Back projection methods for the refinement of SARS-Cov-2 sequence data

As the SARS-CoV-2 virus continues to evolve, viral sequencing data has extensively been used to identify novel genome mutations. Reliable genomic surveillance data is needed in order to identify viral variants of concern. However, individual symptom onset times and infection times are absent from this surveillance data, which often includes only sequencing dates. These delays in reporting can introduce bias in epidemiological models. 

To address this issue, we apply an adapted back projection to our time series data. Back projection was originally developed to estimate the incidence of HIV from AIDS case data, and can infer un-observable features of an event leading to a disease outbreak, such as the period of time between individual infection and diagnosis. 

This approach is used in our statistical model to give a more reliable estimate for distributions of infection times, and improves the quality of genomic surveillance data in regions where sampling is rare or infrequent. Our method therefore aids the detection of variants with concerning traits, such as higher rates of transmission.