A bivariate genetic analysis of HDL- and LDL-cholesterol incorporating measured covariates: a Gibbs sampling application.

We analyzed HDL- and LDL-cholesterol levels as a bivariate phenotype in 27 families as a function of major genes, polygenes, and measured covariates using a Monte Carlo sampling technique called Gibbs sampling. Major genes and polygenes exhibited strong effects, when considered separately. While a major gene versus polygene model could not be clearly differentiated for HDL-C, polygenes appeared to play a stronger role than a major gene for LDL-C. There was no evidence of linkage between the two major genes for HDL- and LDL-C, and the correlation in polygenes was negative. The analysis illustrate the potential applicability of Gibbs sampling to such complex problems as the multivariate analysis of continuous phenotypes.