Simulating within host human immunodeficiency virus 1 genome evolution in the persistent reservoir

Abstract

The complexities of viral evolution can be difficult to elucidate. Software simulating viral evolution provides powerful tools for exploring hypotheses of viral systems, especially in situations where thorough empirical data are difficult to obtain or parameters of interest are difficult to measure. Human immunodeficiency virus 1 (HIV-1) infection has no durable cure; this is primarily due to the virus’ ability to integrate into the genome of host cells, where it can remain in a transcriptionally latent state. An effective cure strategy must eliminate every copy of HIV-1 in this “persistent reservoir” because proviruses can reactivate, even decades later, to resume an active infection. However, many features of the persistent reservoir remain unclear, including the temporal dynamics of HIV-1 integration frequency and the longevity of the resulting reservoir. Thus, sophisticated analyses are required in order to measure these features and determine their temporal dynamics. Here, we present software that is an extension of SANTA-SIM to include multiple compartments of viral populations. We used the resulting software to create a model of HIV-1 within host evolution that incorporates the persistent HIV-1 reservoir. This model is composed of two compartments, an active compartment and a latent compartment. With this model, we compared five different date estimation methods: (Closest Sequence, Clade, Linear Regression, Least Squares and Maximum Likelihood) to recover the integration dates of genomes in our model’s HIV-1 reservoir. We found that the Least Squares method performed the best with the highest concordance (0.80) between real and estimated dates and the lowest absolute error (all pairwise t tests: p < 0.01). Our software is a useful tool for validating bioinformatics software and understanding the dynamics of the persistent HIV-1 reservoir.

HIV Proviral Burden, Genetic Diversity, and Dynamics in Viremic Controllers Who Subsequently Initiated Suppressive Antiretroviral Therapy

“…We also used least-squares dating (LSD) ( 45 ), an approach that aims to minimize the variance between the tree branch lengths and sample dates, that has also recently been used to infer proviral sequence ages ( 46 ). We used the original rooted tree; here it may be useful to reiterate that the root represents the location that maximizes the (Spearman’s) correlation between plasma HIV root-to-tip distances and sampling times and thus does not rely on a strict clock.…”

HIV proviral burden, genetic diversity, and dynamics in viremic controllers who subsequently initiated suppressive antiretroviral therapy

“…We also used Least Squares Dating (LSD) (46), an approach that aims to minimize the variance between the tree branch lengths and sample dates, that has also recently been used to infer proviral sequence ages (47). We used the original rooted tree; here it may be useful to re-iterate that the root represents the location that maximizes the (Spearman’s) correlation between plasma HIV root-to-tip distances and sampling times, and thus does not rely on a strict clock.…”

Bayesian Phylogenetic Inference of HIV Latent Lineage Ages Using Serial Sequences

“…This method relies on a molecular clock, and is not used if the clock is rejected. Jones and Poon developed a related method, estimating mutation rate in the same way but estimated internal node ages and unknown tip ages using a maximum likelihood (ML) approach using a specified mutation rate (15, 16). To et al developed a distance method using a least squares (LS) approach to estimate mutation rates and date internal nodes and tips with unknown ages (17).…”

“…The two methods used by Abrahams et al may be very sensitive to the number of sequences sampled and the sampling times. Simulation studies suggest that LS may out-perform all of these methods (15, 17). An alternative to these existing methods could be developed based on established parametric phylogenetic models that use tip dating for estimating and calibrating phylogenies of viral data, and are potentially more accurate (18, 19).…”

“…In contrast, simulation methods that have been used to evaluate methods for dating integration events largely ignore the underlying population dynamics of HIV. Some assume a constant rate birth-death process while other use a compartmental model with logistic growth (13, 15). Epidemiologists use more complex models, typically ordinary differential equations (ODEs), to describe HIV viral dynamics (21–23).…”