Defining the role of genetics in complex diseases demonstrating familial clustering, yet no simple Mendelian mode of transmission, has immense public health implications because these types of diseases are much more common than the rare Mendelian diseases. The traditional method for identifying disease susceptibility genes is based on selecting. unique pedigrees, often rare, which are most informative for genetic linkage analyses. Still, the analyses are often compromised by locus and allelic genetic heterogeneity, incomplete penetrance, sporadic cases of disease, and unknown or poorly estimated allele frequencies. An alternative method depends on the association of candidate genes with disease in the population. This approach, based on the genetic mechanisms of the disease process, allows estimation of not only relative risks for the candidate gene(s), but also the population attributable risk and gene-environmental interactions. The goal of this project is to make available statistical tools for design and analysis of population candidate-gene association studies. These methods will be useful for a wide variety of diseases. The choice of appropriate controls for association studies has been difficult, mainly because the frequencies of alleles for candidate gene loci may vary across ethnic subpopulations, and biased results can occur when cases and controls have different ethnic backgrounds. Our methods are based on family members as internal controls (with emphasis on parents). Unlike current work in this field, we have developed likelihood-based methods of analysis.
The specific aims of this proposal are: (I) to expand the theoretical development of the candidate gene likelihood-based relative risk models to (a) appropriately and efficiently account for """"""""messy"""""""" data (such as partially missing genotype data), inclusion of multiple cases per family, and inclusion of parental phenotype data; (b) determine the most efficient test for association; (c) extent the relative risk models so that the effects of measured covariates on the genotype risks can be assessed by regression modles - important models will include the effects of individual alleles, allelic interactions (within and between loci), genic interaction with person-specific covariates (i.e., gene-environment interaction), and genomic imprinting; (d) determine theoretical relative risk parameters as functions of genetic parameters to use for study design; (2) to validate by simulations the statistical methods developed in Aim l in terms of robustness to deviations from underlying assumptions and the adequacy of large sample approximations; and (3) to develop and distribute user-friendly computer code for planning and analysis of studies based on the methods developed by this grant.
