it also has one loop that is longer so that there is a gap in the model that is not shown. The correct structure is superimposed on the template and the resulting alignment is compared to the alignment of the model.