However, at low score cut-offs, where high numbers of binding sites are predicted in foreground and background promoters, statistical tests like the exact binomial test can report highly significant P-values for small differences when these are supported by high counts of binding site LDN-193189 distributor instances. Consequently, one cannot start from arbitrarily low PWM score thresholds in order to find the best, most likely higher cut-off. Eventually, one might also like to prioritize binding site motifs and would naturally assign highest priority to motifs with strongest enrichment. In this case, simply calculated odds ratios may result in a different ordering of motifs than corresponding enrichment P-values. We therefore developed an approach that focuses on the magnitude of overrepresentation expressed as a probabilistic estimate of the odds ratio of binding sites and promoter sequences in foreground and background gene sets. This quantity cannot be trapped by highly significant P-values associated with small odds ratios, because it focuses on a estimate of the odds ratio itself. Finally, we assume that the proposed statistic enables more intuitive prioritization of motifs by expressing their importance as elative enrichment in foreground promoters. It may however by perceived as potential drawback to specify the quantile interval as a free parameter, which was set here to 1%.In this study, we applied the new promoter analysis method to transgenic and tumor promoter sets and subsequently compared the results of both progression states. Similar to comparison of GO analyses, this setup enabled us to not only identify highly enriched binding sites in promoters of DE genes, but also to observe differences in importance of certain motifs for transgenic and tumor gene sets. As described in the results section, promoter analysis supported stronger regulation of cell cycle and of lipid metabolism in the tumor state by associating motifs of Atf3, Jun, E2f3, and Pparg with corresponding tumor genes and thus supported our previous findings. Importantly, this part of our study revealed overrepresentation of POU motifs predominantly in transgenic promoter sets. Expression profiles of POU factors as well as of HMG and Forkhead factors whose motifs were also identified through promoter analysis led us to propose Oct4, Tcf7, Lef1, and Foxc1 as regulators of a transcriptional network potentially under control of Wnt signaling. We speculate that the factors contribute to regulation of developmental pathways, which were enriched in both up- and downregulation according to GO analysis. Although network analyses carried out in this work did not reveal a link between Oct4 and EGF-signaling on the level of protein-protein interactions, in-vivo binding fragments for c-Myc, c-Jun.