Abstract

PENA MALAVERA, A; GUTIERREZ, L  and  BALZARINI, M. Statistical models for phenotype-genotype association studies in genetically structured populations. BAG, J. basic appl. genet. [online]. 2016, vol.27, n.2, pp.49-58. ISSN 1852-6233.

Association mapping is used to find specific regions in the genome related to changes in a phenotypic trait. However, it has been found that in genetically structured populations, the number of false positives increases. The aim of this study was to compare the performance of several association mapping statistical models that take into account the underlying population genetic structure. Different statistical strategies developed under the mixed model theory were evaluated. The compared association models included the following matrices to model genetic structure: Q-matrix (probability of membership of each individual to each subpopulation), P-matrix (principal components of marker data capturing the structure variance) and K-matrix (containing genetic relationships between the individuals of the mapping population). The columns of Q-matrix and P-matrix were used in the associative mapping model as fixed effect covariates as well as random effect covariates. We also evaluated models including simultaneously Q-matrix and K-matrix, or either as P-matrix and K-matrix. The reference model (naïve model) was a regression model that did not account for genetic structure. Model comparison criteria were the empirical distributions of p-values, the FDR (False Discovery Rate) and the statistical power. The results suggest that the use of the K-matrix, alone or together with the Q-matrix reduced the false positive rate regardless of the level of genetic divergence among underlying subpopulations.

Keywords : Linear mixed models; Population genetic structure; False discovery rate.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )