Int J Data Min Bioinform. 2015;13(1):50-62. doi: 10.1504/ijdmb.2015.070840.
International journal of data mining and bioinformatics
Ying Shen, Lin Zhang
PMID: 26529907 DOI: 10.1504/ijdmb.2015.070840
Gene function prediction is an important problem in bioinformatics. Due to the inherent noise existing in the gene expression data, the attempt to improve the prediction accuracy resorting to new classification techniques is limited. With the emergence of Gene Ontology (GO), extra knowledge about the gene products can be extracted from GO and facilitates solving the gene function prediction problem. In this paper, we propose a new method which utilises GO information to improve the classifiers' performance in gene function prediction. Specifically, our method learns a distance metric under the supervision of the GO knowledge using the distance learning technique. Compared with the traditional distance metrics, the learned one produces a better performance and consequently classification accuracy can be improved. The effectiveness of our proposed method has been corroborated by the extensive experimental results.
