Contributions to family-based association tests in candidate genes

We study tests for multiple markers in candidate genes using family designs. We conduct extensive simulation studies and confirm our results via applications to real complex disease data.; We propose a new multivariate test FBAT-MM and assess its performance as a part of comparison of competing testing strategies, characterized in this study as combinations of the following factors: genes, statistical tests, tag SNP methods, number of tag SNPs and family designs. We find that tag SNP methods, gene characteristics, and family designs have minimal impact on the best testing strategy. The familywise error rate (FWER) and FBAT-MM test offer the highest power. The best testing strategy that emerged from our analysis is to use the FWER or the multimarker test and select 6-10 tag SNPs using any of the tag SNP methods.; We propose a new permutation test FBAT-MinP. The test combines information from individual marker scores and their pairwise correlations to obtain a large-sample approximation of the sampling distribution of the most extreme test. We find that FBAT-MinP offers uniformly the highest power and is the only test not affected by the multiple testing problem as it gains power when applied to all SNPs relative to using tag SNPs. When tag SNPs are used, FBAT-MinP outperforms the next best test by approximately 4% and the gain increases to 8% when all SNPs are available. Lastly, we propose a new test for continuous traits. The test selects and combines optimal linear combinations of the offspring genotypes. We compare the performance of the new test with that of the test that is currently known to offer the highest overall power FBAT-LC [Xin et al. 2006]. For DSL frequencies larger than 0.25, we find that the new test has performance very similar to that of FBAT-LC, 39 and 41% respectively. Further, for four of the ten analyzed genes, the new test offers higher power.; In agreement with our simulation results, the power performances of FBAT-MM and FBAT-MinP are confirmed via Alzheimer's disease data analysis. Similarly, an analysis of BMI data show that the new test for continuous traits offers similar performance to that of FBAT-LC.