Monte carlo simulation of HIV-1 evolution in response to selection by antibodies

Abstract

The persistence of human immunodeficiency virus type 1 (HIV-1) infection may be in part attributed to its ability to evolve to evade neutralizing antibody (NAb) surveillance. We have tested the prediction that positive selection of V3 is positively correlated with the strength of a patient's NAb response by analyzing a data set containing both sequences from the principal neutralizing domain of HIV-1 (V3) and measures of the strength of the NAb responses of several patients. Only viral sequences from the patient with the strongest NAb response exhibited evidence of significant positive selection. To investigate the nature of selection by NAbs, we simulated the evolution of V3 at the nucleotide level. Two forms of plausible NAb selection were used: negative frequency-dependent selection and negative viral-age-dependent selection. Assuming negative age-dependent selection rather than negative frequency-dependent selection better simulated the temporal pattern of V3 variation from the patient with the strongest NAb response.

Monte Carlo estimates of natural variation in HIV infection

“…The rare exceptions to the ODE approach have been Monte Carlo approaches that include the stochasticity of HIV dynamics at both the epidemiological (BarthJones et al, 2000;Drusano et al, 2000;Gallop et al, 2002;Mode and Sleeman, 2002;Sale et al, 2002) and inhost levels (da Silva and Hughes, 2002;Kamina et al, 2001;Kousignian et al, 2003;Ruskin et al, 2002, Tan andWu, 1998;Tuckwell and Le Corfec, 1998). Drusano et al (2000) used a Monte Carlo simulation to examine the impact of dosing interval on the viral dynamics.…”

“…The simulation, however, can be easily modified to include different cell lifetime distributions, and a number of other biologically realistic effects such as accurate pharmacokinetics, complex adherence patterns, resistance mutations, etc. Because of the vast numbers of cells and virions involved, a detailed Monte Carlo approach such as ours has only recently become computationally feasible, using very large multiprocessor clusters (da Silva and Hughes, 2002).…”