Rogus J J, Haines J L
Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts, USA.
Genet Epidemiol. 1995;12(6):665-9. doi: 10.1002/gepi.1370120624.
Using data simulated to reflect an oligogenic disease, we evaluated screening strategies based on lod-score and weighted pairwise correlation (WPC) analysis with respect to their ability to efficiently identify regions near disease loci. Lod-score analysis was done twice, once assuming a near-recessive mode of inheritance with a high penetrance and again assuming a semidominant mode of inheritance with lower penetrance. Under the near-recessive model, no disease loci were correctly identified, while there was one false positive result. Under the semidominant model, D1G31 was correctly identified, and there were two false positive results. Due to the lack of highly informative families and possible sensitivity to parameter misspecification, this poor performance was not unexpected. WPC, on the other hand, is assumption-free and thus potentially more powerful than a misspecified parametric model, but almost certainly less powerful than a well-specified parametric model. Using WPC modified to handle binary phenotypes with no age-of-onset, we found results closely resembling those under the semidominant model, although no markers near disease loci exceeded the theoretical critical value for WPC.
使用模拟数据来反映一种寡基因疾病，我们基于对数优势比分（lod-score）和加权成对相关性（WPC）分析评估了筛查策略，以考察它们有效识别疾病基因座附近区域的能力。对数优势比分分析进行了两次，一次假设为近乎隐性遗传模式且具有高外显率，另一次假设为半显性遗传模式且外显率较低。在近乎隐性模型下，没有正确识别出疾病基因座，同时有一个假阳性结果。在半显性模型下，D1G31被正确识别，并有两个假阳性结果。由于缺乏信息丰富的家系以及对参数错误设定可能敏感，这种不佳表现并不意外。另一方面，WPC无需假设，因此可能比错误设定的参数模型更强大，但几乎肯定比设定良好的参数模型更弱。使用经过修改以处理无发病年龄的二元表型的WPC，我们发现结果与半显性模型下的结果非常相似，尽管疾病基因座附近没有标记超过WPC的理论临界值。
