Using DNA to predict Education: A Meta-Analytic Review

Abstract

Recent advances in genomics have made it possible to predict people’s differences in educational success from person-specific aggregates of DNA, so-called genome-wide polygenic scores. Here, we review the validity and strength of these DNA-based predictions for educational success across studies. We identified 18 publications that reported associations between polygenic scores, based on the summary statistics of Lee et al.’s (2018) Genome-Wide Association Study for years spent in education, and phenotypic measures of educational success, including academic tests scores, school grades, highest educational qualifications, and years spent in education. A multilevel meta-analysis with k=39 estimates from 16 independent samples (Ntotal = 398,545) yielded an estimate of ρ=.27 (95% CI from .23 to .31) between polygenic scores and educational success. A meta-regression model controlling for sample characteristics, methodological factors, and phenotype measure differences yielded an estimate of ρ=.40 (95% CI from .25 to .54), indicative of a large effect size. This association varied as a function of ancestry, with the prediction being almost thrice as strong for European ancestry than African American ancestry samples. Our findings confirm that DNA-based predictions of educational success are substantial, with a Standard Deviation (SD) increase in polygenic score being associated with a .32 score increase in high school GPA or spending 1.6 more years in full-time education. We discuss how understanding DNA-based predictions may help address students’ heterogeneity in learning abilities and improving teaching and education provisions.