Development of an original algorithm to define serological antibody response thresholds that improve infectious diseases surveillance in Madagascar 3549

Abstract Description Serology is key for monitoring exposure to infectious diseases (ID), especially when molecular diagnostics are limited. A major challenge lies in defining positivity thresholds, particularly for emerging IDs without well-characterized reference samples. Cross-reactivity, from antibody binding to shared epitopes, complicates this process, risking inaccurate seroprevalence estimates. Developing robust methods is crucial for precise serological interpretation and effective disease surveillance. We evaluated Malagasy populational responses to 21 viruses using a Finite Mixture Model to classify antigen-specific responses in different clusters. The pgof parameter determined optimal cluster numbers. Higher counts were predicted for antigenically similar viruses, such as Flaviviruses (4–5 clusters). Thresholds were set on highly specific clusters, yielding seroprevalence estimates ranging from 0.3% for Zika to 8.9% for Dengue. For viruses with minimal epitope similarity, such as Lyssaviruses, single-cluster distribution was predicted, allowing thresholds to be set at Mean + 3 Standard Deviations, resulting in a 1.2% seroprevalence. To better address cross-reactivity, compensation matrices generated through high-affinity antibody removal are being generated, enabling more accurate seroprevalence estimates. This approach underscores the importance of combining statistical models and experimental methods to refine serological thresholds, address cross-reactivity, and enhance surveillance. Funding Sources supported by Agence Française de Developpement Topic Categories Computational and Systems Immunology (COMP)