Comparison of Statistical Models for the Linkage Analysis of A Quantitative Trait

초록

Objectives: This study was undertaken to suggest statistical models for linkage analysis of quantitative trait locus (QTL) with pedigree data. The proposed models involve mixed models based on the adjusted Haseman-Elston model and allow random effects. We compared the performances of the models using the simulated dataset. Methods: The proposed models were applied to the analysis of the Genetic Analysis Workshop 12 (GAW12) simulated dataset for a QT which contains 38,850 sib-pairs from 1,150 pedigrees. We considered a response variable as sibship sample mean corrected cross-product of quantitative trait (smHE) proposed by Wang et al (2001). The proposed models consist of a random effect for correlation among siblings from a same pedigree, and one for the correlation among siblings from same parents. In addition, we considered several types of adjusting trait factors included in the proposed models. We also compared the performance of the proposed models with that of weighted least squares (WLS) regression models. Results: A model after adjusting for each QT (or QT2) of siblings yielded good power for detecting linkage of major markers. The models with a response variable as smHE showed better performance than the adjusted Haseman-Elston models. The power in each mixed model was generally high and higher than that in independence model for detecting linkage. Weighted models showed similar power to or lower power than unweighted models. The plots of residuals were improved via the mixed models with smHE. Conclusions: We conclude that the mixed models with smHE adjusted for each QT (or QT2) of siblings seem to work better than the adjusted Haseman-Elston model to identify linkage on considering all candidate markers at the same time. The residual plots followed asymptotically normal distribution in these proposed models for these data. The proposed models seem not only quite useful in detecting linkage for the QT but also quite flexible to use.