Sandra Bohn

This paper develops and evaluates ensemble polygenic risk score models for disease prediction, showing improved accuracy and portability across diverse populations when combined with clinical features.

Selphi improves genotype imputation accuracy, especially for rare variants, enhancing GWAS discovery and polygenic risk scores.

Orchestra improves local ancestry inference, revealing admixture patterns in Latin Americans, Ashkenazi Jews, and selection signals in British populations.

This study benchmarks phasing and imputation software using three datasets, finding Beagle 5.2 most accurate for common markers and IMPUTE5 for low-frequency markers, while also creating an automated pipeline for customized chip design.