# A joint Bayesian hierarchical model for estimating SARS-CoV-2 diagnostic and subgenomic RNA viral dynamics and seroconversion

**Authors:** Tracy Q. Dong, Elizabeth R. Brown  
**arXiv:** [2301.03714](https://arxiv.org/abs/2301.03714)  
**Date:** 2023-01-09  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2301.03714v1/

## Key points

- Understanding viral dynamics and immunizing antibodies of SARS-CoV-2 is crucial for better therapeutic and prevention strategies for COVID-19.
- The study presents a Bayesian hierarchical model that estimates diagnostic RNA viral load, subgenomic RNAs (sgRNA) viral load, and the rate and timing of seroconversion.
- The model considers the relationship between viral load and antibody data, identifies potential correlates of viral load characteristics and propensity for seroconversion.
- The model is applied to data from the COVID-19 PEP study, which compared hydroxychloroquine to placebo-like control for prevention of SARS-CoV-2 infection.
- No differences were found between hydroxychloroquine and placebo-like control in preventing SARS-CoV-2 infection.
- The study describes the model formulation and inference procedure in detail in Section 3, analyzes data from the COVID-19 PEP study in Section 4, and summarizes results.
- A cross validation exercise demonstrates the model's ability to impute sgRNA viral trajectories for individuals with only diagnostic viral load data.
- This comprehensive Bayesian hierarchical model provides valuable insights into virus dynamics and antibody response in individuals exposed to SARS-CoV-2.

## Summary

Understanding the viral dynamics and immunizing antibodies of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is crucial for devising better therapeutic and prevention strategies for COVID-19. In this study, the authors present a Bayesian hierarchical model that jointly estimates the diagnostic RNA viral load reflecting genomic materials of SARS-CoV-2, the subgenomic RNAs (sgRNA) viral load reflecting active viral replication, and the rate and timing of seroconversion reflecting presence of antibodies. The proposed method takes into account the dynamical relationship and correlation structure between the two types of viral load, allows for borrowing of information between viral load and antibody data, and identifies potential correlates of viral load characteristics and propensity for seroconversion. To demonstrate the features of the joint model, the authors apply it to data collected in the COVID-19 PEP study. The COVID-19 PEP Study was a double-blinded, household-randomized controlled trial comparing hydroxychloroquine to placebo-like control for prevention of SARS-CoV-2 infection. The study enrolled 829 initially asymptomatic household contacts and healthcare workers recently exposed (<96 hours) to persons with laboratory-confirmed SARS-CoV-2 infection from 41 U.S. states. The study found no differences between hydroxychloroquine and placebo-like control arms in preventing SARS-CoV-2 infection. The authors describe their model formulation and inference procedure in detail in Section 3. They then apply their methods to analyze data from the COVID-19 PEP study in Section 4 and summarize their results. They also conduct a cross validation exercise to illustrate the model's ability to impute sgRNA viral trajectories for individuals who only had diagnostic viral load data. Overall, this study presents a comprehensive Bayesian hierarchical model that integrates multiple aspects of SARS CoV 2 dynamics and seroconversion by applying it to real world data which provides valuable insights into understanding virus dynamics as well as antibody response in individuals exposed to SARS CoV 2 which can inform more effective therapeutic or preventive strategies against COVID 19.