Department of Statistics, UCLA

A smooth response surface algorithm is developed as an elaborate data mining technique for analyzing gene expression data and constructing gene regulatory network. A three-dimensional smooth response surface is generated to capture the biological relationship between the target and activator-repressor. This new technique is applied to functionally describe triplets of activators, repressors and targets, and their regulations in gene expression data. A diagnostic strategy is built into the algorithm to evaluate the scores of the triplets so that those with low scores are kept and a regulatory network is constructed based on this information and existing biological knowledge. The predictions based on the identified triplets in two yeast gene expression data sets agree with some experimental data in the literature. It provides a novel model with attractive mathematical and statistical features that make the algorithm valuable for mining expression or concentration information, assist in determining the function of uncharacterized proteins, and can lead to a better understanding of coherent pathways.