Lo Shaw-Hwa, Zheng Tian
Department of Statistics, Columbia University, New York, NY 10027, USA.
Hum Hered. 2002;53(4):197-215. doi: 10.1159/000066194.
The mapping of complex traits is one of the most important and central areas of human genetics today. Recent attention has been focused on genome scans using a large number of marker loci. Because complex traits are typically caused by multiple genes, the common approaches of mapping them by testing markers one after another fail to capture the substantial information of interactions among disease loci. Here we propose a backward haplotype transmission association (BHTA) algorithm to address this problem. The algorithm can administer a screening on any disease model when case-parent trio data are available. It identifies the important subset of an original larger marker set by eliminating the markers of least significance, one at a time, after a complete evaluation of its importance. In contrast with the existing methods, three major advantages emerge from this approach. First, it can be applied flexibly to arbitrary markers, regardless of their locations. Second, it takes into account haplotype information; it is more powerful in detecting the multifactorial traits in the presence of haplotypic association. Finally, the proposed method can potentially prove to be more efficient in future genomewide scans, in terms of greater accuracy of gene detection and substantially reduced number of tests required in scans. We illustrate the performance of the algorithm with several examples, including one real data set with 31 markers for a study on the Gilles de la Tourette syndrome. Detailed theoretical justifications are also included, which explains why the algorithm is likely to select the 'correct' markers.
复杂性状的定位是当今人类遗传学最重要且核心的领域之一。最近，人们的注意力集中在使用大量标记位点的基因组扫描上。由于复杂性状通常由多个基因引起，逐个测试标记来定位它们的常见方法无法捕捉到疾病位点间相互作用的大量信息。在此，我们提出一种反向单倍型传递关联（BHTA）算法来解决这个问题。当有病例 - 亲代三联体数据时，该算法可用于对任何疾病模型进行筛选。它通过在对重要性进行全面评估后，一次消除一个重要性最低的标记，从而识别出原始较大标记集的重要子集。与现有方法相比，这种方法有三个主要优点。首先，它可以灵活地应用于任意标记，无论其位置如何。其次，它考虑了单倍型信息；在存在单倍型关联的情况下，它在检测多因素性状方面更强大。最后，就基因检测的更高准确性和扫描所需测试数量的大幅减少而言，所提出的方法在未来全基因组扫描中可能会被证明更有效。我们用几个例子说明了该算法的性能，包括一个针对抽动秽语综合征研究的有31个标记的真实数据集。还包括详细的理论依据，解释了为什么该算法可能会选择“正确”的标记。
