In order to find gene regulatory networks from microarray data, it is
important to first find direct regulatory relationships between pairs of
genes. We propose a new method for finding potential regulatory
relationships between pairs of genes from microarray time series data and
apply it to expression data for cell-cycle related genes in yeast. We compare
our algorithm, dubbed the event method, with the earlier correlation method
and the edge detection method by Filkov et al. When tested on known
transcriptional regulation genes, all three methods are able to find similar
numbers of true positives. The results indicate that our algorithm is able to
identify true positive pairs that are different from those found by the two
other methods. We also compare the correlation and the event methods using
synthetic data and find that typically, the event method obtains better
results.