Reconstruction of historical malaria transmission in Senegal using multiplex serocatalytic models

Abstract The advent of multiplexing technologies, allowing antibodies to hundreds of antigens to be measured in a single test, has led to enormous increases in the amount of data generated by serological surveys. New modelling methods are required to exploit this data. This study extends serocatalytic models to consider up to three antibody responses targeting the same pathogen simultaneously. These models were fitted to data from cross sectional serological surveys of Plasmodium falciparum malaria in the Senegalese villages of Dielmo and Ndiop, and model predictions were validated against 22 years of longitudinal epidemiological data. The most accurate reconstruction of historical clinical incidence of P. falciparum was provided by a combination of antibodies to Apical Membrane Antigen 1 (AMA1) and Glutamate-Rich Protein (GLURP). This model estimated a 76% (95% CrI: 61% - 86%) drop in transmission in 2004 (95% CrI: 2001 – 2008) coinciding with changing anti-malarial treatment. Multiplex serocatalytic models provided more accurate estimates of past clinical incidence than singleplex serocatalytic models, with the most accurate multiplex model (AMA1 + GLURP) outperforming all singleplex models (Kruskal Wallis p < 0.01). Finally, models with three antigens did not provide more accurate estimation than models with two antigens. Significance Statement Multiplex serological assays are accelerating the uptake of serology as a surveillance tool for malaria. However, there has been a lag in the development of analytic tools for the analysis of this rich multiplex data on measurements of antibodies to multiple malaria antigens. In this study, new mathematical models are developed to account for multiplex data. We demonstrate that these models provide more accurate estimates of past malaria transmission in Senegal, with lower levels of uncertainty compared to models based on data from antibodies to a single antigen. By providing more reliable estimation, we anticipate that our work can form the basis of an important new tool for malaria surveillance.