Speaker
Description
Effective viral transmission network analysis is crucial for controlling virus spread. Cluster size is key for prioritizing interventions. Traditional methods often miss network extent due to sampling biases. This study investigates mean pairwise genetic distance (MGEND) as a proxy for estimating "true" cluster size.
Viral samples from an HIV clinic in Mexico City were classified into clusters using HIV-TRACE. We then conducted a subsampling experiment on large clusters, to analyze the stability of MGEND after subsampling clusters to different levels down to 10% of their original sizes. Additionally, we examined the empirical relationship between cluster size and MGEND across multiple clusters (clusters larger than 10 individuals) using an exponential regression model. Clusters with negative residuals from this first regression were considered undersampled. They then were removed to perform a second exponential regression only on well-sampled clusters (clusters with positive residuals). The empirical equation of the second regression was then used to calculate an expected size of all clusters with negative residuals from the first and second regressions.
MGEND remained stable despite substantial downsampling of the five largest clusters in our dataset (down to 20%), suggesting that MGEND is a reliable proxy for cluster size even when clusters are undersampled. The exponential regression revealed a significant positive correlation (p< 2.14E-06) between cluster size and MGEND. Considering MGEND remained constant despite susampling, it is possible that negative-residual clusters were undersampled. The second exponential regression allowed to calculate an expected size for undersampled clusters.
MGEND is a robust proxy for real cluster size in viral transmission networks, despite undersampling. It offers a novel approach to identifying undersampled clusters and estimating their expected size, via exponential regressions. However, caution must be taken as expected-size estimation assumes that clusters with the largest size-to-MGEND ratios were those well represented. Still, this approach can can assign higher priority to the relatively larger clusters identified; which is crucial for targeted contact tracing and outbreak management.
| Expedited Notification | Yes, I want to opt-in for Expedited Notification |
|---|
