In this competition workshop, participants will address the challenge of selecting and parameterizing models for complex biological systems. Biological systems present unique hurdles due to their complexity and variability. Even in the face of this complexity, methods to predict the future state of biological systems (and populations) from limited time series data is an ongoing challenge in medical and public heath areas.

The workshop will emphasize real-time forecasting as new data becomes available. For example, forecasting disease progression requires incorporating new data as it becomes available. Understanding disease spread within a population necessitates real-time tracking of case numbers and severity. As public health interventions are implemented, forecasts must be adjusted for real-time changes.

Data collection poses a significant challenge, as continuous monitoring is rarely feasible. Some variables cannot be measured directly, and biological data are often noisy, complicating model precision and reliability.

Prior to the competition, participants will develop any model they wish that is capable of modeling a viral pandemic. During the competition, time series data on infection occurrence will released sequentially. The participants' challenge is to predict the future course of the infection in the population. 

Influenza test case

We use yearly influenza A epidemic as a case study to model the spread of an acute infection within a population. Influenza is a recurring infection that results in tens of thousands of hospitalizations and between 5,000 and 50,000 deaths yearly in the USA [Wu 2011, Kandula 2019, Brauer 2019, Kalchev 2024]. The influenza virus mutates rapidly, and epidemiologic data suggests that infection with one strain may only temporarily, on a scale of a few months, protect from the second strain [Ferguson 2003]. The yearly strains of influenza vary widely in both their infectivity (e.g., their rate of spread) and the severity of the disease (e.g., the number of cases severe enough to require hospitalization or lead to death). Health organizations update influenza vaccines yearly attempting to predict the upcoming year’s most infectious and most common strains. The effectiveness of the vaccines, as well as the percentage of the population vaccinated, varies widely from year to year but is usually below 50% [CDCb 2024]. This combination of factors makes each influenza season unique. There are worldwide efforts to predict the severity of the yearly influenza epidemic. A key challenge arises once the influenza season begins: using early, limited data to forecast its future course. These predictions are crucial for helping public health officials prepare and respond effectively to the evolving yearly disease course.