Prediction of Contact Maps Using Support Vector Machines
BIBE '03 Proceedings of the 3rd IEEE Symposium on BioInformatics and BioEngineering
Striped sheets and protein contact prediction
Bioinformatics
PROFcon: novel prediction of long-range contacts
Bioinformatics
| Hi-index | 0.00 |
In structural Bioinformatics a challenge is still to predict structure from sequence and one promising way to solve this problem is the prediction of protein contact maps. Several step forwards have been described in the last years; however still the prediction is not enough accurate to allow a reliable reconstruction of the 3D structure from the contact map and improvements on previous result are required. Here we focus on the prediction of a specific structural class of proteins, the so-called alpha proteins since specifically for these proteins the prediction accuracy of the contact maps is still lower than those of the other classes. In this paper we address the prediction of the helical residue contacts in all-alpha proteins in order to improve the existing methods on this specific task. Using a non redundant set of 300 all-alpha proteins we achieve an accuracy of 22% and a coverage of 19%, a result that overpasses by 7 percentage points those obtained by the best performing contact map predictors, presently available.