Ghosh Saurabh
Human Genetics Unit, Indian Statistical Institute, Kolkata, India.
Genet Epidemiol. 2014 Sep;38 Suppl 1:S63-7. doi: 10.1002/gepi.21827.
Our working group studied methods for joint analyses of multiple phenotypes using the data provided by Genetic Analysis Workshop 18. Two data sets were available: one containing genotypes obtained from a real human whole-genome sequencing study along with longitudinal measurements on systolic and diastolic blood pressure, age, sex, medication use, and tobacco smoking; and the other a simulated data set using the same set of genotypes and phenotype structure as the real data set. The nine sets of investigators in our working group focused on the statistical challenges posed by association analyses of multivariate phenotypes; they applied a wide spectrum of statistical methods, such as linear mixed models, copula models, and semiparametric regression models for simultaneous analyses of longitudinal data on the two blood pressure phenotypes at the genome-wide level. In this report, we discuss the various strategies explored by the different investigators whose common goal was improving the power to detect association with multivariate phenotypes.
我们的工作组利用遗传分析研讨会18提供的数据，研究了多种表型联合分析的方法。有两个数据集可供使用：一个包含从真实人类全基因组测序研究中获得的基因型，以及收缩压和舒张压、年龄、性别、药物使用和吸烟情况的纵向测量数据；另一个是模拟数据集，其使用与真实数据集相同的基因型和表型结构。我们工作组的九组研究人员关注多变量表型关联分析带来的统计挑战；他们应用了广泛的统计方法，如线性混合模型、copula模型和半参数回归模型，以在全基因组水平上同时分析两种血压表型的纵向数据。在本报告中，我们讨论了不同研究人员探索的各种策略，他们的共同目标是提高检测与多变量表型关联的能力。
