Figure 2

ChIP-chip procedure (A). MA plot for a control experiment, after normalization (B; M = log2(Red/Green); A = log2(Red*Green)/2). Fit of the gamma+normal model to the log ratio of red versus green channels (C). The red portion of the histogram shows probes that belong to the unbound distribution with p < 0.05. The green portion of the histogram are probes that belong to the bound distribution with p < 0.05. The grey areas in between are ambiguous. Our model allows us to annotate promoters as being active or inactive at different confidence levels (D). "High on" indicates strong evidence for RNA Polymerase II binding in both replicates (the probes fall within the green portion of panel C); "Medium on" indicates strong evidence for RNA Polymerase II binding in one replicate, and weak evidence in the other (i.e., the probes fall outside of the red area in panel C). "Low on" indicates weak evidence in both replicates. "Low off" indicates inconsistency between the replicates, and finally "Strong off" indicates a high probability that no binding occurred (i.e., probes fall within the red portion of panel C). "Low on" indicates weak evidence in both replicates. "Low off" indicates inconsistency between the replicates, and finally "Strong off" indicates a high probability that no binding occurred (i.e., probes fall within the red portion of panel C).