Quantifying Sample Collection Time Uncertainty Improves Epidemiological Modeling of SARS-CoV-2 Evolution

Viral sequence data has been used extensively to monitor SARS-CoV-2 evolution over the course of the Covid-19 pandemic. Reliable genomic surveillance data is needed to gain insights into patterns of SARS-CoV-2 evolution, to inform public health measures, and to aid in vaccine developments. However, reported Covid-19 cases in public databases rarely record precise infection times, which can introduce uncertainty in epidemiological and statistical modeling. To address this issue, we adapted a back projection algorithm to better estimate the underlying progression of SARS-CoV-2 evolution over time. We use this algorithm to infer the duration between individual infection and sampling events. We integrated this approach with a statistical model to provide robust estimates of natural selection for variants with increased transmission. Our approach can handle sparsely sampled datasets and is therefore useful for analyzing regions where sampling is infrequent. As the virus becomes endemic in our populations and countries scale back on pandemic surveillance measures, approaches like these become increasingly important tools for effective viral surveillance. In this talk I will present results from our simulations that validate this method, as well as applications to real-world viral case data. This method could aid in the reliable detection of SARS-CoV-2 variants of concern with higher transmission rates.