TY - JOUR
T1 - The relationship between population attributable fraction and heritability in genetic studies
AU - Wang, Tao
AU - Hosgood, H. Dean
AU - Lan, Qing
AU - Xue, Xiaonan
N1 - Funding Information:
This research was supported, in part, by R21 CA202529 (to TW) and by the Albert Einstein Cancer Center Biostatistics Core Grant 5P30CA013330-45 (to TW and XX).
Publisher Copyright:
© 2018 Wang, Hosgood, Lan and Xue.
PY - 2018/10/1
Y1 - 2018/10/1
N2 - Population attributable fraction (PAF) has been widely used to quantify the proportion of disease risk in a population that can be attributed to risk factors in epidemiological studies. However, the use of PAF has been limited in assessing the contribution of genetic variants. Most notably, the PAF estimate is typically much larger than other commonly used measures, such as heritability, thereby raising the concern that PAF may overestimate the genetic contribution. In this paper, we show that PAF is a one-to-one function of heritability, and explain why PAF is larger than heritability. Further, we present an estimation procedure based on the summary statistics from genome-wide association studies (GWAS) to estimate the PAF of multiple correlated genetic variants for a binary outcome. Currently available estimation procedures only apply to a single variant or to multiple genetic variants that are independent from each other. Our simulation studies verified the relationship between PAF and heritability, and showed that the proposed estimation procedure for these two measures performed well. Finally, we applied the proposed method to the published data of two lung cancer GWAS to estimate the PAF and heritability of several newly identified variants. Our results demonstrate that the PAF estimate is a useful measure of the genetic contribution to the development of the disease. We hope this paper serves as an advocate for a wider use of PAF in genetic studies.
AB - Population attributable fraction (PAF) has been widely used to quantify the proportion of disease risk in a population that can be attributed to risk factors in epidemiological studies. However, the use of PAF has been limited in assessing the contribution of genetic variants. Most notably, the PAF estimate is typically much larger than other commonly used measures, such as heritability, thereby raising the concern that PAF may overestimate the genetic contribution. In this paper, we show that PAF is a one-to-one function of heritability, and explain why PAF is larger than heritability. Further, we present an estimation procedure based on the summary statistics from genome-wide association studies (GWAS) to estimate the PAF of multiple correlated genetic variants for a binary outcome. Currently available estimation procedures only apply to a single variant or to multiple genetic variants that are independent from each other. Our simulation studies verified the relationship between PAF and heritability, and showed that the proposed estimation procedure for these two measures performed well. Finally, we applied the proposed method to the published data of two lung cancer GWAS to estimate the PAF and heritability of several newly identified variants. Our results demonstrate that the PAF estimate is a useful measure of the genetic contribution to the development of the disease. We hope this paper serves as an advocate for a wider use of PAF in genetic studies.
KW - GWAS
KW - Genetic epidemiology
KW - Heritability
KW - Population attributable risk
KW - Summary statistics
UR - http://www.scopus.com/inward/record.url?scp=85055316662&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85055316662&partnerID=8YFLogxK
U2 - 10.3389/fgene.2018.00352
DO - 10.3389/fgene.2018.00352
M3 - Article
AN - SCOPUS:85055316662
SN - 1664-8021
VL - 9
JO - Frontiers in Genetics
JF - Frontiers in Genetics
IS - OCT
M1 - 352
ER -
