High-definition likelihood inference of genetic correlations across human complex traits
Evaluating the genetic correlations across our phenotypes is of essential importance for understanding disease etiology and other potential causality. A new method—published in the journal Nature Genetics—vastly improves estimates of genetic correlations using established resources from genome-wide association studies.
The research team consisting of Zheng Ning, Yudi Pawitan, and Xia Shen at the Department of Medical Epidemiology and Biostatistics has published a new technique, revealing substantially more genetic correlations between human phenotypes.
The authors derived a full likelihood from the joint distribution of genome-wide genetic effects so that genetic correlation can be estimated with good precision. They compared the new method with the state-of-the-art technique and found that the estimation efficiency is substantially improved, equivalent to a 2.5-fold increase in sample sizes of genomic studies.
The discoveries include significant genetic correlations across a number of human behavioral and disease phenotypes. Based on the UK Biobank resource, the power of the new method revealed about 35% more genetic correlations across 30 human traits. This includes 57 genetic correlations that could not be identified using the state-of-the-art technique, including the genetic correlations between human behaviors such as sleep duration, risk taking, and educational attainment, etc. and complex diseases such as diabetes, depression, and cancer.
The new method uses summary-level instead of individual-level data, obtaining comparable estimates. This vastly speeds up statistical inference for genomic big data. The study powers up future high-throughput genomic studies in large human populations, helping scientists understand the connection between various human complex traits.