Estimating disease prevalence in large datasets using genetic risk scores

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Clinical classification is crucial for disease prevalence estimation but often complex and resource-intensive.
• Advancements in population-level genetic data offer opportunities for novel stratification methods.
• Genetic risk scores (GRS) are increasingly available and relevant for health research.

Purpose of the Study:

• To propose and evaluate a generalizable mathematical framework for disease prevalence estimation using genetic risk scores.
• To compare the performance of different GRS distribution-based methods for prevalence estimation.
• To assess the robustness of genetic stratification for various disease and cohort characteristics.

Main Methods:

• Developed a mathematical framework for disease prevalence estimation utilizing GRS.
• Compared methods including distribution means, Earth Mover's Distance, kernel density estimates, and an Excess method.
• Evaluated performance across different disease prevalences, cohort sizes, and GRS discriminative power.

Main Results:

• Genetic stratification effectively produces robust disease prevalence estimates.
• Accurate estimates are achievable even for rare diseases, smaller cohorts, and less discriminative GRS.
• The proposed framework demonstrates general utility and resilience.

Conclusions:

• Genetic stratification offers a powerful, unbiased alternative to traditional clinical classification for prevalence estimation.
• These techniques provide valuable insights into disease prevalence and characteristics.
• The approach is broadly applicable, enhancing epidemiological research capabilities.