Integration of mutual information and text mining methods for extracting gene-gene interactions from gene expression data

Mutual information algorithms have been used for the identification of gene-gene interactions in gene expression data. These methods have been hindered by a high false-positive rate. We explored the use of free-text abstracts as an additional source of information for assessing the biological relevance of predicted gene interactions. Our results suggest that the performance of a mutual information algorithm on this task can be enhanced by using text mining methods to refine the initial set of predictions.