BMP7 was previously reported to participate in regulation of apoptosis in Semaxanib vascular smooth muscle cells. Given their association with apoptosis and their progression state-specific expression profiles, alpha4-integrin and BMP7 may represent constituents of switch mechanisms of carcinogenesis. The network clusters reveal regulatory circuitries that might be explored for novel therapeutic interventions. Indeed, PPARgamma antagonists are being investigated as treatment of various malignancies including liver cancers. Regulatory cascades targeting PPAR-gamma through upstream kinases and phosphatases, such as M3/6, JNK1, MEKK2, MKP3, MEK2, or ERK2, of which M3/6, MEKK2, and MFP-3 were induced during carcinogenesis, suggest additional possibilities for drug development. Furthermore, the ligand of insulin and insulin-like receptors, IGF-2, was strongly upregulated in tumor cells, whereas there were moderate changes in transgenic cells. The potential role of this ligand in autocrine regulation of cancer cell growth was previously discussed in the literature and further analyzed in our study. Transcription factors are important contributors to coordinated gene expression changes like those observed in the study data. It is a standard approach to test for overrepresentation of TF binding sites in promoters of coregulated genes versus a background of promoters. We quantified binding site enrichment by the 0.01- quantile of the ratio of two Beta distributions modeling the odds ratio of predicted binding sites and promoters and foreground and background gene sets. The 0.01-quantile value, in the following denoted q-value, estimates the value of the odds ratio, so that the true ratio is higher with 99% probability. For each of the 578 TRANSFAC PWMs, the algorithm started with a low PWM score threshold and iteratively adjusted the cut-off to maximize the q-value. In addition, binding site enrichment was tested in promoters of upregulated genes associated with cell cycle and of downregulated genes associated with lipid metabolism. In Figure 6, q-values of TRANSFAC motifs optimized for transgenic foreground sets are plotted against q-values of corresponding tumor foreground sets. Furthermore, we extracted the top PWMs ordered by q-values in Table S4. Identifiers of TRANSFAC matrices whose dots are highlighted in Figure 6 are bold-typed in Table S4. Extraction of matrices followed the rule to show the top 15 PWMs, or all with at least 2-fold enrichment in either transgenic or tumor set, or all PWMs highlighted in Figure 6, whichever resulted in the largest number of motifs. We also extracted transcription factor genes according to identified PWMs and performed upstream network analysis with transcription factor sets. As a result, promoter analysis revealed TF motifs specifically.