Linkage analysis has been the successful primary tool for mapping many Mendelian traits and some complex traits until the genetic analysis paradigm shifted from rare Mendelian disease mapping using family data to the common variant-common disease mapping mainly using unrelated case-control data, i.e., genome-wise association studies (GWAS). However, the emerging availability of sequencing data and the inability to detect rare risk variants by GWAS has led to a renewed interest in linkage and other family-based
methods.
Olson (1999)’s Conditional-Logistic (CL) model retains the nice property of the LOD score formulation, and has advantages over other methods to make it an appropriate choice for complex trait mapping. However, the asymptotic distribution of the CL-LR statistic with constraints on the model parameters is unknown. Also, the method assumes independence of pairs; therefore, significance levels are biased, resulting in increased type I error rates, with data consisting of a pedigree with multiple ARPs.
The goal of this dissertation is to address these issues, and several steps of work are done. First, a web-based tool that pipelines the informatics process for pedigree data is developed. Given pedigree data, it provides a convenient “one-stop-shop” for pedigree informatics: descriptive statistics, genetic similarity coefficients, the variance-covariance values for similarity coefficients, a plot of pedigree structure, and a visualization of identity coefficients. Next, three approximations to the asymptotic null distributions of the CL-LR statistics are developed and compared with the empirical null distributions by simulation using independent ASPs. Then, the impact of the pedigree structure on the null distribution of CL-LOD scores is investigated for different analysis models, suggesting a promising indicator to be used in a weighting scheme. Lastly, new score tests in the CL model framework are developed accounting for the non independence of multiple pairs within a pedigree. The performance of the score tests is evaluated and compared with those of other existing methods by simulation. The work will provide a valuable addition to improve the current genetic analysis of complex traits, thereby contributing to the identification of genes for disease traits.