I'm performing a bioinformatic study using a method to detect selection within cattle genome.
A previous author developed this method with a statistic calculated from DNA genotypes. The method consists in performing a null distribution with values calculated from between chromosomes genotype data and stablish a critical threshold from this null distribution. Then multiple comparisions between the threshold and experimental values within chromosomes are performed and experimental values greater than the threshold are deemed as significant deviations.
The original author of the method set the critical threshold as the 1-(1-x) percentil, where x is the number of tests (of experimental values to test), leaving an expectation of only one false positive. The problem is that while the original author performed one and a half million tests, i have to test around 53 millions (my nulo distribution has 60 million values). This leads to a percentil that is almost 100% and the critical threshold is an outlier value. So in summary, the huge number of comparisions leads to a huge type II error by doing the multiple test correction.
I want to ask for an alternative method to perform my study. Thanks for your attention.