Genetic insights into learning: Using polygenic scores to predict school performance independent of cognitive ability

Complex traits, such as the ones seen in the behavioural sciences, are highly polygenic, meaning they are influenced by many genes of small effect (Abdellaoui et al., 2023). Genome-wide association studies (GWAS) aim to quantify the effects of individual variants to understand the genetic architecture of such traits. An important application of GWAS is constructing polygenic risk scores (PGS) - composite indices of genetic risk using inherited DNA variants (Plomin & von Stumm, 2018).

The genome does not systematically change throughout one's life course, meaning a PGS is able to predict equally well at birth and at adolescence. The most predictive PGS in the behavioural sciences is educational attainment (years of education), explaining 16% of the variance in age 16 GCSE achievement (Allegrini et al., 2019). Beyond prediction, PGS provide some of the strongest evidence for the importance of the environment, by controlling for individual differences in underlying genetic risk (Plomin & Von Stumm, 2022).

Although the predictive power of PGS is a major advance, a significant challenge in genome-wide association is that it captures all variants associated with educational attainment, regardless of whether they are causal for the trait (Abdellaoui et al., 2023). Assortative mating and population stratification are two major sources of confounding, which can inflate estimates of heritability and lead to spurious associations (Selzam et al., 2019).

Further, it is difficult to extricate direct genetic effects and indirect genetic effects, which represent the effects of the non-transmitted parental alleles mediated by the environment (Eilertsen et al., 2021). However, another issue in genome-wide association is that it detects variants for traits correlated with educational attainment, such as mental health risk, cognitive ability, and social-economic status (Abdellaoui et al., 2023), which is a benefit in terms of prediction but a disadvantage in terms of explanation.

Previously, the first supervisor has examined decomposing genetic effects into within and between variance components, finding that SES was a major driver of between-family differences in cognitive traits through gene-environment correlation (Selzam et al., 2019). A limitation of current cognitive PGS is that they are derived from genome-wide association studies of extremely general traits - general school performance and general cognitive ability.

The first supervisor has previously explored the genetics of specific cognitive abilities, finding that when residualized on g (SCA.g) they remained substantially heritable (Procopio et al., 2022). As an extension of this research, the first supervisor has also applied multi-polygenic score prediction (previously developed by Krapohl et al., 2018) to predict SCA.g, explaining 4.4% of the variance.

To examine the importance of the environment using genetically sensitive designs, the first supervisor has previously found that average differences in school achievement between schools mirror the genetic differences between them (Smith-Woolley et al., 2018). Controlling for prior ability, which itself is mediator for genetic effects, OFSTED school ratings similarly explain little to no variance in individual pupils' achievement (Von Stumm et al., 2021).

While the environment is important (explaining the other half of variance in achievement), the primary factor driving age-to-age stability in achievement is genetics (Rimfeld et al., 2018). (491 words) Aim of the investigation (up to 150 words) State primary research question and where appropriate the primary hypotheses being tested.

We have three overarching research questions, described below: Maximizing polygenic score prediction Assess nonshared environment (NSE) factors from infancy to adulthood in achievement and ach.g. Chart genotype-environment (rGE) effects on enviro