Testing the Impact of Trait Prevalence Priors in Bayesian-Based Genetic Prediction Modeling of Human Appearance Traits

The study examined the impact of using trait prevalence-informed priors in Bayesian-based genetic prediction models for human appearance traits, such as eye, hair, and skin color, hair structure, and freckles. It found that incorporating these priors could improve prediction accuracy, but the effect varied across different traits and categories. Mis-specified priors often reduced accuracy compared to models without priors. The study highlighted the challenge of limited data on trait prevalence, which made the practical use of such priors infeasible at the time. Accurate specification of prevalence-informed priors was crucial for enhancing prediction models, but the lack of comprehensive prevalence data posed a significant limitation. The research emphasized the need for unbiased estimates of trait prevalence across diverse populations and suggested that future studies should focus on identifying more causal genetic factors to improve model performance.