Leveraging synthetic and genetic data to improve epidemic forecasting
Dave Osthus, Alexander C. Murph, Emma E. Goldberg, Lauren J. Beesley, William M. Fischer, Nidhi Parikh, Lauren A. Castro
Abstract
Forecasting infectious disease outbreaks is hard. Forecasting emerging infectious diseases with limited historical data is even harder. In this paper, we investigate ways to improve emerging infectious disease forecasting when little pathogen-specific training data are available. Specifically, we explore two sources of information that may be available near the start of an emerging disease outbreak: synthetic data and genetic information.
Introduction
Over the past decades, infectious disease forecasting has transformed from an academic curiosity into a critical tool for public health preparedness and response. This growth has been driven by advances in modeling [2–4] and data availability [5], and by the recognition that timely forecasts can guide resource allocation, inform policy, reduce the cost burden associated with emerging health threats, and improve public health outcomes [6].
Materials and method
This section provides details on the study design, data sources, synthetic data generation process, forecasting model, and evaluation metrics used in the results above.
Results
We evaluated weekly 1–4 week ahead forecasts of COVID-19 cases for U.S. states and Puerto Rico from June 2020 through December 2022. The study used COVID-19 as a retrospective test case for pandemic forecasting models trained without COVID-19 data, allowing us to evaluate whether historical respiratory disease data, synthetic outbreak data, and SARS-CoV-2 variant information could improve forecasts under emerging-outbreak constraints. Full study-design details are provided in Section 4.
Discussion
In this paper, we sought to answer two high-level questions: (1) Can synthetic data be used to train deep learning models that achieve state-of-the-art infectious disease forecasting performance? and (2) Can genetic information be used to improve infectious disease forecasting models? We conducted an exercise to answer these questions with an eye towards the next pandemic by limiting ourselves to only using training data available prior to January 1st, 2020 for forecasting COVID-19 cases. We affirmatively answered both questions. Synthetic data was a valuable training data source, as models trained on only synthetic data provided impressive forecasting performance.
Acknowledgments
We gratefully acknowledge all data contributors, i.e., the authors and their originating laboratories responsible for obtaining the specimens, and their submitting laboratories for generating the genetic sequence and metadata and sharing via the GISAID Initiative, on which some of this research is based. Research presented in this article was supported by the Laboratory Directed Research and Development program of Los Alamos National Laboratory under project number 20240066DR.
Citation: Osthus D, Murph AC, Goldberg EE, Beesley LJ, Fischer WM, Parikh N, et al. (2026) Leveraging synthetic and genetic data to improve epidemic forecasting. PLoS Comput Biol 22(8): e1014630. https://doi.org/10.1371/journal.pcbi.1014630
Editor: Benjamin M. Althouse, University of Washington, UNITED STATES OF AMERICA
Received: March 18, 2026; Accepted: July 24, 2026; Published: August 27, 2026
Copyright: © 2026 Osthus et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data and code used to perform the analysis presented in this paper can be found here: https://github.com/lanl/precog/tree/main/synthetic_and_genetic_forecasting.
Funding: This work supported by the Laboratory Directed Research and Development program (https://www.lanl.gov/science-engineering/science-programs/ldrd) of Los Alamos National Laboratory (20240066DR to DO, AM, EG, LB, WF, NP, LC). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
