Resampling-based analyses of the effects of combinations of HIV genetic mutations on drug susceptibility.

Viral genotype data aid in understanding the development of antiretroviral drug resistance and in identifying appropriate treatments. Using HIV-1 protease sequences and measures of in vitro sensitivity to the drug amprenavir, we develop a novel statistical approach that can be used to investigate combinations of mutations that alter drug susceptibility. Our method is based on the use of order statistics whose null distributions are estimated through resampling and used for formal hypothesis testing. We present a step-down testing method that preserves the overall family-wise error rate in finite samples via an application of the monotonicity condition of Romano and Wolf. Simulations demonstrate that the power of this new approach is comparable to a traditional resampling method; however, this approach can also be used as a visual diagnostic that may be informative even when specified hypotheses are not rejected, for example, in suggesting candidate regression models. Analysis of the data from the Stanford HIV database shows that while M46I/L mutations are associated with drug resistance, addition of the L88D/S mutation leads to hypersusceptible virus. Further addition of T90M/L mutations results in highly resistant virus. Use of this order statistics method allows the investigation of how mutations act in the presence of others and may suggest mechanisms by which resistance occurs or is reversed through the accumulation of mutations.