We had previously developed methods for qualitative traits using genotypes for affected individuals and their parents, and if possible their grandparents. The log-linear approach we had proposed provides a powerful likelihood ratio test and estimation of relative penetrances for variant alleles, without requiring knowledge of the genetic model. It is also robust against bias due to population stratification. The method can incorporate parent-of-origin effects, and maternally-mediated genetic effects, and allows for the possibility that one or both parental genotypes, or even the genotype of the affected offspring, may be missing. ? ? In dissertation research with a student at Duke, we estimated the amount of power improvement that could be achieved by using unaffected offspring to reduce the genotype ambiguity for missing parents.? ? Statistical methods for assessing maternal genetic effects on a qualitative outcome have been written in a chapter, for inclusion in a book related to maternal genetic effects being edited by Laura Mitchell.? ? Case-parent-based studies cannot directly assess effects of exposures, which is a serious limitation; traditional case-control studies have a different set of limitations. For example, they are subject to bias due to selection of controls and they cannot directly assess maternally mediated effects or imprinting. We developed a hybrid design that includes both nuclear families and unrelated population control parentss, and thereby offers the best of both approaches. This hybrid design is considerably more powerful than either traditional approach and also provides a test for bias due to population stratification, a problem that can invalidate a population-based case-control study of genetic effects. If this problem is detected in a given hybrid study, the investigator retains the ability to fall back to the case-parents component of the study, which will remain valid for testing genetic effects and gene-by-environment interaction. In work with a visiting student (Vermeulen), we have now extended the hybrid design to allow for a scenario where the population controls are mother-child pairs, rather than mother-father pairs. This is being further evaluated through power calculations and simulations and will soon be written up for publication. ? ? We recently developed and published a data mining approach (TRIMM for triad-based multi-markers). This method makes use of multi-marker studies of nuclear families, with affected individuals and their parents. The approach provides a way to test a set of linked markers simultaneously for association with disease risk and also a way to nominate a set of risk-haplotype tagging SNPs specific to the phenotype under study. The method does not require resolution of phase ambiguity, but relies directly on genotype information,i.e. the number of copies of a variant SNP inherited at each locus. It can handle many loci simultaneously, and can handle sporadically missing genotype data. Also, a simple extension allows for detection of haplotypes that may have acted through the maternal genome/phenotype during gestation with the affected offspring. In work about to be submitted, we are now using likelihood methods to estimate both offspring and maternal relative risks associated with one and two copies of a particular candidate haplotype. The method yields unbiased estimation for adequate sample sizes, and when multiple tightly-linked SNPs are studied can allow the analyst to distinguish between a protective effect of one candidate haplotype and a deleterious effect of that haplotypes complement. Such Yin-Yang haplotype pairs have recently been reported to be surprisingly common in the human genome, so it is important to be able to resolve that inferential conundrum. ? ? We are now also extending the method to allow identification of super-multiplicative interactions between haplotypes and environmental exposures, again using nuclear family data from cases and their parents. Early results of our simulations are promising, though detection of gene-by-environment interaction remains a challenging problem. ? ? In another project, we are estimating the asymmetry that would exist in family history data secondary to the existence of a maternally-mediated genetic effect. We applied this strategy to family history data from the Sister Study, and found evidence that maternal grandmothers of young-onset (under 50) cases of breast cancer were more likely to have had breast cancer than were their paternal grandmothers. This suggests there may be maternally-mediated genetic risk factors for breast cancer, or that mitochondrial variants play a role.
Showing the most recent 10 out of 42 publications
