Genomic Surveillance Algorithm Improves Early Detection of Emerging Variants

Existing analytic pipelines often struggle to scale across regions while preserving critical transmission links, slowing response during outbreaks. Lower-resource settings are disproportionately affected by these limitations. A new study shows a scalable algorithm that accelerates and lowers the cost of genomic surveillance.

Researchers at Texas A&M University developed the Iterative Block Particle Filter (IBPF), a computational framework designed to optimize global genomic surveillance. The framework aims to strengthen local, community-based sequencing capacity while preserving the regional interactions needed to track variant spread. The work was carried out with collaborators from multiple international institutions.

IBPF builds on the particle filter, also known as the Sequential Monte Carlo (SMC) method, which is widely used but hampered by the “curse of dimensionality” when applied to high-dimensional spatial networks. The new approach uses an iterative scheme in which the output from one data block becomes the input for the next, enabling localized updates without fragmenting critical connections between neighboring regions and travel hubs. By maintaining dependencies across spatial units while controlling filtering error locally, IBPF scales to high-dimensional spatiotemporal data common in pathogen surveillance.

In the study, the team implemented large-scale multi-strain models using real-world inputs, including epidemiological records, vaccine information, and high-resolution international air travel data. The framework tracked dozens of regions and multiple viral strains and outperformed common filtering algorithms. Notably, it reduced the time between detecting a disease variant and sequencing it. 

The results indicate that surveillance resources can be prioritized toward major international travel hubs, a strategy the authors note may be especially useful where long-term, large-scale sequencing efforts are not feasible. The paper, “Optimizing global genomic surveillance for early detection of emerging SARS-CoV-2 variants,” was published in Nature Communications.

Beyond COVID-19, the authors state that IBPF is compatible with more general systems and could support broader surveillance and forecasting applications. They also note the methodology’s relevance to other dynamic networks in which complex interactions evolve over time. The research group has made its code publicly available.

“In the machine learning community, many algorithms are designed to perform well only on specific test data sets. Our approach is different because we have theoretical performance guarantee that it works reliably across broad spatiotemporal data,” said Patricia Ning, assistant professor in the Department of Statistics at Texas A&M University.

“If we can optimize the allocation, we can spend money more efficiently. That would allow us to detect new variants much earlier without increasing the budget,” said Jifan Li, doctoral candidate in the Department of Statistics at Texas A&M University.

Related Links
Texas A&M University