Abstract
Assays using ELISA measurements on serially diluted serum samples have been heavily used to measure serum reactivity to SARS-CoV-2 antigens and are widely used in virology and elsewhere in biology. We test a method using Bayesian hierarchical modelling to reduce the workload of these assays and measure reactivity of SARS-CoV-2 and HCoV antigens to human serum samples collected before and during the COVID-19 pandemic. Inflection titers for SARS-CoV-2 full-length spike protein (S1S2), spike protein receptor-binding domain (RBD), and nucleoprotein (N) inferred from 3 spread-out dilutions correlated with those inferred from 8 consecutive dilutions with an R2 value of 0.97 or higher. We confirm existing findings showing a small proportion of pre-pandemic human serum samples contain cross-reactive antibodies to SARS-CoV-2 S1S2 and N, and that SARS-CoV-2 infection increases serum reactivity to the beta-HCoVs OC43 and HKU1 S1S2. In serial dilution assays, large savings in resources and/or increases in throughput can be achieved by reducing the number of dilutions measured and using Bayesian hierarchical modelling to infer inflection or endpoint titers. We have released software for conducting these types of analysis.