Detecting genetic structure from molecular marker data in the context of association mapping

PEÑA MALAVERA, A.; BRUNO, C.; TEICH, I.; BALZARINI, M.

La Falda, Córdoba.

Congreso; Cuarta Escuela de Matemática y Biología; 2010

Association mapping is used extensively in human, animal, and plant genetics. It is an effective strategy to identify loci that govern complex traits in a context of abundant availability of genomic information, such as that produced by different types of DNA markers like multilocus-multiallelics microsatellites (SSR). Association mapping is based on the assumption that if a mutation increases the observation of a characteristic in a group of individuals, then it is expected that the allele associated with such feature will be more frequent among those individuals who share it, being identified by using allele frequencies and polymorphism rates at the population level. Before initiating association mapping studies it is necessary to empirically describe the underlying genetic structure and to estimate its magnitude. Genetic structure exists whenever subgroups of individuals differ systematically in their allele frequencies for different loci but in many cases it is not known a priori. When carrying out an association analysis without considering the effects of genetic structure, the risk of detecting spurious associations between markers, and between them and the phenotype of interest, increases. In structured populations there might be a high proportion of significant associations even when many markers are not linked to any locus of the character of interest. In order to infer genetic population structure and to obtain a rational basis for marker-assisted selection, several statistical and bioinformatics methods for classification in multidimensional spaces including hierarchical (UPGMA), non-hierarchical (K-means) and recent methods (SOM) can be applied to multilocus-multiallele data. In this study we will compare the power to detect genetic structure of these different multivariate analytical approaches, give an illustration of its applications, and discuss how they can impact on association mapping. The analyses will be evaluated in simulated and real data sets with a priori known genetic structure.