Description
AbstractMitigating transmission of SARS-CoV-2 has been complicated by the inaccessibility and, in some cases, inadequacy of testing options to detect present or past infection. Immunochromatographic lateral flow assays (LFAs) are a cheap and scalable modality for tracking viral transmission by testing for serological immunity, though systematic evaluations have revealed the low performance of some SARS-CoV-2 LFAs. Here, we re-analyzed existing data to present a proof-of-principle machine learning framework that may be used to inform the pairing of LFAs to achieve superior classification performance while enabling tunable False Positive Rates optimized for the estimated seroprevalence of the population being tested.