Genetic association analysis in family samples with missing data

Collections of families are useful for genetic studies because their members share a common background, both genetic and environmental, meaning that differences between them are likely to be due to the genes being studied. Family studies can also determine whether a mother?s genes affect the health of her child through inheritance by the child or presence in the mother. For practical reasons it is often difficult to collect data from all the members of a family, so statistical methods can be used to fill in the missing data. In recent years development of such methods has fallen behind similar methods designed for unrelated subjects. This project will extend our previous work on missing data in family studies, developing several new methods that will be applied to a number of collaborative studies.

Family-based association analysis uses samples of cases and their relatives to draw inferences on disease association: most commonly, cases and their parents, or sib pairs (including twins) are used. Their advantages include matching of subjects for genetic and environmental background, and the ability to distinguish the effects of maternally inherited genes from those paternally inherited. Major collections of family data are in progress and require rigorous statistical methods for their epidemiology. This project will extend our previous work on likelihood based analysis of samples with missing data. We will develop and implement methods for distinguishing the effects of genes transmitted by the mother from those carried by her. We will develop proper analysis of quantitative traits when subjects have been sampled on the basis of their trait values. We will devise appropriate methods for analysis of copy number variants in family data. In each of these applications we will allow for missing genotype data, in particular missing parents in nuclear families, in a statistically optimal fashion. These methods will be applied in a number of collaborative studies. We will also translate our methods into a general epidemiological framework to allow more widespead application outside of genetics.