In recent years, a number of machine learning approaches to literature-based gene function annotation have been proposed. However, due to issues such as lack of labeled data, class imbalance and computational cost, they have usually been unable to surpass simpler approaches based on stringmatching. In this paper, we propose a principled machine learning approach based on kernel classifiers. We show that kernels can address the task's inherent data scarcity by embedding additional knowledge and

more »

... e propose a simple yet effective solution to deal with class imbalance. From experiments on the TREC Genomics Track data, our approach achieves better F1-score than two state-of-the-art approaches based on string-matching and cross-species information.