To identify windows with recombination rates significantly higher than mean background rates, that is, recombination hotspots, we used the sequenceLDhot program (Fearnhead 2006; http://www.maths.lancs.ac.uk/∼fearnhea/Hotspot/).ac.uk/∼fearnhea/Hotspot/
The current data set suggested that a statistical threshold using fewer than three standard deviations above the mean background count rate can increase the true positive rate without exceeding a 50% false positive rate.
For the current data set shown in Figure 3, the best combination of specificity and sensitivity, as indicated by the + symbol, occurred at a true positive rate of 65% and false positive rate of 32%, and corresponds to a probe positive threshold level of the mean background count rate plus 4.3 standard deviations.
Unlike the ratiometric threshold, in theory, the three-sigma statistical threshold approaches a 1-to-1 target-to-background ratio as the mean background count rate increases.
The ROC curve was generated by varying the number of standard deviations that must be exceeded above the mean background count rate for a probe positive threshold.
The operating point on the data curve for three standard deviations above the mean background count rate (i.e., three-sigma statistical threshold) indicated by the * symbol in Figure 3, corresponds to a true positive rate of 71% and false positive rate of 40%.
A statistical threshold for probe positivity of the mean background count rate plus two standard deviations, for example, would reduce the statistical confidence level of a true positive and true negative measurement from a 85.6% probability for three-sigma statistical threshold to a 76.0% probability for the for two-sigma statistical threshold.
Therefore, it is our opinion that the three-sigma statistical threshold criteria (i.e., probe positive threshold level of the mean background count rate plus 3 standard deviations) for determination of gamma detection probe positivity represents a good balance of sensitivity, selectivity, and statistical confidence levels for positive and negative probabilities.
The background rates for the P generations varied from 5.9% to 13.0% with a mean value of 8.2%, whereas the F1 generation background rate varied from 1.4%too 19.8% with a mean value of 13.5%.
