Prediction of regulatory networks: identification of transcription factor-target relationship from gene ontology information and gene expression data

Defining regulatory networks, linking transcription factors (TFs) to their targets, is a central problem in post-genomic biology. Here we apply an approach based on Nearest Neighbour (NN) Algorithm to predict the targets of a transcription factor by combining gene ontology (GO) and gene expression data. In particular, we used NN algorithm to predict the regulatory targets for 36 transcription factors in the Saccharomyces cerevisiae (Qian J. et al., 2003, Bioinformatics. 19(15):1917-26) based on the gene ontology and microarray expression data from various physiological conditions. We trained and tested our NN algorithm on a data set which contains a number of both positive and negative examples. The overall success rate by the jackknife test for the dataset was 97%, and that for the regulatory targets(positive) was 58%, suggesting that such a hybrid approach particularly by incorporating the knowledge of gene ontology) may become a useful high-throughput tool in the area of regulatory networks modelling