The goal of this thesis is to find transcription factor binding site motifs through the use of microarray expression data, protein-DNA binding data from chromatin immunoprecipitation on a microarray chip, and conservation of the genomic sequence among species. The techniques described here were applied to data from Arabidopsis thaliana. In gene lists obtained using different techniques, over-represented sequence motifs were searched using Gibbs sampling algorithms in well-tested implementations. From these genes, at least some have to have their expression directly regulated by a transcription factor, and their promoter region will contain a binding site for it. This is a measurable property of the sequence that can be learned by a machine learning algorithm. Using expression profiles of the genes as labels, a support vector machine with an integrated kernel learned to distinguish between regulated and unregulated genes on the basis of differences in the motifs, in terms of the motif ’s surrounding sequence, distribution, relative distance to the transcription start, pairwise distance, conservation in related plantae, and other sequence characteristics. This information can be used to infer a network of genes responding to the change in expression of a transcription factor, if the logic of the experiments is taken into account. With the motif now known we can go on to analyze other genes that have this motif and may be part of the network, but were not contained in the original list because, for instance, their expression level was below the microarray sensitivity or they were not present on the chip. Network creation ultimately remains a manual process, because modeling the experimenter’s rationale in designing these experiments is a difficult enough task on its own. To test the performance of the selected approach, we try to predict an initial network of genes responding to heat shock, a response that is well characterized and the primary response genes’ regulatory motif is known. The work flow proposed is able to find this motif and ranks it among the top three for the heat shock microarray data. The approach was also trained on data obtained from loss- and gain of function experiments centered around the transcription factor WUS CHEL. The results on this original research data show that we not only find direct targets of transcription factors, but must assume that most alterations in expression we observe are a secondary response provoked by expression change in transcription factors regulated directly. Secondary targets are difficult to discriminate without additional data, so we introduce a method to rank genes by their being bound by a transcription factor. For this, we use data from chromatin immunoprecipitation on a chip. The results for WUS CHEL point out likely targets of primary and secondary regulation, including proposals for a binding motif. The approach shows a lot of promise as it performs as expected on data with known results; confirmation for the WUSCHEL transcription factor targets will be done experimentally on the basis of the list obtained through this method.