New analytic approaches and endpoints in human HIV vaccine correlate studies

摘要 艾滋病毒疫苗效力试验因接触率低和保护水平低而变得复杂。 然而，尽管有这些障碍，感染风险(COR)的关键相关因素已经从失败的试验中定义出来 符合总体疗效标准。关于保护性免疫反应、宿主遗传学和 通过对无效和略有效果的试验的分析得出病毒易感性。尤其是，在中和抗体的同时 广度一直被认为是疫苗介导的保护的关键，Fc介导的效应器的广度 功能只是最近才开始探索的。鉴于自然感染的丰富数据，被动抗体 转移研究探索作用机制、临床前疫苗疗效研究和先前的人类HIV-1 支持抗体效应器功能的潜在作用的疫苗效力试验有助于预防 这些活动代表了疫苗研究中调查和优化的重要途径 和发展。 为了确定这些COR，病例对照分析通常通过比较以下因素的反应谱来进行 感染者(病例)和未感染者(对照)。但是，“控制”或未受感染的Subject类是 缺乏保护性反应且没有接触病原体的混合个体，以及 那些具有保护性反应并且要么暴露或没有暴露的人。对于效果不佳的疫苗， 预计大多数对照受试者表现出的反应表型与 案子。因此，传统的COR分析受到受保护对象的稀释，而这些不受保护的对象 而是未曝光的对象。我们建议评估新的机器学习(ML)技术，这些技术可以稳健地 根据免疫原性(或其他)数据推断疫苗接种个人的保护状态，以便 在这些具有挑战性的环境中促进相关的发现。 因此，我们提出了串联方法，以开发对艾滋病毒疫苗有效性的新见解。 ·开发新的分析方法，将不仅与这一艾滋病毒，而且与其他艾滋病毒有关的分析联系起来 疫苗试验和其他方面，以及 ·通过收集新的体液免疫反应数据，确定抗体效应物的广度和效力 用于HVTN702试验的功能。

SUMMARY HIV vaccine efficacy trials have been complicated by low rates of exposure and low levels of protection. Yet, despite these barriers, crucial correlates of infection risk (CoR) have been defined from trials that failed to meet overall efficacy criteria. Much has been learned about protective immune responses, host genetics, and viral susceptibility from analysis of ineffective and marginally effective trials. In particular, while neutralizing Ab breadth has long been considered key to vaccine-mediated protection, breadth of Fc-mediated effector functions has only more recently begun to be explored. Given the rich data from natural infection, passive Ab transfer studies probing mechanism of action, preclinical vaccine efficacy studies, and prior human HIV-1 vaccine efficacy trials that support the potential role of Ab effector functions to contribute to protection from infection, these activities represent an important avenue of investigation and optimization in vaccine research and development. To define these CoR, case-control analysis is typically conducted by comparing the response profiles of infected (case) and uninfected (control) subjects. However, the “control", or uninfected subject class is a mixture of individuals that lack the protective response and were simply not exposed to the pathogen, and those that possess the protective response and either were or were not exposed. For poorly effective vaccines, the majority of the control subjects are expected to show a response phenotype indistinguishable from the cases. Thus, traditional CoR analysis suffers from the dilution of the protected subjects with these unprotected but unexposed subjects. We propose to evaluate novel machine learning (ML) techniques that can robustly infer the protection status of vaccinated individuals on the basis of immunogenicity (or other) data in order to facilitate correlates discovery under these challenging circumstances. Accordingly, we propose tandem approaches to develop new insights into HIV vaccine efficacy · by developing new analytical approaches to correlates analysis relevant not only to this but other HIV vaccine trials and beyond, and · by collecting new humoral immune response data defining the breadth and potency of antibody effector function for the HVTN702 trial.