Understanding drivers of phylogenetic clustering in molecular epidemiological studies of HIV.

Abstract

Despite declarations that the so-called end of AIDS is near, the global human immunodeficiency virus (HIV) epidemic continues to grow. This is the case within localized epidemics in the resource-rich world, as well as so-called generalized epidemics in the resource-limited setting. With the relative lack of success of widespread prevention approaches, focus has turned to the finer granularity of the epidemics [1]. Even within generalized epidemics in sub-Saharan Africa, evidence points to increasing heterogeneity in transmission [2]. More-precise determination of the characteristics of individuals continuing to spread the virus—for instance, whether their infection is undiagnosed, diagnosed and untreated, or diagnosedandtreated—isneededtoguide prevention to reduce transmission to manageable levels. It is within this context that the use of viral genetic sequences can add important value to inference of transmission dynamics and inform targeted prevention strategies [3]. Sequence data are becoming increasingly used in epidemiological studies for avariety of pathogens, with recent recommendations for how such studies are to be reported [4]. Whereas the major focus for implementing molecular epidemiological approaches for HIV should be toward reducing the devastating epidemics in Africa and Asia, the vast majority of HIV sequence data derives from North America and Europe, mainly as a result of widespread HIV genotypic drug resistance testing. Nevertheless, sequence data from resource-rich settings represent an invaluable resource for developing methods that can be applied globally. This is particularly the case when sequence databases cover a significant fraction of individuals infected with HIV, as exemplified in countries such as the United Kingdom, Switzerland, and the Netherlands. Use of phylogenetics to identify the likely source of specific transmission events is a well-trodden path in HIV research [5], particularly in relation to small, targeted epidemiological investigations. However, when applying such approaches to data sets sampled at a regional or national level, the sampling fraction is too low to detect significant numbers of direct transmissions [6].Nevertheless, so-called clusters of highly