Genome-wide association study for empirically derived metabolic phenotypes in the Framingham Heart Study offspring cohort

We used data reduction and clustering methods to identify five phenotypically homogeneous groups of study participants with similar profiles for cardiovascular disease risk factors. We constructed both qualitative (binary subgroup membership) and quantitative traits (probability of subgroup membership) for each individual. The Cluster 1 comprised individuals who were generally healthy and had some history of smoking. Cluster 2 was dropped from the analyses due to the preponderance of missing data. Cluster 3 was used as the control group, healthy non-smokers. Members of Cluster 4 had features of the metabolic syndrome and were generally not as obese as Cluster 5. Obesity was the hallmark of Cluster 5, the members of which also had some features of the metabolic syndrome. We then examined the genetic associations with both qualitative and quantitative representations of these empirically derived traits. Genetic analyses of the qualitative traits were conducted, comparing each of the affected groups with the unaffected cluster alone and, to increase statistical power, the unaffected group and healthy smokers combined. One single-nucleotide polymorphism on chromosome 4 met a conservative genome-wide significance level, but the effect was muted when we accounted for population stratification. The results for the quantitative traits were similar, with a small number of genome-wide significant findings muted by control for admixture. The directional findings will provide the basis for hypothesis generation for syndromes such as the metabolic syndrome and obesity.