Editors' ChoiceAIDS

The Power of Epistasis

See allHide authors and affiliations

Science Translational Medicine  20 Apr 2011:
Vol. 3, Issue 79, pp. 79ec56
DOI: 10.1126/scitranslmed.3002508

The notoriously high mutation rate of the human immunodeficiency virus (HIV) grants it survival advantages against drugs designed to combat this AIDS-causing pathogen. Knowledge of the mutations that cause drug resistance is essential if researchers are to devise ways to combat this deadly adaptation; however, insufficient analytical methods hampered the ability of scientists to examine the collaborative effects of large numbers of mutations. Now, Hinkley et al. introduce a new analytical approach for unraveling the repercussions of multiple-mutation synergy.

The importance of epistasis—a synergistic interaction among gene variations—in phenotypical manifestations, relative to individual mutations, has long been debated. Comprehensive analysis of epistasis among mutations requires testing multiple interactions. Such analysis is computationally challenging and raises concerns of statistical model overfitting—the inclusion of excessive parameters that results in a description of relationships occurring in a particular data set but not necessarily reflective of true biology. Hinkley et al. modeled viral replicative capacity (that is, fitness) in the absence of drugs and in the presence of 15 different drugs for 1859 amino-acid variants in the HIV-1 protease and reverse transcriptase genes. In order to optimize the predictive ability of the model, the authors used a generalized kernel-ridge regression method for testing the main effects of individual mutations on fitness and the corresponding consequences of all pairwise interactions between mutations. The results of these analyses were coupled with knowledge of HIV-1 protease structure in order to identify patterns in the strength of epistatic effects across functional domains.

Models that included epistatic interactions had an average predictive power of 54.8% and in all drug conditions had on average 18.3% better predictive power than did models with main effects alone. Intragenic interactions accounted for greater improvements in predictive power than did intergenic (between protease and reverse transcriptase) interactions. Epistatic effects in protease mutations were strongest when at least one mutation occurred in any of three functional domains of the protein, and strength was correlated with physical proximity of the variants to each other in the three-dimensional protein structure. This work underscores the importance of deciphering multiple epistatic interactions in drug resistance studies. An accurate knowledge of drug resistance mechanisms holds the promise of better therapeutic interventions.

T. Hinkley et al., A systems analysis of mutational effects in HIV-1 protease and reverse transcriptase. Nat. Genet. 27 March 2011 (10.1038/ng.795). [PubMed]

Stay Connected to Science Translational Medicine

Navigate This Article