Elucidation of Transcriptional Regulatory Relationships Via Information Theoretic Clustering and Consensus Nucleotide Motif Extraction

The process of extracting biologically relevant data from large-scale genetic experiments often begins with clustering genes based on their co-regulatory expression patterns over the course of a given experiment. An examination of the proposal that clustering genes exhibiting both positively- and negatively-correlated expression provides insight into functional relationships not extracted by traditional similarity metrics is presented here. This proposal is analyzed through application of the information theoretic concept of mutual information, a similarity metric which can detect both correlated and uncorrelated associations in gene expression data, adapted to the agglomerative-hierarchical average-linkage method of clustering microarray data. This clustering technique was applied to two micro array data sets, one measuring yeast cell cycle activity over 170 minutes and two complete cycles and another measuring mouse dendritic cell maturation over a 24-hour time-course following stimulation with poly(I:C). As expected from this method, the resulting clusters for both data sets exhibited both positively- and negatively-correlated expression patterns. For the mouse data, separate analysis of the gene ontology functions inherent in each cluster 's up- and down-regulated genes indicates functionally disjoint activities occurring simultaneously, perhaps controlled by a common regulatory event in the upstream signaling pathway, which could include nucleosome modification leading to changes in promoter accessibility. As a further attempt to elucidate functional relationships inherent in coregulated genes, a tool called Site-MAPS is presented that extracts conserved nucleotide motifs in the non-coding regions of clustered genes. This approach utilizes probabilistic suffix trees to extract consensus motif candidates from upstream or downstream noncoding regions, statistically quantifies their significance, then consolidates and displays the results in a web-based interface. Application of this method to co-regulated genes within the clusters described above, as well as a validation set of clusters consisting of yeast genes with known consensus binding sites in their non-coding regions, successfully identified known regulatory motifs.

Grantor dc:publisher

Wake Forest University

Year dc:date.issued

2008

Author dc:creator

• Fye, Jason

Handle dc:identifier.uri

http://hdl.handle.net/10339/32730

OAI identifier oai:identifier

oai:wakespace.lib.wfu.edu:10339/32730