Individual differences in educational achievement are highly heritable (60%) throughout the school years, and this high heritability is largely explained by genetically influenced cognitive and non-cognitive factors (Krapohl et al. 2014).
The best early predictor of academic achievement is family socioeconomic status (SES), with children from poorer SES families doing worse at school compared to children from higher SES families (Bradley & Corwyn, 2002). However, this prediction is the same for all children within a household, yet academic achievement often differs between family members. Intelligence (g) is by far the best individual-specific predictor of educational achievement, both contemporaneously and longitudinally. However, intelligence as measured in infancy or early childhood is not as reliable and explains less than 3% of the variance in exam grades at the end of compulsory education. Converesely, we have previously shown that we can predict up to 9% of the variance in exam performance at age 16 using a single polygenic score derived by aggregating the effects of educational attainment-associated DNA variants identified through a discovery genome-wide association (GWA) study (Okbay et al. 2016; Selzam et al. 2016). Importantly, this DNA predictor is individual specific (unlike family SES) and it is available at birth or even prenatally (unlike g).
Here, we use the UK representative Twins Early Development Study (TEDS) sample of over 4,000 unrelated individuals to explore the variance explained in exam scores (GCSE) at the end of compulsory education by g across development (from age 2 to 16). We compared these estimates to the variance explained by DNA (polygenic score) using the summary statistics from a 2018 educational attainment GWA with a sample of 1.1 million participants (Lee et al. 2018). We will also explore the variance explained by genome-wide polygenic score (GPS) and g across development when controlling for SES. Our preliminary findings show that DNA provides the best child-specific prediction of school achievement available at birth, explaining over 14% of the variance. Until age 7, the prediction from DNA is stronger than from IQ and from previous educational achievement. DNA has slightly more predictive power than family SES measured at birth; importantly, DNA explains additional variance in exam grades when family SES is controlled. This prediction from DNA will improve with more powerful GWA studies and using multi-polygenic score approaches. Eventually, it will be possible to identify children who are likely to have difficulties at school, which will foster hope for early intervention.
References
Bradley RH & Corwyn RF Annual Review of Psychology 2002
Krapohl E. et al. Proceedings of the National Academy of Sciences 2014
Lee J. et al. Nature Genetics (in press)
Okbay A. et al. Nature 2016
Selzam S. et al. Molecular Psychiatry 2016
