A Broad Class of Efficient Tests of Association That Combine Family-Based and Case-Control Tests of Association

• Activity Number: 277
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Jane Cerise*+ and William C.L. Stewart
• Companies: Columbia University and Columbia University
• Keywords: Association ; Exome Sequencing ; Meta-Analysis ; GWAS

Abstract

A common approach to the study of complex traits is to analyze the genotypes of affected families and ethnically matched controls. In this setting, researchers often report a p-value for a test of no segregation distortion, as well as a p-value for a test of equal genotype frequencies among cases and controls. However, if both tests use a shared set of affecteds, these p-values are correlated and the overall evidence for association is unclear.

To address these problems, we propose a class of powerful tests of association. Specifically, Z-mixed is the scaled sum of Z-seg (e.g. TDT) and Z-freq (e.g. trend test). Since the scaling factor in Z-mixed accounts for the correlation, Z-mixed provides a clear and accurate summary of the overall evidence for association. Using simulated data, we show that Z-mixed is more powerful than Z-seg or Z-freq alone; that our analytic expression for the scaling factor is correct; and that the type I error of Z-mixed is controlled. To demonstrate the utility of our approach, we applied Z-mixed to 18 hypertension families and 113 HapMap controls. Our new method is implemented in the software package EAGLET, and is freely available from the web.