Using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain

summary

The gist

H5N1 highly pathogenic avian influenza (HPAI) has been circulating around the globe in previously unseen patterns since 2020 The fit to the final model with the most informative spatial scales shows…

In short

Researchers used machine learning to find which spatial scales—like grid areas or administrative units—best predicted genetic distances between H5N1 virus samples. Findings showed that game bird and wild bird abundance were key predictors in the first period, while farm counts became more important later. This highlights how different environmental scales influence virus spread patterns.

Key concepts

Genetic Distance
This measures how different two influenza virus sequences are from each other, based on nucleotide substitutions. It is used here as a proxy for the distance viruses traveled or the transmission distance between infected individuals.
Spatial Scales of Aggregation
This refers to different ways of grouping data geographically, such as using large grid areas versus smaller administrative zones. The study tests which grouping method best captures the underlying factors driving virus spread.
Random Forest Models
These are sophisticated machine learning tools used to predict outcomes based on various input variables. They were employed here to determine the most important spatial and biological factors influencing the genetic distance between virus samples.

Terminology used across episodes

This episode discusses

The paper

Using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain · Read on arXiv

Roslin Institute, University of Edinburgh · School of Physics and Astronomy, University of Edinburgh

H5N1 highly pathogenic avian influenza (HPAI) has been recently circulating in previously unseen patterns. As the underlying causes are uncertain, we need a better understanding of the drivers of virus circulation, as they underpin the spread, and inform about the impact of HPAI. We combine infection and bird population data, using Random Forest models to quantify the relationship between spatially aggregated values of possible infection risk factors and viral phylogenetic data from HPAI outbreaks to define scales of interactions that influence virus circulation. As genetic proximity is correlated to transmission proximity, the genetic relationships of sequences and relevant variables indicate the scale and resolution of interactions driving transmission. We used H5N1 Sequences from Great Britain divided into two periods: Dec 2020-May 2022, and Jun 2022-Oct 2024. Final model with selected spatial scales shows that for the first period various wild bird abundance variables were the strong predictors of genetic distance, and for the second period, these were geographical distance, and collection date. Variables related to infections have different spatial scales for analysed periods, while variables describing the environment have more consistent scales. The differences in predictors for the periods bears further investigation, however may be related to differences in the virus, areas of spread, or seasonal factors. This is the first phylogenetic study indicating possible role for game birds in the circulation of HPAI and supports experimental studies but must be viewed with caution as our study does not indicate direct causation. We show that the scale of spatial aggregation is variable dependent and reflects the scale of underlying processes. In the future our findings may be used to study dynamics of HPAI and design targeted surveillance and intervention strategies.

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: I'm Ines.

Marcus: And I'm Marcus, here with Yuki.

Ines: We're looking at the paper: "Using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain".

Marcus: Here's the gist of the paper. H5N1 highly pathogenic avian influenza (HPAI) has been circulating around the globe in previously unseen patterns since 2020 The fit to the final model with the most informative spatial scales shows that for the first period game bird abundance and were the strongest predictors of genetic distance between samples Our analyses show that HPAI cases and wild birds’ abundance are highly predictive of genetic distances when aggregated over grid-based areas The differences in predictors between the two periods bears further investigation, however may be related to differences in the virus, geographical areas of spread, and seasonal factors such as the game bird hunting season in the UK Study Objective The study combines infection and bird population data using Random Forest Decision-Tree models to quantify relationships between spatially aggregated values of possible correlates with infection risk at different geographical scales and viral phylogenetic data to define scales of interactions that may influence virus circulation The authors propose a different approach that directly exploits the relationships between genetic (patristic) distance between the virus sequences, as this is highly correlated with the transmission distance between infected individuals They ask what scales of aggregation are most informative and consider whether simple gridded aggregation against aggregation according to administrative units is more relevant as an indicator of whether human, or natural activity is an underlying driver Scale Selection Methodology The selected scales are determined by assuming that the best scale is the one that on average provides aggregated variable that is the most important for the genetic distance determination with highest importance score and results in the best model fit with highest R2 The final model was constructed using variables describing HPAI H5N1 case aggregates, farm counts and wild abundance calculated over selected scales They used three metrics to select the scales of spatially aggregated variables comparing RF runs with variables defined on different scales Key Findings on Drivers The most informative factors for genetic distance prediction were period-dependent; for the first period, game bird abundance and wild bird abundance were the strongest predictors of genetic distance between samples For the second period, geographical distance, but also collection date and farm count identified as most informative In period 1 (December 2021 to May 2022), game bird and raptor abundance appeared important in predicting genetic distances, suggesting a bridging role between observed samples with significant undetected transmission among them In the later period (June-October 2022), farm count played an important role, and areas of higher abundance of gulls and corvids indicated much shorter genetic connections Implications for Intervention The variability of scale across risk factors and time periods illustrates the importance of considering relevant spatial scale in understanding transmission in complex disease ecosystems The methodology developed can be used to study geographical scales of complex pathogen-host-environmental systems like avian influenza combining pathogen sequence data with many variables reported in spatial resolution In the future, these investigations may be used to provide geographical scales that are the most useful to studying dynamics of HPAI spread and the scale for targeted surveillance and intervention strategies Data Integration The analysis integrated HPAI H5N1 sequences from period 03rd December 2020 to 19th October 2022 with all positive HPAI H5N1 cases from the same period, poultry farm data as of December 2022 and wild bird abundance data from eBird platform from year 2021 Data processing involved assigning grid areas (considering all resolutions from H to XXL) and 2011 Census Output Areas to each case The genetic distance between paired sequences was calculated for each pair using the Tamura and Nei 1993 model Temporal Distance Optimization A temporal distance cutoff excluding the sequences that were sampled over 30 days from each other was implemented to exclude pairs most likely because of the time of emergence This threshold was selected based on temporal distance optimization where it aimed to select the model with temporal distance threshold that had the lowest importance without compromising the model fit The results from this genetic distance threshold optimisation are presented on Figure 3 Overall, the methodology developed can be used to study geographical scales of complex pathogen-host-environmental systems like avian influenza combining pathogen sequence data with many variables reported in spatial resolution The variability of scale across risk factors and time periods that we have identified illustrates the importance of considering relevant spatial scale in understanding transmission in complex disease ecosystems In the future the investigations of scales can be implemented to improve spatial sampling strategies, define targeted interventions and the zones they are implemented in, or create more accurate risk maps The study indicates that HPAI cases and wild birds’ abundance are highly predictive of genetic distances when aggregated over grid-based areas References Apley, D.

Ines: (2018).

Marcus: ALEPlot: Accumulated Local Effects (ALE) Plots and Partial Dependence (PD) Plots.

Ines: https://cran.r-project.org/web/packages/ALEPlot/index.html Barroso, P., López-Olvera, J.

Paper summary: Ines: So, we're looking at this paper called Using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain. The main idea here is they want to figure out what spatial scales are most informative when trying to link infection risk to genetic distance between virus samples.

Marcus: Right, so the core thesis is that they combine infection data with bird population data using Random Forest models and viral phylogenetic data to define these interaction scales and see how they predict spread. It’s about finding the right way to look at things geographically when studying this virus.

Yuki: From a population genetics standpoint, it’s interesting because H5N1 has been spreading in previously unseen patterns since two thousand twenty and this study tries to give us a better understanding of those drivers for the spatiotemporal patterns of circulation that underpin the global spread <ref:2411.10424#pg1,in previously unseen patterns since 2020>.

Ines: Exactly. They are using these models to quantify the relationship between spatially aggregated values of possible correlates with infection risk at different geographical scales, while also defining scales based on viral phylogenetic data that might influence virus circulation.

Marcus: The paper claims they found period-dependent predictors, meaning what drives the genetic distance prediction changes depending on when you look at the data. For instance, for the first period they studied, game bird abundance and wild bird abundance were the strongest predictors of genetic distance between samples.

Yuki: That suggests that in that initial phase, having information about where those birds are abundant really helped us connect different virus samples through their genetic distance. What does that imply about how the virus moves across those landscapes?

Ines: It suggests a bridging role between observed samples with significant undetected transmission among them, because game bird and raptor abundance appeared important in predicting genetic distances during that first period.

Marcus: Then they shift focus for the later period, June to October two thousand twenty-two and farm count played an important role <ref:2411.10424#pg1>. They also found that areas with higher abundance of gulls and corvids indicated much shorter genetic connections there.

Yuki: So, it seems like the relevant ecological factor shifts over time—from bird abundance early on to farm activity later—which makes sense if you're looking at how different parts of the ecosystem interact with the virus.

Ines: The paper is interested in whether simple gridded aggregation against aggregation according to administrative units is more relevant as an indicator of whether human or natural activity is the underlying driver.

Marcus: They tested that by comparing their models, and they found that while the R squared value describes the fit of the whole model rather than just one variable, generally the differences from their random forest models were quite small.

Yuki: That's a bit of a caution there. It sounds like while they found interesting predictors, it’s important to remember that changing only one abundance variable doesn't necessarily mean you have the best fit for the whole picture.

Ines: Right, because as for abundance variables they changed the scales of all abundance species group variables together and that R squared change doesn't indicate the goodness of fit when only one of the abundance variables is changed. They also used three metrics to select these scales and found that the R squared was the most variable, while importance based metrics were more consistent.

Marcus: That tells us that relying on just one perfect spatial scale might be too narrow; you need to look at how different aggregation methods perform across all those variables.

Yuki: I wonder what this means for us in terms of real-world monitoring, since the paper is looking at risk mapping for avian influenza in Great Britain. It suggests that our approach to defining geographical scales needs to be flexible based on the time period and the specific drivers we are interested in.

Ines: Precisely. This whole study is about using sequence data to study spatial scales of interactions driving spread of Highly Pathogenic Avian Influenza in Great Britain, and it shows that the interaction structure isn't static across different ecological factors or time frames.

Marcus: So, to wrap up on this paper, the authors are pointing toward a more nuanced view where the most useful spatial scale depends entirely on what you are trying to predict—whether it's infection risk based on bird density or farm activity.

Yuki: It really highlights that understanding H5N1 spread requires integrating those different ecological perspectives across different time frames, which is crucial for long-term risk assessment for both wild and domestic poultry populations.

Ines: Yeah, the implications are that we can use this methodology to study geographical scales of complex pathogen host environmental systems, combining sequence data with many variables reported in spatial resolution. It helps define the best scales for targeted surveillance and intervention strategies moving forward.

Conclusion: Ines: So, we’re wrapping up our look at this paper about Highly Pathogenic Avian Influenza in Great Britain by looking at sequence data and spatial scales of interaction.

Marcus: Yeah, the authors are using Random Forest models to link infection risk to genetic distance between virus samples, trying to figure out which geographical scales matter most.

Ines: The conclusion they draw is that the way you define those spatial scales really depends on what you're looking at and when you look at it.

Marcus: They found that for one time period, like the first one they studied, bird abundance—both game birds and wild birds—were the biggest drivers for predicting genetic distance between samples.

Ines: That suggests that in that initial phase, knowing where those specific animals are abundant really helped connect different virus samples through their genetic distance.

Marcus: But then things shift for a later period, like June to October twenty twenty two, and farm count becomes more important in predicting those distances.

Ines: So the key finding is that the most informative factor isn't fixed; it changes based on the time period you analyze.

Marcus: Right, and they also pointed out that these differences might be because of different virus strains or maybe seasonal factors like when people are hunting game birds in places like the UK.

Ines: The paper suggests we can use this method to study how complex systems, like an influenza outbreak, work spatially by combining sequence data with many different variables reported at various resolutions.

Marcus: It gives us a framework for figuring out which geographical scales are most useful when we want to set up targeted surveillance or intervention strategies for these outbreaks.

Ines: So the next thing we need to look at is how this study can specifically help define those actionable zones for stopping the spread of this virus.

More episodes

← Home