Parameter Estimation in Disease Forecasting for Ecological Models

[Janeteey]

Keywords: Disease Ecology, Markov Chain Monte Carlo (MCMC), Partially Observed Markov Processes (POMP), Non-negative Binomial Distribution, Forecasting, West Nile Virus (WNV).

Developing models to understand the transmission of pathogens in disease ecology is a critical field of study that helps us understand the spread of diseases and how to prevent them. Mathematical modeling is an important tool in epidemiology, prediction, and prevention as it can be used to simulate the spread of disease and predict the effectiveness of different control strategies. Epidemiological and ecological problems, as well as other scientific disciplines, have increased the need to connect mathematics to real-life situations through mathematical modeling using ecological models. It is very important because it allows to describe the evolution of a real-life situation (epidemics, ecological problems, disease, etc.) in a population or an ecological system, a human or animal body, and to test several control measures that can be envisaged. Modeling the disease risk based on historical surveillance and contemporary environmental data and forecasting future risk using predictive models and continued epidemiological analysis.

In this project, we plan to use Ordinary Differential Equations (ODEs) to describe the rate at which individuals flow between states (a simple SIR model) to forecast a vector-borne disease (West Nile Virus). To use the right parameters in the model fitting to the data (climate data from NEON and mosquito abundance data). In this project, we will consider two parameter estimation inferences, which are: 1. Partially Observed Markov Processes (POMP) and 2. MCMC with a non-negative binomial distribution.

References: https://www.mdpi.com/1424-2818/14/5/320 https://www.researchgate.net/publication/237105745_INTRODUCTION_TO_POMP_INFERENCE_FOR_PARTIALLY-OBSERVED_MARKOV_PROCESSES https://www.health.ny.gov/diseases/west_nile_virus/fact_sheet.htm https://en.wikipedia.org/wiki/West_Nile_fever https://www.cdc.gov/westnile/index.html https://reu.ecology.uga.edu/wp-content/uploads/2016/08/Waring.pdf https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7307980/ https://www.birs.ca/workshops/2017/17w5107/files/1_Monday_Jennifer%20Hoeting.pdf https://www.biorxiv.org/content/10.1101/125880v1.full.pdf

[kayodeoshinubi]

One thing we can add to this project is to study the effect of climate change on mosquitoes abundance using temperature and how this affect the population growth rate.

[Janeteey]

Nice. On Tuesday, May 30, 2023 at 09:19:18 PM MDT, Kayode Oshinubi @.***> wrote:

One thing we can add to this project is to study the effect of climate change on mosquitoes abundance using temperature and how this affect the population growth rate.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

[rqthomas]

One of the first steps in this project is to organize the NEON mosquito data into a clean time series ("targets"). The process of doing this would effectively be starting a new theme for the NEON Ecological Forecasting Challenge which would be awesome (https://doi.org/10.1002/fee.2616).

[nonlinearnature]

Would be interested to think of non-parametric / data-driven approach to compliment the ODE approach, ala https://onlinelibrary.wiley.com/doi/full/10.1111/ele.13652

[noamross]

@nonlinearnature Happy to discuss, we have a few projects of this sort, that use nonparametric models (boosted trees/BARTs and GAMs) to forecast disease outbreak probability (spatially) and the progress of outbreaks, in one case also using ODEs also to compare/ensemble.

[nonlinearnature]

To iterated, I'd really be interested to see if/how we could extract equivalent insights from ML forecasts about parameters/rates rather than just an equivalent forecasting end-product.

[kayodeoshinubi]

One of the first steps in this project is to organize the NEON mosquito data into a clean time series ("targets"). The process of doing this would effectively be starting a new theme for the NEON Ecological Forecasting Challenge which would be awesome (https://doi.org/10.1002/fee.2616).

Great suggestion.

[kayodeoshinubi]

@nonlinearnature That is a good idea

[kayodeoshinubi]

@nonlinearnature Happy to discuss, we have a few projects of this sort, that use nonparametric models (boosted trees/BARTs and GAMs) to forecast disease outbreak probability (spatially) and the progress of outbreaks, in one case also using ODEs also to compare/ensemble.

@noamross I am also happy to discuss this approach.

[rqthomas]

It might be useful to invite (either in person or virtually) the NEON point-person for the mosquitoes. @sokole who is this person?

[sokole]

@rqthomas That would be Sara Paull, but I think she might be out of town at a conference next week.

[rqthomas]

@sokole and I chatted with Sara Paull and she shared this information about the mosquito data product

There are very few positive West Nile virus pools (55 total, ~30 come from the STER site in 2015). For that reason the mosquito pathogen data product may look different in future years.

[kayodeoshinubi]

@sokole @rqthomas perhaps we can look into ticks data depending on what we can lay our hands on.