CASP experiment

Currently, methods of structure prediction from sequence perform poorly. The results of the biannual CASP experiment (Critical Assessment of techniques for protein Structure Prediction see http //predictioncenter.llnl. gov/) are equivocal to say the least. A recent report on the improvements of aligning target sequences to a structural template (67) indicates that over the last four CASP competitions there was no significant improvement in quality in this key step in the prediction process. Alignment remains the major source of error in all models based on less than 30% se-... 

The standard measure for accuracy is Q3 defined as the number of residues correctly predicted divided by the number of residues. Q3 can also be computed for the residues of individual proteins Q3P or of individual secondary structures QH, QE, and QC. Other methods use certain types of correlation coefficients or take the type and position of secondary structure elements into account (proposed by Rost [168] and further developed for the CASP experiments [169]). One example is the SOV (segment overlap) score, i.e. the average over all secondary structure segments of the ratio of the overlap of predicted and true secondary segment and the overall length of the two segments. 

Other successful structure prediction methods based on sequence alone are FASTA [15, 18], hidden Markov models (HMMs) [25, 26, 28, 179], intermediate sequence search [182], and iterative profile search [29]. Sequence analysis methods are discussed in detail in Chapter 2. The results of such methods in the CASP experiment are described in [157, 183, 184],... 

State-of-the-art methods for protein structure prediction are continuously discussed and evaluated in the bi-annual CASP experiments (see the accompanying special issues of the Journal Proteins [329]), the LiveBench [330] and EVA [331] experiments. The major available methods, programs and servers are collected in Tables 6.2 and 6.3. 

A blind prediction is a computer-based construction of 3D models of a protein sequence for which no structure is known at the time. CASP (comparative assessment of structure prediction methods [109, 274-276]) is a worldwide contest of protein structure prediction that takes place every two years. During the CASP experiments, a set of automated numerical evaluation tools have been implemented [110, 111, 277, 278] to cope with the large number of predictions in a way that is as objective as possible. In fact, the experiment is also devoted to the research and development of such unbiased methods. However, there is still quite some controversy on the criteria to judge protein structure predictions and the corresponding models [339]. 

The core fragment threading approach, using carefully defined structural core models, a calibrated scoring function, and a statistical estimation of P-values for the scores, appears to be one of the most successful methods in the CASP-3 experiments [196, 197]. 

Theoretical predictions are risky. Therefore for almost all such prediction experimental validation is required. Nevertheless, often the models can indicate appropriate ways for validation or further experiments. These experiments can be expected to be time-consuming, and expensive. Furthermore, the protein actually needs to be available for the suggested experiments. All of this limits the applicability of experimental validation. Therefore, it is mandatory to reduce errors as much as possible and to indicate the expected error range via computer-based predictions. This is not a trivial problem for structure prediction, though. An estimation of the performance and accuracy of the respective methods can be obtained from large scale comparative benchmarking, from successful blind predictions and from a community wide assessment experiment (CASP [109, 229]/ CAFASP [283]). These are addressed in turn in the following ... 

B. Kramer, M. Rarey, and T. Lengauer. Casp-2 experiences with docking flexible ligands using flexx. PROTEINS ... 

---