Emma E. Goldberg | Searcharxiv

arXiv2026

Leveraging Synthetic and Genetic Data to Improve Epidemic Forecasting

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 under operational constraints. Specifically, we explore two options likely to be available near the start of an emerging disease outbreak: synthetic data and genetic information. For this investigation, we conducted an experiment where we trained deep learning models on different combinations of real and synthetic data, both with and without genetic information, to explore how these models compare when forecasting COVID-19 cases for US states. All models are developed with an eye towards forecasting the next pandemic. We find that models trained with synthetic data have better forecast accuracy than models trained on real data alone, and models that use genetic variants have better forecast accuracy compared to those that do not. All models outperformed a baseline persistence model (a feat only accomplished by 7 out of 22 real-time COVID-19 cases forecasting models as reported in [38]) and multiple models outperformed the COVIDHub-4_week_ensemble. This paper demonstrates the value of these underutilized sources of information and provides a blueprint for forecasting future pandemics.

stat.AP↗

arXiv2014

Model Inadequacy and Mistaken Inferences of Trait-Dependent Speciation

Species richness varies widely across the tree of life, and there is great interest in identifying ecological, geographic, and other factors that affect rates of species proliferation. Recent methods for explicitly modeling the relationships among character states, speciation rates, and extinction rates on phylogenetic trees- BiSSE, QuaSSE, GeoSSE, and related models - have been widely used to test hypotheses about character state-dependent diversification rates. Here, we document the disconcerting ease with which neutral traits are inferred to have statistically significant associations with speciation rate. We first demonstrate this unfortunate effect for a known model assumption violation: shifts in speciation rate associated with a character not included in the model. We further show that for many empirical phylogenies, characters simulated in the absence of state-dependent diversification exhibit an even higher Type I error rate, indicating that the method is susceptible to additional, unknown model inadequacies. For traits that evolve slowly, the root cause appears to be a statistical framework that does not require replicated shifts in character state and diversification. However, spurious associations between character state and speciation rate arise even for traits that lack phylogenetic signal, suggesting that phylogenetic pseudoreplication alone cannot fully explain the problem. The surprising severity of this phenomenon suggests that many trait-diversification relationships reported in the literature may not be real. More generally, we highlight the need for diagnosing and understanding the consequences of model inadequacy in phylogenetic comparative methods.

q-bio.PE↗