Leagan Gaeilge
Text Version
Printer Friendly
Sitemap
Warning: Your browser doesn't support all of the features in this Web site. Please view our accessibility page for more details.
HIV-1 evolution
Much of my research related to HIV has involved the development of probabilistic models of evolution that are tailored to understanding the relationship between viral sequence evolution and host immune responses or drug treatment. In addition, my group has provided bioinformatics support for the analysis of HIV-1 subtype C by South African researchers on the CHAVI and Caprisa projects. This work often makes use of HyPhy package, developed by Prof. Sergei Kosakovsky-Pond. HyPhy has a batch language that we use to define our own evolutionary models.
Most, if not all of the papers linked in the text are open access. If you cannot access a paper, you are welcome to request a pdf: cathal.seoighe _at_ nuigalway d_ot ie.
Some specific research projects
A phylogenetic Hidden Markov Model for epitope detection (see article in Molecular Biology and Evolution)
We have developed a probabilistic framework to integrate evolutionary information into the detection of immune epitopes. We have constructed a mutation-selection model of immune escape, dependent on the genotype of the infected host. The genotype of the host at immune loci influences the epitopes that are recognized and thus the location of selective pressure on the virus for immune escape. We exploit this to enhance epitope prediction in a HMM framework. The image below (from the paper) shows predicted and known epitopes for a number of HLA-alleles. For a figure legend and a more complete set of predictions see the paper
A model of amino acid switching applied to HIV-1 evolution (see article in PLoS Pathogens)
Positive Darwinian selection is often inferred from protein-coding nucleotide sequences by comparing the rate of nonsynonymous substitutions (dN) to the synonymous substitution rate (dS). If we make the typical assumption that synonymous substitutions are selectively neutral, then when dN is significantly greater than dS this implies that the average effect of a nonsynonymous substitution on the fitness of the organism is positive; hence, that nonsynonymous substitutions are being driven to fixation by positive Darwinian selection. Although dN >> dS implies positive selection, the converse is not true and in many interesting cases where a subset of the nonsynonymous mutations are driven to fixation by selection will not have dN > dS. We (and others) refer to dN >> dS as diversifying selection, the case in which any nonsynonymous mutation will, on average, confer a selective advantage. It may also be possible to use a comparison of dN and dS to detect selection when, on average, dN < dS at a site. One case in which this might occur is when there is selection for mutations between a subset of amino acids. If the subset of amino acids can be identified a priori then we can look specifically for dNs >> dS, considering only the rate at which nonsynonymous mutations between this subset of amino acids occur. Consider the example of immune escape and reversion in a virus. In the presence of a specific immune response targeted against a viral epitope the virus undergoes selection for immune escape, while in hosts that do not mount this specific immune response the virus may experience selection to revert to wild type (in the case that the immune escape incurred a fitness cost to the virus). This pattern of immune escape and reversion is often found at important T cell epitopes in HIV-1. We developed a model for selection affecting mutations between a subset of amino acids. This is described in full in the article linked above and depicted graphically below:
This model was somewhat more efficient at picking up sites that evolve under positive selection pressure, particularly sites where there is strong selection, but limited diversity. We consider that sites such as these are interesting for vaccine research because they represent regions of the virus that are likely to be targeted effectively by the immune response but where the virus has limited options for immune evasion. Such sites are illustrated below:
Analysis of HIV-1 evolution in acute infection and the influence of APOBEC (see publication in PLoS Pathogens, 2009)
Through CHAVI we studied the evolution of HIV-1 subtype B immediately following transmission to new hosts. Interestingly the vast majority of new HIV-1 infections appear to result from transmission of a single virion or virally-infected cell (see for example, Keele et al, PNAS, 2008). For this study we focussed only on patients that had been infected by a single viral variant (homogeneous infections). The objective was to search for consistent patterns of diversification across multiple patients, such as might be expected, for example, if there is a .selective sieve. upon transmission (as has been hypothesized). Evolution from the transmitted form of the virus to a form that is adapted for expansion in a newly infected host might be detectable as parallel evolutionary patterns across multiple infected patients.
This study found evidence of very early immune pressure, particularly evident at sites in the virus that are subject to hypermutation (as a consequence of the host innate immune APOBEC system). We also found some evidence of consistent patterns of evolution, particularly in the fusion domain of gp41, which remains unexplained. We have subsequently observed the same pattern in HIV-1 subtype C (unpublished).
Purifying selection acting on synonymous sites in HIV-1 (see publication in Virology Journal)
The assumption that synonymous substitutions are neutral is often badly violated, as there are several regions of the HIV genome that function at the nucleotide level. An example is the Rev Responsive Element (RRE) in the env gene. Rev binds to the RRE in order to facilitate the export of unspliced RNA from the nucleus, an essential part of the HIV-1 life cycle. In addition to scattered functional sites the HIV-1 RNA genome also has substantial conserved secondary structure, which affects the assumption of neutrality of the synonymous sites. When synonymous sites are functional, the synonymous rate of evolution is an underestimate of the neutral rate and this can result in false positive and misleading inferences of positive Darwinian selection. The figure (opposite) shows a survey of the synonymous substitution rate across HIV-1 inferred using HyPhy.
A model of directional selection applied to the evolution of drug resistance in HIV-1 (publication in Molecular Biology and Evolution)
This model was designed to detect evidence of selection pressure from longitudinally sampled pairs of sequences (in this particular application the pairs of sequences were from mothers and infants in a study of Nevirapine to prevent mother to child transmission of HIV-1). The model was encoded in R. An R workspace, including the code and the data from the manuscript is available upon request. A student based in Cape Town is working on a more user friendly implementation of this model.
School of Mathematics, Statistics and Applied Mathematics
National University of Ireland, Galway, University Road, Galway, Ireland.
Phone: +353 (0)91 492332 (direct) , +353 (0)91 524411 x2332 (switchboard)
Fax: +353 (0)91 494542 Email: Mary.Kelly@nuigalway.ie
This page was last updated Monday, February 15
National University of Ireland, Galway, University Road, Galway, Ireland.
Phone: +353 (0)91 492332 (direct) , +353 (0)91 524411 x2332 (switchboard)
Fax: +353 (0)91 494542 Email: Mary.Kelly@nuigalway.ie
This page was last updated Monday, February 15