PQG Working Group

Brian Cleary

Broad Fellow
Broad Institute of MIT and Harvard

Using viral loads and within-host models to improve COVID-19 surveillance

Limited testing capacity has been an ongoing problem throughout the COVID-19 pandemic. Pooled testing is a faster and less expensive diagnostic approach compared to individual testing, but there are important tradeoffs in sensitivity, efficiency and logistics to consider. In this talk, we outline an approach combining within-host hierarchical models, compartmental models of pathogen spread, and viral load data to identify optimized pooling protocols that are robust to the changing state of an epidemic. This study highlights the importance of considering within-host biology and individual-level heterogeneity when evaluating epidemiological surveillance strategies. We will also introduce HYPER, a new pooled testing method based on hypergraph factorization. HYPER is designed to be easy to implement and adapt, while also producing pools that are balanced and efficient. We will discuss what hypergraph factorizations are and how they generate the pooling designs used in HYPER.