Poster (download) Ivan A. Kuznetsov1, Sergei A. Slavskii2, Tatiana I. Shashkova3, Georgii A. Bazykin4, Tatiana I. Axenovich5, Fyodor A. Kondrashov6, Yurii S. Aulchenko71Center of Life Sciences, Skolkovo Institute of Science and Technology, Moscow, Russia; Laboratory of Theoretical and Applied Functional Genomics, Novosibirsk State University, Novosibirsk, Russia, Ivan.Kuznetsov@skoltech.ru2Laboratory of Theoretical and Applied Functional Genomics, Novosibirsk State University, Novosibirsk, Russia, slavsky@phystech.edu3Laboratory of Theoretical and Applied Functional Genomics, Novosibirsk State University, Novosibirsk, Russia, deppt002@gmail.com4Center of Life Sciences, Skolkovo Institute of Science and Technology, Moscow, Russia, g.bazykin@skoltech.ru5Laboratory of recombination and segregation analysis, Institute of Cytology and Genetics SB RAS, Novosibirsk, Russia, tatiana.aksenovich@gmail.com6Institute of Science and Technology, Vienna, Austria, fkondrashov@gmail.com7Laboratory of recombination and segregation analysis, Institute of Cytology and Genetics SB RAS, Novosibirsk, Russia, yurii@bionet.nsc.ru The classical approach to analysis of polygenic quantitative traits assumes use of normal approximation and the additivity of effects. For more than a century, adult height served as an exemplary trait justifying this type of analysis and serving as a test case. We demonstrate that the classical approach to analysis of height has met its limits in contemporary large populations. In particular, we demonstrate the existence of weak, but highly significant non-additive interactions of genetic and environmental factors. In the conventional model, the better fit to the data that was achieved by accounting for these interactions came at the expense of the mean-model’s increased complexity. The complexity of the model could be reduced if log-normal approximation was used.

Read More