Tion profile of cytosines within TFBS should be negatively correlated with TSS expression.KN-93 (phosphate) overlapping of TFBS with CpG “traffic lights” may affect TF binding in various ways depending on the functions of TFs in the regulation of transcription. There are four possible simple scenarios, as described in Table 3. However, it is worth noting that many TFs can work both as KPT-9274 biological activity activators and repressors depending on their cofactors.Moreover, some TFs can bind both methylated and unmethylated DNA [87]. Such TFs are expected to be less sensitive to the presence of CpG “traffic lights” than are those with a single function and clear preferences for methylated or unmethylated DNA. Using information about molecular function of TFs from UniProt [88] (Additional files 2, 3, 4 and 5), we compared the observed-to-expected ratio of TFBS overlapping with CpG “traffic lights” for different classes of TFs. Figure 3 shows the distribution of the ratios for activators, repressors and multifunctional TFs (able to function as both activators and repressors). The figure shows that repressors are more sensitive (average observed-toexpected ratio is 0.5) to the presence of CpG “traffic lights” as compared with the other two classes of TFs (average observed-to-expected ratio for activators and multifunctional TFs is 0.6; t-test, P-value < 0.05), suggesting a higher disruptive effect of CpG "traffic lights" on the TFBSs fpsyg.2015.01413 of repressors. Although results based on the RDM method of TFBS prediction show similar distributions (Additional file 6), the differences between them are not significant due to a much lower number of TFBSs predicted by this method. Multifunctional TFs exhibit a bimodal distribution with one mode similar to repressors (observed-to-expected ratio 0.5) and another mode similar to activators (observed-to-expected ratio 0.75). This suggests that some multifunctional TFs act more often as activators while others act more often as repressors. Taking into account that most of the known TFs prefer to bind unmethylated DNA, our results are in concordance with the theoretical scenarios presented in Table 3.Medvedeva et al. BMC j.neuron.2016.04.018 Genomics 2013, 15:119 http://www.biomedcentral.com/1471-2164/15/Page 7 ofFigure 3 Distribution of the observed number of CpG “traffic lights” to their expected number overlapping with TFBSs of activators, repressors and multifunctional TFs. The expected number was calculated based on the overall fraction of significant (P-value < 0.01) CpG "traffic lights" among all cytosines analyzed in the experiment."Core" positions within TFBSs are especially sensitive to the presence of CpG "traffic lights"We also evaluated if the information content of the positions within TFBS (measured for PWMs) affected the probability to find CpG "traffic lights" (Additional files 7 and 8). We observed that high information content in these positions ("core" TFBS positions, see Methods) decreases the probability to find CpG "traffic lights" in these positions supporting the hypothesis of the damaging effect of CpG "traffic lights" to TFBS (t-test, P-value < 0.05). The tendency holds independent of the chosen method of TFBS prediction (RDM or RWM). It is noteworthy that "core" positions of TFBS are also depleted of CpGs having positive SCCM/E as compared to "flanking" positions (low information content of a position within PWM, (see Methods), although the results are not significant due to the low number of such CpGs (Additional files 7 and 8).within TFBS is even.Tion profile of cytosines within TFBS should be negatively correlated with TSS expression.Overlapping of TFBS with CpG "traffic lights" may affect TF binding in various ways depending on the functions of TFs in the regulation of transcription. There are four possible simple scenarios, as described in Table 3. However, it is worth noting that many TFs can work both as activators and repressors depending on their cofactors.Moreover, some TFs can bind both methylated and unmethylated DNA [87]. Such TFs are expected to be less sensitive to the presence of CpG "traffic lights" than are those with a single function and clear preferences for methylated or unmethylated DNA. Using information about molecular function of TFs from UniProt [88] (Additional files 2, 3, 4 and 5), we compared the observed-to-expected ratio of TFBS overlapping with CpG "traffic lights" for different classes of TFs. Figure 3 shows the distribution of the ratios for activators, repressors and multifunctional TFs (able to function as both activators and repressors). The figure shows that repressors are more sensitive (average observed-toexpected ratio is 0.5) to the presence of CpG "traffic lights" as compared with the other two classes of TFs (average observed-to-expected ratio for activators and multifunctional TFs is 0.6; t-test, P-value < 0.05), suggesting a higher disruptive effect of CpG "traffic lights" on the TFBSs fpsyg.2015.01413 of repressors. Although results based on the RDM method of TFBS prediction show similar distributions (Additional file 6), the differences between them are not significant due to a much lower number of TFBSs predicted by this method. Multifunctional TFs exhibit a bimodal distribution with one mode similar to repressors (observed-to-expected ratio 0.5) and another mode similar to activators (observed-to-expected ratio 0.75). This suggests that some multifunctional TFs act more often as activators while others act more often as repressors. Taking into account that most of the known TFs prefer to bind unmethylated DNA, our results are in concordance with the theoretical scenarios presented in Table 3.Medvedeva et al. BMC j.neuron.2016.04.018 Genomics 2013, 15:119 http://www.biomedcentral.com/1471-2164/15/Page 7 ofFigure 3 Distribution of the observed number of CpG “traffic lights” to their expected number overlapping with TFBSs of activators, repressors and multifunctional TFs. The expected number was calculated based on the overall fraction of significant (P-value < 0.01) CpG "traffic lights" among all cytosines analyzed in the experiment."Core" positions within TFBSs are especially sensitive to the presence of CpG "traffic lights"We also evaluated if the information content of the positions within TFBS (measured for PWMs) affected the probability to find CpG "traffic lights" (Additional files 7 and 8). We observed that high information content in these positions ("core" TFBS positions, see Methods) decreases the probability to find CpG "traffic lights" in these positions supporting the hypothesis of the damaging effect of CpG "traffic lights" to TFBS (t-test, P-value < 0.05). The tendency holds independent of the chosen method of TFBS prediction (RDM or RWM). It is noteworthy that "core" positions of TFBS are also depleted of CpGs having positive SCCM/E as compared to "flanking" positions (low information content of a position within PWM, (see Methods), although the results are not significant due to the low number of such CpGs (Additional files 7 and 8).within TFBS is even.