Prediction of post-translational phosphorylation of proteins using multiple kernel learning(Biometrics1)

概要

論文の詳細を見る
Phosphorylation is one of the most important post-translational modifications which regulate the activity of protein. The problem of predicting phosphorylated sites is the first step of understanding various biological processes that initiate the actual function of proteins in each pathway. In the problem of predicting phosphorylated sites, the bottleneck in prediction accuracy is a way of extracting useful features from the protein sequence. Since there is no optimal feature extraction method minimizing information loss, we should choose one of available methods carefully. Using multiple features rather than one can improve the prediction accuracy. Some kinds of heuristics in bioinformatics have been widely used in order to combine multiple features, including majority voting and single concatenated feature vector. We propose an optimal way of integrating multiple features in the framework of multiple kernel learning. Our numerical experiments confirm that our new feature extraction methods with multiple kernel learning are very useful for predicting phosphorylated sites.
社団法人電子情報通信学会の論文
2006-11-17