Gjessing H K, Lie R T
Division of Epidemiology, Norwegian Institute of Public Health, Norway.
Ann Hum Genet. 2006 May;70(Pt 3):382-96. doi: 10.1111/j.1529-8817.2005.00218.x.
Case-parent triad data are considered a robust basis for studying association between variants of a gene and a disease. Methods evaluating statistical significance of association, like the TDT-test and its extensions, are frequently used. When there are prior hypotheses of a causal effect of the gene under study, however, methods measuring penetrance of alleles or haplotypes as relative risks will be more informative. Log-linear models have been proposed as a flexible tool for such relative risk estimation. We demonstrate an extension of the log-linear model to a natural framework for also estimating effects of multiple alleles or haplotypes, incorporating both single- and double-dose effects. The model also incorporates effects of single- and double-dose maternal haplotypes on a fetus during pregnancy. Unknown phase of haplotypes as well as missing parents are accounted for by the EM algorithm. A number of numerical improvements to maximum likelihood estimation are also implemented to facilitate a larger number of haplotypes. Software for these analyses, HAPLIN, is publicly available through our web site. As an illustration we have re-analyzed data on the MSX1 homeobox-gene on chromosome 4 to show how haplotypes may influence the risk of oral clefts.
病例-双亲三联体数据被认为是研究基因变异与疾病之间关联的坚实基础。经常使用评估关联统计显著性的方法，如传递不平衡检验（TDT检验）及其扩展方法。然而，当存在关于所研究基因因果效应的先验假设时，将等位基因或单倍型的外显率作为相对风险进行测量的方法会提供更多信息。对数线性模型已被提议作为进行此类相对风险估计的灵活工具。我们展示了对数线性模型的一种扩展，将其应用于一个自然框架，用于估计多个等位基因或单倍型的效应，同时纳入单剂量和双剂量效应。该模型还纳入了孕期单剂量和双剂量母体单倍型对胎儿的影响。通过期望最大化（EM）算法来处理单倍型的未知相位以及缺失的双亲。还对最大似然估计进行了一些数值改进，以便于处理更多数量的单倍型。用于这些分析的软件HAPLIN可通过我们的网站公开获取。作为示例，我们重新分析了4号染色体上MSX1同源框基因的数据，以展示单倍型如何影响口腔腭裂的风险。
