Comparison of binding sites identified by direct subtraction of mock control and the proposed normalization without mock control. The two biological replicates for MSL binding in embryos (blue and green) show a large overlap (85%) in the identified sites. The new normalization method involving rotation and lowess-curve fitting (red) was applied to each replicate. In both cases, the new method identifies nearly all of the common sites (98% and 97%) found by direct subtraction and many of the sites found by only one. Also, the new method for sample 1 gives a greater percentage of overlap with mock-subtracted sample 2 than mock-subtracted sample 1 does. Similarly, the new method for sample 2 gives a greater percentage of overlap with mock-subtracted sample 1 than mock-subtracted sample 2 does.