The Human Immunodeficiency Virus (HIV) has a widespread effect across the world as it continually infects more and more people. Therefore, a major focus of immunologists and virologists is to treat the virus. However, the virus tends to evolve very quickly inside a host to avoid immune and drug selective pressure and diversifies with low selection pressure among hosts (Vracken 2). Within hosts, the virus is heavily affected by a variety of immune responses, causing the virus to develop escape mutations quite quickly. However, inter-host evolution does not appear to be driven by a similar selective process -- strain variety in HIV is most often a result of the founder effect", in which by chance the first strain in the area is the most prevalent (Lemey 2).

There are four main types of antiretroviral drugs that can treat HIV - Nucleoside Reverse Transcriptase Inhibitors (NRTIs), Non-nucleoside Reverse Transcriptase Inhibitors (NNRTIs), Protease inhibitors (Pls), and integrase inhibitors (INIS). This study focuses on NRTIs, the first antiretroviral drug developed. However, the methods used here can easily be extended to any type of antiretroviral drug.

As a result of its high variability and fast evolution, antiretroviral drug treatment for HIV is often unsuccessful. Many modern drug treatments are a regimen of combined and changing antiretroviral drugs to try and circumvent HIV's quick evolution.

This research aims to provide an easier way to predict the fitness (relative ability to survive and reproduce) of different strains of HIV under drug stress. This model can be used to predict further drug resistance mutations as well as predict conserved domains in the amino acid sequence. This will allow for easier treatment of HIV through antiretroviral drugs as well as inform scientists in drug design.

In this project, a basic model of the fitness of each strain of HIV was used based on the ising Spin Model in particle physics (Shekhar 1). This statistical model can predict the relative fitness of each new sequence of HIV based on its similarity to a consensus sequence, taking into account common double mutations. This model has been shown to also correlate with in vitro fitness of HIV (Ferguson 3). The model developed in this project is two Ising Spin Models - one using parameters created with drug-naive sequences and one using parameters with drug stressed sequences. This allows the model to differentiate between resistance mutations and mutations common in a drug-naive virus.

The result of this research is a model that is able to predict HIV evolution under drug stress, leading to some inferences about new resistance mutations. The model predicts 22 single mutations that confer resistance to NRTIs and 4,291 pairs of mutations that cause resistance. Further research is necessary to test this model in vitro with genetically engineered reverse transcriptase mutants in order to check if the mutations predicted by the model do indeed cause antiretroviral resistance.