Skip to Main Content
The advance of high-throughput experimental technologies poses continuous challenges to computational data analysis in functional and comparative genomics studies. Gene ontology (GO) annotation and transcriptional profiling using gene expression array have been two of the major approaches for system-wide analysis of gene functions and gene interactions. In the literature, extensive studies have been reported in each aspect. Yet there is a lack of efficient algorithm that discover associative patterns across these two data domains. We proposed a mixture model associative artificial neural network to tackle this deficiency. The algorithm inherits the theoretical foundation of adaptive resonance associative map (ARAM), with essential redefinition of pattern similarity measures and learning functions. The proposed algorithm is capable of clustering data based on both GO semantic similarity and expressional correlation, for the purpose of systematically discovering genome-wide, highly correlated gene groups, which in turn suggest similar or closely related functions. We applied the proposed algorithm to the analysis of the Saccharomyces cerevisiae (yeast) dataset and obtained satisfactory results.