Inference of genealogical relationships among individuals from genetic markers

Individuals in a population may have among themselves various genealogical relationships (such as sib and parent-offspring relationships). Knowledge of these relationships is essential in many areas of research in behavioural, ecological and evolutionary genetics and in conservation biology. Although pedigree or ecological data can be used to determine relationships, such data are rarely available from most natural populations. In such cases, we can genotype individuals at a number of marker loci and infer their genealogical relationships from the pattern of similarity (allele sharing) among the multi-locus genotypes of the individuals. Powerful likelihood methods have been developed to partition a sample of individuals into distinctive genetic groups (defined by structured relationships) of variable sizes by maximising the likelihood of marker data. However, these methods are limited in application to either small problems in which the number of sampled individuals is small, or simple problems in which only sibships are inferred in a one-generation sample of individuals. Built on my previous work, this project aims to develop statistical methods for inferring parentage and sibships jointly in a large two-generation sample of individuals from marker data, and for assessing the uncertainties of the inferences. Typing errors and mutations in marker data will be accounted for in inferring relationships and be identified simultaneously by the methods. These extensions will make the group likelihood methods more flexible, robust and powerful in inferring relationships among individuals from marker data in practice. We will first develop the methodology, and then use extensive simulations to investigate the statistical properties of the method and its robustness when some assumptions are violated. Some empirical data sets with known relationships will be analysed by the proposed methods to further check their performance in realistic situations and to demonstrate their usefulness. The final goal is to develop a software package implementing the proposed group likelihood methods and to make it available free on the World Wide Web to the scientific community.

Using genetic markers to infer the genealogical relationships (e.g. sibship) among individuals in a population is becoming an important tool in ecology and evolutionary and conservation biology. Powerful group likelihood methods have been developed to partition a sample of individuals into distinctive genetic groups (defined by one or more types of relationships organised in a specific structure) of variable sizes by maximising the likelihood of marker data. However, these methods are limited in application to either small problems in which the number of sampled individuals is small, or simple problems in which only sibships are inferred in a one-generation sample of individuals. Built on my previous work, this project aims to develop statistical methods for inferring parentage and sibships jointly in a large two-generation sample of individuals from marker data, and for assessing the uncertainties of the inferences. Typing errors and mutations in marker data will be accounted for in inferring relationships and be identified simultaneously by the methods. These extensions will make the group likelihood methods more flexible, robust and powerful in inferring relationships among individuals from marker data in practice. We will first develop the methodology, and then use extensive simulations to investigate the statistical properties of the method and its robustness when some assumptions are violated. Some empirical data sets with known relationships will be analysed by the proposed methods to further check their performance in realistic situations and to demonstrate their usefulness. The final goal is to develop a software package implementing the proposed group likelihood methods and to make it available free on the World Wide Web to the scientific community.