Figure 5

Accuracy of automated sister chromatid exchange (SCE) detection by Bioinformatic Analysis of Inherited Templates (BAIT). (a) By comparing the number of SCE events identified by BAIT to those determined manually, we calculated the percentage of computational calls that were incorrect (false positives) or not detected (false negatives). Filtering the data by only including bins that deviated minimally from the mean changed the results, with highly conservative filtering increasing the level of false negatives, and very broad filtering increasing the level of false positives. (b) The frequency of (left) false positives and (right) false negatives with respect to library background. Cleaner, high-quality libraries with < 1% of reads mapping incorrectly had a lower false-positive rate than libraries with medium background (<5% incorrectly mapped reads), and an even lower rate than libraries with high background (<10% incorrectly mapped reads). Error bars are ± standard deviation.