Statistical methods for the analysis of multi-antibody data to inform malaria control and elimination strategies

抗体数据为疾病控制和监测提供信息的潜力正日益得到认识。流感、沙眼、淋巴丝虫病和疟疾是积极进行血清监测的传染病的例子。因此，一直主张建立世界血清库，以促进下一代血清监测工具。在过去十年中，由于疟疾传播的减少，疟疾死亡和病例的减少证明了血清监测对疟疾的相关性增加。在低传播环境中，旨在检测人类或蚊子感染情况的传统疟疾指标估计的不确定性可能会显著增加。此外，这些指标受到抽样范围和疟疾传播季节性的强烈影响，这一事实加剧了这一问题。血清学研究克服了这些限制，因为它们的目标是量化暴露，而不是感染。因此，世界卫生组织(WHO)正在其消除疟疾指南中考虑血清学评估。血清流行病学分析的普遍做法是估计疟疾传播强度，独立处理多种抗体反应的数据。该项目将专注于开发多元统计方法，以克服这种方法的局限性，充分利用多抗体数据中的信息力量。总体而言，该项目有三个主要目标：(1)使用机器学习技术在多重数据中选择信息丰富的抗体；(2)将现有的无阈值方法扩展到多变量环境；(3)在疾病消除前的背景下研究多变量血清学结果的表现。该项目中开发的统计方法还将应用于其他传染病(例如新冠肺炎)，在这些疾病中，血清学评估也是其控制和消除战略的优先事项。

The potential of antibody data to inform disease control and surveillance is being increasingly recognised. Influenza, trachoma, lymphatic filariasis and malaria are examples of infectious diseases where sero-surveillance is actively performed. As a consequence, the creation of a World Serum Bank has been advocated to facilitate the next generation of sero-surveillance tools.In the last decade, the relevance of sero-surveillance for malaria has increased due to reductions in malaria transmission evidenced by decreasing numbers of malaria deaths and cases. In low transmission settings, the uncertainty in the estimates of conventional malaria metrics aimed at detecting the presence of infection in humans or mosquitoes, can increase significantly. In addition, this issue is exacerbated by the fact these metrics are strongly affected by the sampling frame and the seasonality of malaria transmission. Serological studies overcome such limitations, because they aim to quantify exposure rather than infection. As a result, serological assessment is now being considered by the World Organization (WHO) in their guidelines for malaria elimination.The prevailing practice in sero-epidemiological analyses is to estimate malaria transmission intensity treating the data from multiple antibody responses independently. This project will focus on the development of multivariate statistical methods that overcome the limitations of this approach to fully borrow the strength of information across multi-antibody data. Overall, the project has three main objectives: (i) selection of informative antibody in multiplex data using machine learning techniques, (ii) extending existing threshold-free methodology to a multivariate setting, and (iii) study of the performance of multivariate serological outcomes in the context of disease pre-elimination elimination. The developed statistical methods in this project will also be deployed to other infectious diseases (e.g., COVID-19), where serological assessment is also a priority in their control and elimination strategies.