Statistical Genetics Research on the Utility of Polygenic Embryo Screening
In Lencz et al, eLife (2021) we again used statistical genetics modeling, this time to examine the utility of embryo screening for reducing the risk of a single polygenic disease (see our Q&A). Given our results for traits (Karavani et al, 2019), we were surprised to observe large estimated relative risk reductions, but only when modeling a “selection strategy” of implanting the lowest scoring embryo. Given that polygenic risk scores have low sensitivity to detect future cases, using the scores only to exclude high risk embryos is not effective in reducing the risk. For example, when selecting the embryo with the lowest polygenic risk score for schizophrenia (out of five), our model and simulations estimated more than 40% relative risk reduction. On the other hand, absolute risk reduction can be quite small for diseases, like schizophrenia, that affect only a small percentage of the population. We developed an online risk reduction calculator for the clinical and scientific community to explore the expected risk reduction under various settings.
In Karavani et al, Cell (2019) we used statistical genetics modeling to study the utility of embryo screening for polygenic traits (see our Q&A and coverage in Science). We found that using polygenic scores available at the time, utility is limited. For example, selecting the highest scoring embryo (out of five viable embryos) would increase the mean height of the child (compared to random selection) by ~2.5cm or the mean IQ by ~2.5 points. In practice, even this modest increase is not guaranteed, as the realized adult trait may vary widely around its mean. In fact, the trait may even end up lower than random selection, as we demonstrated using very large nuclear families genotyped and phenotyped for height.