Variance Component Score Statistics for QTL Mapping

Variance Components based models are commonly used for linkage and association mapping of quantitative traits. Score Tests based on these models are generally more robust to various modeling assumptions than the corresponding likelihood ratio tests. They are also computationally much simpler than the likelihood ratio tests, making them the natural choice for whole genome scans, which have become increasingly common with the emergence of high-throughput genotyping technologies. However the popularity of score statistics havebeen limited, due to several practical issues, such as lack of availability of software and guidelines for choice of score statistic variants. In this dissertation, we develop novel score statistics for both linkage and association mapping, elucidate the theoretical properties of these and of the existing variants, and also compare some of the existing and proposed score variants using simulation. Analytical arguments and simulation results are used to develop guidelines for choice of appropriate score variants under different practical situations. In this dissertation, we are primarily concerned with identifying robust and powerful score statistics for detecting genetic susceptibility loci for complex diseases by mapping underlying quantitative phenotypes. Unlike Mendelian disorders, complex diseases in humans typically have a large number of modest genetic effects, which cumulatively have a significant impact on the disease. The work in this dissertation is aimed at maximizing the power of genome scans to detect more of these small genetic effects. This is of considerable public health significance, as the identified genetic variants can be followed up to gain important insights into the etiology of the disease, which can further lead to development of screening tests and preventive and therapeutic interventions for complex diseases.