Supervised hierarchical clustering of gene expression using tangent transformation

Genes are in perfect coherence under all or a subset of conditions if they are (1) regulated in the same or opposite direction; and (2) increase and decrease expression levels in an approximately fixed proportion to each other. While extensive work has been done on gene expression data, existing algorithms often do not capture these two properties of coherence simultaneously. To enable the simultaneous measurement of profile coherence in regulation proportion as well as regulation direction under either all or a subset of conditions, we propose a novel tangent transformation method, which can be applied for supervised gene expression analysis. Experimental results indicate that our tangent transformation is effective in a supervised hierarchical clustering of gene expression.