Kernel Methods for Protein Function Prediction

1. INTRODUCTION From a computer scientist's point of view, protein function prediction can be regarded as a classification problem, namely to correctly classify a newly discovered protein into its functional class. Besides, this computational function prediction often means dealing with different data formats, data types and data sets with missing entries. Kernel methods are a popular method from Machine Learning which can handle all these problems of classification, data compatibility, data integration, and data completion. Numerous applications of kernel methods to protein function prediction tasks have been published over recent years. In the following, we will review applications, potential and limits of kernel methods in this field.