Hakhamanesh Mostafavi

This paper introduces PGSUS, a method to assess the influence of stratification and assortative mating effects on polygenic scores, revealing that direct genetic effects often contribute little to PGS variation.

This paper compares how GWAS and rare variant burden tests prioritize genes for 209 traits in the UK Biobank, showing they focus on different genes due to distinct prioritization criteria.

This paper introduces a Bayesian framework combining population genetics and machine learning to estimate gene constraint, improving accuracy for short genes and providing a flexible tool for gene-level property estimation.

The paper introduces conditional frequency spectra to study selection on complex traits in biobanks, finding evidence for stabilizing or purifying selection on GWAS-associated variants.

The paper develops a probabilistic model to explain genetic architectures of complex traits using scaling laws, finding that traits share similar selection coefficients but differ in scaling parameters.

The paper tests for genetic interactions affecting causal variant effect sizes in gene expression and LDL-C across European and African American populations, finding significant evidence in gene expression but not LDL-C.

GWAS hits (44 UK Biobank traits) and GTEx v8 cis-eQTLs (38 tissues) differ systematically: eQTLs cluster near transcription start sites of weakly constrained genes; GWAS hits favour constrained, multifunctional genes.

The paper introduces a Bayesian model comparison approach, MRP, for rare variant association studies using summary statistics, demonstrating improved power and discovery of gene-trait associations across multiple phenotypes.

The paper shows that polygenic scores have varying prediction accuracy within a single ancestry group based on factors like age, sex, and GWAS study design.

The paper explains that population genetic measures like pLI reflect selection on heterozygotes, not dominance or haploinsufficiency.