Towards predicting coiled-coil protein interactions
RECOMB '01 Proceedings of the fifth annual international conference on Computational biology
A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
| Hi-index | 0.00 |
Peptide-based vaccines, in which small peptidesderived from target proteins (epitopes) are used toprovoke an immune reaction, have attracted considerable attention recently as a potential means both of treating infectious diseases and promoting thedestruction of cancerous cells by a patient's ownimmune system. With the availability of large sequence databases and computers fast enough for rapidprocessing of large numbers of peptides, computeraided design of peptide-based vaccines has emerged asa promising approach to screening among billions ofpossible immune-active peptides to find those likely toprovoke an immune response to a particular cell type.In this paper, we describe the development of threenovel classes of methods for the prediction of classI epitopes. Each one of the three classes of methodsgives a specific set of insights into the epitope prediction problem. We present a quadratic programmingapproach that can be trained on quantitative as well asqualitative data. The second method uses linear programming to counteract the fact that our training datacontains mostly positive examples. The third class ofmethods uses sequence profiles obtained by clusteringknown epitopes to score candidate peptides. By integrating these methods, using a simple voting heuristic,we achieve improved accuracy over the state of the art.