Integrative genetic risk prediction using non-parametric empirical Bayes classification.

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Genetic risk prediction is crucial for individualized medicine but limited by small sample sizes in complex diseases.
• Integrating data from existing studies can increase effective sample size, but accessing individual-level genotype data is often challenging, especially for rare diseases.

Purpose of the Study:

• To propose a novel approach for integrative genetic risk prediction of complex diseases with binary phenotypes.
• To develop a method that can leverage auxiliary study summary statistics without requiring individual-level data.

Main Methods:

• A tuning parameter-free, non-parametric empirical Bayes procedure is introduced to accommodate potential heterogeneity between target and auxiliary diseases.
• The method is designed to be trained using only summary statistics from auxiliary studies.

Main Results:

• Simulation studies demonstrated that the proposed integrative method outperforms non-integrative and existing integrative classifiers in predictive accuracy.
• Application to pediatric autoimmune diseases showed substantial reduction in prediction error for specific target/auxiliary disease pairings.

Conclusions:

• The developed method offers a powerful approach to enhance genetic risk prediction for complex diseases by effectively integrating external summary statistics.
• The method is implemented in the R package 'ssa', facilitating its application in genetic research.