Beyond similarity-based methods to associate genes for the inference of function

Abstract

The function(s) of a novel gene or gene product can be inferred by associating the gene or gene product with those whose functions are known. It is now common practice to associate two genes if they have similar sequences. In recent years, computational methods have been developed that associate genes on the basis of features beyond similarity, using a variety of biological data beyond single-gene sequences. This review describes several promising methods that associate genes or gene products. These associative methods employ similarity of sequences and structures, features from whole-genome analysis, co-expression patterns from microarray and EST data, interacting properties from proteomic data, and links from literature mining. Finally, we outline issues surrounding the validation and integration of these methods.