Beta turn is an important element of protein structure which plays an important role in protein folding and stability and molecular recognition. In the past three decades, numerous beta turn prediction methods have been developed based on various strategies. At present, it is difficult to say which method is better than the other. It is because of the fact that these methods have been developed on different data sets. Thus, it is important to evaluate the performance of new beta turn prediction algorithms and compare the performance with the already existing methods. The assessment of performance of a method will be important, both for users as well as developers, since it allows one to find the best method for their work and the developer to improve the method.|
The aim of BTEVAL server is to evaluate beta turn prediction algorithms on a uniform data set of 426 proteins or subsets of these proteins. It is the new data set in which no two protein chains have more that 25% sequence identity and each chain contains minimum one beta turn. The PROMOTIF program has been used to assign beta turns in these proteins. The performance of a method can be evaluated in terms of 4 different measures: Qtotal(or prediction accuracy), is the percentage of correctly classified residues; Qpred(or probability of correct prediction) is the percentage of correct prediction of beta turns; Qobs (or percentage coverage) is the percentage of observed beta turns that are correctly predicted and MCC which accounts for both over and under-predictions. The performance of a method can be compared with the existing statistical algorithms such as Chou-Fasman algorithm, Thornton's algorithm, 1-4 & 2-3 correlation model, Sequence coupled model and Gorbturn and neural network based method BTPRED. Thus, the server provides a benchmarking of different beta turn prediction methods.