Detecting the positions of nonconsensus amino acids in HIV patients by marginal likelihood thresholding

We show how marginal likelihood thresholding can be applied in the context of multiple hypothesis testing, proposing a rule to select the tuning parameter involved. For detecting the positions of nonconsensus amino acids in patients suffering from two different HIV variants, we use a logistic regression framework and see that our results are in line with those from standard and advanced procedures controlling the false discovery rate, i.e. the proportion of incorrectly rejected null hypotheses.