The majority of Genome-Wide Association Studies (GWAS) focus on disease risk. Relatively less attention is paid to phenotypes such as disease-progression or severity. These phenotypes are longitudinal, requiring multiple data points over time, and cannot be analyzed through traditional association studies. We have developed a novel method to address this issue.
Our method analyzes a set of potentially heterogeneous longitudinal data and attempts to place the data into homogeneous subgroups. Also, our method allows for prediction of both the number of subgroups and their trajectories. The Bayesian posterior probabilities of belonging to a given subgroup are used as quantitative traits in traditional association methods. The method is applicable to family-based and population-based data sets and can incorporate environmental covariates.
Using simulations, we have documented that our method maintains proper type I error. Also, simulation results suggest that the method exhibits high power to localize disease loci (greater than 80% power in a variety of alternative hypotheses, including: single causal loci full penetrance, single causal loci reduced penetrance, and multiple causal loci). When using population-based data, the method has proven robust to the problem of population stratification.
Link to Paper
A novel method for analyzing genetic association with longitudinal phenotypes
Londono D, Chen KM, Musolf A, Wang R, Shen T, Brandon J, Herring JA, Wise CA, Zou H, Jin M, Yu L, Finch SJ, Matise TC, Gordon D: A novel method for analyzing genetic association with longitudinal phenotypes. Stat Appl Genet Mol Biol 2013;12:241-261. PMID: 23502345