GeneNetwork analysis of each trait resulted in the generation of over 3500 sets of genes that are associated with trait values either via genetic correlation of gene expression in one of the five target tissues or through positional candidacy in significant or suggestive QTL. These gene–phenotype associations were integrated via a combinatorial analysis which represents the data as a bi-partite graph consisting of gene vertices and phenotype vertices. Here, a ‘phenotype’ vertex represents each list of correlates or candidates, and the gene vertices represent each list member. Edges connecting these vertices represent significant gene–phenotype associations. The union of all lists contained occurrences of approximately 33 000 genes and ESTs. However, many of these genes are represented only once, and through convergent analysis can be eliminated. By analyzing the number of phenotypes to which each gene was associated, we were able to identify those genes which most broadly and reliably related to the behavioral functions assessed. A ranked ordering of genes was generated based on each gene's connectivity to phenotypes. It should be noted that this ranking is somewhat biased by
