Is the Model of Causality a Good Idea?

In this episode of Unsupervised Learning Razib talks to James Lee, a professor of psychology at the University of Minnesota. Lee is a co-author of a new paper in Nature, Polygenic prediction of educational attainment within and between families from genome-wide association analyses in 3 million individuals. A landmark in the field of cognitive genomics, this publication is the result of years of collaboration between two dozen researchers. 

Over the course of the episode, they deep dive into the results from the publication that Lee in particular finds fascinating. But first, Razib brings up a recent controversy related to Paige Harden’s book The Genetic Lottery and the science that undergirds its thesis. Evolutionary geneticists Graham Coop and Molly Prezworski recently wrote a review of Harden’s book in Evolution, Lottery, luck, or legacy. A review of “The Genetic Lottery: Why DNA matters for social equality”. They argue that Harden overplays her hand in terms of what “polygenic risk scores” can tell us about our future life trajectory (and in particular her focus on education outcomes), as well as their social utility. Harden responds with a piece titled Forests and Trees, contending that Coop and Prezworski mischaracterize her position and seem to hold behavior genetics to an unreasonably high standard of evidentiary validity. In buttressing the science in The Genetic Lottery, Lee expounds on the importance of the finding that genetic positions associated with something like higher educational attainment seem highly correlated with regions of the genome associated with neurological development in particular.

Next, Razib asks what aspect of the new paper Lee found most interesting, and he points to the section on the nature of dominance, the characteristic whereby certain genetic variants express a trait when present in a single copy, as opposed to two copies (recessive traits). These arguments go back to Sewall Wright and R. A. Fisher’s debates about the nature of dominance from a century ago, a divergence in viewpoints at the very founding of population genetics as a field. Lee favor’s Wright’s view that dominance is a function of the physiological mechanism of gene expression; a gene that produces proteins will still produce sufficient quantities in even a single copy. In contrast, most of the authors of  Polygenic prediction of educational attainment within and between families from genome-wide association analyses in 3 million individuals favored Fisher’s idea that dominantly expressed genes sweep to selection faster, and so that view is tacitly supported in their conclusions.

During the rest of the discussion, Lee expounds on a wide range of topics that touch on behavior genomics, from whether rare variants of large effect will come to be seen as important, to why heritability estimates using family-based designs are so much lower for educational attainment than conventional population-wide statistics, and the relevance of the results from this latest work for evolutionary genetics. Lee makes the case that the synthesis of genomics and behavior genetics makes for a fascinating story of scientific discovery that will help illuminate our understanding of human nature in the 21st-century, far beyond the field’s utility in predicting individual traits.