Bayesian reconstruction of HIV transmission trees from viral sequences and uncertain infection times

Hesam Montazeri, Susan J. Little, Niko Beerenwinkel, Victor DeGruttolaarXiv 2018 · DOI · arXiv

Abstract

Genetic sequence data of pathogens are increasingly used to investigate transmission dynamics in both endemic diseases and disease outbreaks; such research can aid in development of appropriate interventions and in design of studies to evaluate them. Several methods have been proposed to infer transmission chains from sequence data; however, existing methods do not generally reliably reconstruct transmission trees because genetic sequence data or inferred phylogenetic trees from such data are insufficient for accurate inference regarding transmission chains. In this paper, we demonstrate the lack of a one-to-one relationship between phylogenies and transmission trees, and also show that information regarding infection times together with genetic sequences permit accurate reconstruction of transmission trees. We propose a Bayesian inference method for this purpose and demonstrate that precision of inference regarding these transmission trees depends on precision of the estimated times of infection. We also illustrate the use of these methods to study features of epidemic dynamics, such as the relationship between characteristics of nodes and average number of outbound edges or inbound edges-- signifying possible transmission events from and to nodes. We study the performance of the proposed method in simulation experiments and demonstrate its superiority in comparison to an alternative method. We apply them to a transmission cluster in San Diego and investigate the impact of biological, behavioral, and demographic factors.

Introduction

In the world of infectious disease, knowing "who infected whom" is essential for public health. Reconstructing a transmission tree allows researchers to identify "superspreaders," evaluate the effectiveness of interventions, and understand the biological drivers of an outbreak. However, while sequencing pathogen genomes has become routine, turning those sequences into a reliable map of transmission remains a massive computational challenge.

A new study by Montazeri et al. presents a Bayesian framework that bridges this gap by combining viral sequences with a frequently overlooked piece of data: the uncertainty of infection times.

The Challenge: Phylogeny vs. Transmission

There is a fundamental disconnect between a phylogenetic tree (which shows the evolutionary history of the virus) and a transmission tree (which shows the history of host-to-host infections). Because viruses continue to evolve within a host before being passed on, a single phylogeny can be consistent with many different transmission scenarios.

Comparison of transmission trees and phylogenies, illustrating that multiple transmission scenarios can correspond to the same phylogeny.

The authors demonstrate mathematically that a single phylogeny with $n$ leaves can correspond to up to $2^{n-1}$ possible transmission trees. Genetic data alone is simply not enough to resolve the "who-infected-whom" question. Existing methods often struggle because they either rely solely on genetic distance or fail to account for the fact that we rarely know the exact moment someone was infected. Clinical data usually provides an "Estimated Date of Infection" (EDI) with a wide margin of error.

Methodology: Integrating Timing and Genetics

The authors developed a novel Bayesian inference method using Markov chain Monte Carlo (MCMC) sampling. The core innovation lies in integrating "noisy" timing data with genetic sequences to produce a more reliable transmission map.

Key Components:

  • Soft Constraints: Instead of assuming fixed infection dates, the model uses infection intervals (e.g., a 95% confidence interval) as "soft constraints" in the prior distribution.
  • Tree Mapping: The researchers created specific algorithms to transform a proposed transmission tree into a timed phylogenetic tree. This allows them to use the standard Felsenstein algorithm to calculate the likelihood of the observed genetic sequences.
A transmission tree (a) and its corresponding phylogenetic tree (c) after mapping.
  • MCMC Sampler: The model explores the space of possible trees using moves such as Subtree-Pruning-Regrafting, child-parent exchanges, updating infection times, and adjusting the viral substitution rate.

Data and Validation

The method was validated and applied using two primary sources:

  1. Simulated Data: Using a Susceptible-Infected-Removed (SIR) epidemic model, the authors generated synthetic outbreaks. They tested various tree sizes (15 to 100 nodes) and varying levels of timing uncertainty ($W$).
  2. San Diego HIV Cluster: A real-world dataset of 19 subjects from a large transmission cluster in San Diego, California, including viral sequences and EDI data based on clinical markers.

Results and Implications

The most significant finding is that even moderately accurate infection intervals drastically reduce the number of possible transmission trees, allowing the model to converge on the truth.

  • High Accuracy: In simulations with precise timing data, the model recovered approximately 90% of the true transmission edges, outperforming popular alternatives like phybreak.
  • Real-World Application: In the San Diego cluster, the MCMC chain identified two primary tree structures.
  • Outdegree as a Metric: By reconstructing the tree, the authors can calculate the outdegree (the number of people one person infected). This allows researchers to statistically test whether certain factors—like a high viral load—actually lead to more transmissions. While the 19-person cluster study didn't find a statistically significant link ($p=0.09$), the method provides a framework for evaluating such hypotheses.

Limitations and Future Directions

The authors acknowledge several simplifying assumptions:

  • The "Closed System" Assumption: The model assumes all individuals in the transmission chain have been sampled, whereas in reality, many links are often missing.
  • Within-Host Diversity: The current model does not account for the complex "cloud" of viral variants (quasi-species) within a patient.
  • Superinfection: The model does not currently account for individuals being infected multiple times by different sources.

By moving beyond simple "genetic distance" and embracing the uncertainty of clinical timing, this work provides a more rigorous way to turn genomic data into actionable public health intelligence.

Summary updated

Abstract

Genetic sequence data of pathogens are increasingly used to investigate transmission dynamics in both endemic diseases and disease outbreaks; such research can aid in development of appropriate interventions and in design of studies to evaluate them. Several methods have been proposed to infer transmission chains from sequence data; however, existing methods do not generally reliably reconstruct transmission trees because genetic sequence data or inferred phylogenetic trees from such data are insufficient for accurate inference regarding transmission chains. In this paper, we demonstrate the lack of a one-to-one relationship between phylogenies and transmission trees, and also show that information regarding infection times together with genetic sequences permit accurate reconstruction of transmission trees. We propose a Bayesian inference method for this purpose and demonstrate that precision of inference regarding these transmission trees depends on precision of the estimated times of infection. We also illustrate the use of these methods to study features of epidemic dynamics, such as the relationship between characteristics of nodes and average number of outbound edges or inbound edges-- signifying possible transmission events from and to nodes. We study the performance of the proposed method in simulation experiments and demonstrate its superiority in comparison to an alternative method. We apply them to a transmission cluster in San Diego and investigate the impact of biological, behavioral, and demographic factors.
BibTeX
@misc{montazeri2018bayesian,
  author = {Hesam Montazeri and Susan J. Little and Niko Beerenwinkel and Victor DeGruttola},
  title = {{Bayesian reconstruction of HIV transmission trees from viral sequences and uncertain infection times}},
  year = {2018},
  eprint = {1801.07660},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1801.07660},
}
Figure