Consistency of genetic analyses in longitudinal data: observations from the GAW13 Framingham Heart Study data.

Abstract

This paper examines the consistency of genetic analyses across time, both in the context of replicating results from one data collection point to the next, and from the perspective of modeling longitudinal processes. This summary originates from the examination of findings from nine papers from Genetic Analysis Workshop (GAW) 13 that reported on analyses of longitudinal data of a variety of traits from the Framingham Heart Study. These analyses include both assessments of consistency of aggregate genetic effects, in the form of estimation of heritability and relative risk of disease, as well as localization of quantitative trait loci (QTLs) by genome-wide linkage screens. Consistency varied widely by trait, possibly reflecting differences in measurement error, secular trends, or underlying biological features such as genotype x age interaction. Quantitatively, comparing magnitudes of estimates across age or time, heritability estimates showed greater consistency than LOD scores. However, qualitatively, the same regions of interest were often identified in genome scans from different time points or different ages. Estimates of sibling recurrence risk, on the other hand, showed little consistency. Heritabilities were greater when participants were matched by age than when they were matched by date of examination. Multivariate approaches, either in use of multiple traits or in use of multiple measures of the same trait, appeared to provide stronger genetic signals both for relative risk and for linkage. Finally, modeling of longitudinal processes provided evidence for genotype x age interactions that may partially explain variation in results of genetic analyses across time or age.