Table 3

Graph-based iterative Group Analysis of gene expression during the yeast diauxic shift.

The evidence network was constructed either from GeneOntology information (nodes are connected if they share a GeneOntology annotation) or from enzyme activity information obtained from Swissprot http://www.expasy.ch. In the latter case, genes are connected if their encoded proteins convert the same substrate (as product or educt, i.e. the direction of the reaction is not taken into account here). This type of network is much smaller (only 744 genes), as only genes coding for enzymes are included. All groups that are changed with a minimal p-value smaller than 1/[number of annotated genes] are shown, sorted by significance. The corresponding E-value as estimated by the analysis of 100 random permutations of the data is also shown. The employed threshold for inclusion in the table is very generous and does not guarantee that all subgraphs shown are statistically significant. The local minimum anchoring each regulated neighborhood is indicated by its genetic locus name (for overlapping neighborhoods, only the best-ranking minimum is shown). Descriptive group names were added manually. Groups that correspond to processes discussed in the original paper are highlighted in italics. It can be seen that the highest ranking group in each case is the largest and contains the central biological processes detected by DeRisi et al. (1997) and by iGA (see Table 1 and 2). N, number of genes in each subgraph.

Breitling et al. BMC Bioinformatics 2004 5:100   doi:10.1186/1471-2105-5-100