Lectures and seminars PhD student seminar: Yuying Li

Abstract

Genetic correlation is a key parameter in understanding the shared genetic architecture of complex traits. However, existing methods typically provide only global estimates, potentially overlooking important regional variations. We developed HDL-L, an extension of the high-definition likelihood (HDL) approach, to address these limitations by performing genetic correlation analyses in small, approximately independent linkage disequilibrium (LD) blocks. Compared to the state-of-the-art method LAVA, HDL-L not only yields more consistent heritability estimates but also provides more efficient genetic correlation estimates, with a significantly lower risk of false positives and superior computational performance. In an analysis of 30 UK Biobank phenotypes, HDL-L identified 109 significant local genetic correlations, reflecting its high resolution and robustness.
To further demonstrate its utility, HDL-L can be used for regional genetic colocalization analyses. HDL-L outperformed COLOC and SuSiE in detecting genetic colocalization, as evidenced by higher AUC in simulation analysis and higher rediscovery rates in real data analysis. When applying HDL-L on cis-pQTL regions for 2,826 plasma proteins and 200 diseases, we identified 76 reliable protein targets for the diseases.
Collectively, we showcase HDL-L as a powerful and efficient solution for fine-scale genetic correlation and colocalization analyses, accelerating discoveries in personalized medicine, drug development, and broader genetic research.