Reconstructing HIV Epidemics from HIV Phylogenetics

项目摘要 艾滋病毒继续在少数群体中传播，主要是从不知道自己的人那里传播。 感染更有效地找到未确诊的艾滋病毒感染者接受治疗，并监测 预防工作，需要更好的流行病学技术。我们的项目汇集了一个团队 来自临床、分子生物学、流行病学、数学和进化的经验丰富的研究人员 领域的我们将通过结合进化理论、多尺度、 动态建模、人工智能以及大规模临床和序列数据。在这次更新中，我们将 扩展我们以前关于HIV宿主内进化过程如何与流行病学相互作用的工作， 动力学在量化了传播史和由此产生的艾滋病毒感染之间的联系之后， 我们将HIV的进化和流行病学之间的关系概念化为三个层次： 宿主内、传播和人群流行水平。因为艾滋病毒生物学的基本过程 和进化在很大程度上被忽略的建模流行病的水平，在目标1，我们检查宿主内 影响多样化的过程。我们将包括重组，选择，和潜伏在一个新的聚结 宿主内模型，以评估对流行病学水平的影响。我们还将量化潜力- 承载着多方向的选择压力。在目标2中，我们关注的是发生在 传输我们将发展一个新的最大似然法的基础上向前的时间概率模型 该方法改进了对多个传输之间的传输方向和传输时间的推断， 主机，并开发一种传输异质性检测方法，以评估整体可能的 感染者之间的传播异质性，以及检测在遗传学上的超 扩散可能已经发生。在aim 3中，我们将开发机器学习方法来处理超大数据 组（103-106例患者），并使用额外的临床和人口统计学数据来增强遗传学， 重建潜在的传播历史所有这三个目标都将涉及旨在发展 并改进下一代可编程动态应用程序的方法。

Project Summary HIV continues to spread, episodically, among minority groups, and mostly from people unaware of their infection. To more efficiently locate undiagnosed people living with HIV for treatment, as well as to monitor prevention efforts, better epidemiological techniques are needed. Our project brings together a team of experienced researchers from clinical, molecular biology, epidemiological, mathematical, and evolutionary fields. We will develop innovative epidemiological methods by combining evolutionary theory, multi-scale dynamic modeling, artificial intelligence, and large-scale clinical and sequence data. In this renewal, we will expand on our previous work on how HIV within-host evolutionary processes interact with epidemiological dynamics. Having quantified the link between transmission history and the resulting HIV phylogeny among hosts, we conceptualize the relationship between the evolution and epidemiology of HIV into three levels: within-host, at transmission, and on the population epidemic level. Because essential processes of HIV biology and evolution have been largely ignored when modeling the epidemic level, in aim 1 we examine within-host processes that affect diversification. We will include recombination, selection, and latency in a new coalescent within-host model to evaluate the impact on the epidemiological level. We will also quantify potential within- host multi-directional selection pressures. In aim 2, we focus on mechanisms that occur around the time of transmission. We will develop a new maximum likelihood method based on a forward-time probabilistic model of transmission that improves the inference of transmission direction and time of transmission among multiple hosts, and develop a transmission heterogeneity detection method to both assess overall possible transmission heterogeneity among infected persons, as well as to detect where in a phylogeny super- spreading may have occurred. In aim 3, we will develop machine learning methods to handle very large data sets (103-106 patients), and use additional clinical and demographic data to augment phylogenies in order to reconstruct the underlying transmission history. All three aims will involve advancements aimed at developing and improving methods for the next generation of phylodynamic applications.