Knowledge-based analysis of microarray gene expression data by using support vector machines

We introduce a method of functionally classifying genes by using gene expre ssion data from DNA microarray hybridization experiments. The method is bas ed on the theory of support vector machines (SVMs). SVMs are considered a s upervised computer learning method because they exploit prior knowledge of gene function to identify unknown genes of similar function from expression data. SVMs avoid several problems associated with unsupervised clustering methods, such as hierarchical clustering and self-organizing maps. SVMs hav e many mathematical features that make them attractive for gene expression analysis, including their flexibility in choosing a similarity function, sp arseness of solution when dealing with large data sets, the ability to hand le large feature spaces, and the ability to identify outliers. We test seve ral SVMs that use different similarity metrics, as well as some other super vised learning methods, and find that the SVMs best identify sets of genes with a common function using expression data. Finally, we use SVMs to predi ct functional roles for uncharacterized yeast ORFs based on their expressio n data.