Genomic prediction of cognitive traits in childhood and adolescence

Abstract

Recent advances in genomics are producing powerful DNA predictors of complex traits, especially cognitive abilities. Here, we leveraged summary statistics from the most recent genome-wide association studies of intelligence and educational attainment to build prediction models of general cognitive ability and educational achievement. To this end, we compared the performances of multi-trait genomic and polygenic scoring methods. In a representative UK sample of 7,026 children at age 12 and 16, we show that we can now predict up to 11 percent of the variance in intelligence and 16 percent in educational achievement. We also show that predictive power increases from age 12 to age 16 and that genomic predictions do not differ for girls and boys. Multivariate genomic methods were effective in boosting predictive power and, even though prediction accuracy varied across polygenic scores approaches, results were similar using different multivariate and polygenic score methods. Polygenic scores for educational attainment and intelligence are the most powerful predictors in the behavioural sciences and exceed predictions that can be made from parental phenotypes such as educational attainment and occupational status.

Genetic effects on educational attainment in Hungary

“…While we are unaware of any study which quantifies between-country variations in the genetic signal of social class or other vectors of shared environmental effects on educational achievement, due to the highly divergent histories of these countries we find it unlikely that social stratification-driven effects on educational attainment are associated with similar genetic variants in Hungarian, Estonian and English samples. The degree of reduction in the PGS heritability in former Warsaw Pact countries relative to Anglo-Saxon samples (9-11% in the current sample, over 10% in a larger British sample (Allegrini et al, 2019;Selzam et al, 2019) versus 6-6.5% in young Hungarians and Estonians) is consistent with the roughly 50% reduction seen in recent within-family studies Selzam et al, 2019) which by design exclude social stratification, and the remaining proportion may index true genetic effects. We note, however, that genetic nurture (a true genetic effect on the parents causing a behavioral phenotype which increases offspring educational attainment) may still contribute to PGS heritability in the latter countries .…”

“…Large-scale genome-wide association (GWA) studies linked specific genetic variants to educational attainment (Rietveld et al, 2013;Okbay et al, 2016;Lee et al, 2018). Polygenic scores based on GWAS results confirmed the predictive value of these genetic variants (also termed polygenic score [PGS] heritability), typically accounting for up to 10% of the phenotypic variance in educational attainment itself Allegrini et al, 2018), cognitive abilities (de Zeeuw et al, 2014;Selzam et al, 2016;Allegrini et al, 2018), social mobility (Ayorech et al, 2017) and overall socioeconomic success (Belsky et al, 2016;Belsky et al, 2018) in independent samples. However, the association between genetic variants identified by GWAS and phenotypes may be mediated or moderated by environmental variables in at least two ways.…”

“…Another possible test of social stratification-free PGS heritability estimation is the use of populations in which the genetic signal of social class is presumably different than in the GWAS discovery population. The validity of education attaiment PGSs has been demonstrated among others in samples of Icelanders , Estonians and African Americans Lee et al, 2018;Rabinowitz et al, 2019), usually with significant but lower validity than in cohorts similar to GWAS discovery samples (Allegrini et al, 2018).…”

Screening human embryos for polygenic traits has limited utility

“…For example, initial evidence suggests that currently available datasets might be able to explain up to 50% of the variance in height by using LASSO, and that a similar doubling of explained variance is also possible for cognitive phenotypes [27]. Additionally, the use of multiple related phenotypes has been demonstrated to enhance the predictive power of PS [47]; for example, the combination of educational attainment and intelligence GWAS may permit a doubling of cognitive ps 2 [48]. Finally, it has recently been suggested that enrichment of certain subcategories of functional variation (e.g., coding, conserved, regulatory, and LD-related genomic annotations) in GWAS results can be leveraged to further enhance prediction accuracy [49,50].…”

Making the most of Clumping and Thresholding for polygenic scores

“…We prefer to use AUC over Nagelkerke's R 2 because AUC has a desirable property of being independent of the proportion of cases in the validation sample; one definition of AUC is the probability that the score of a randomly selected case is larger than the score of a randomly selected control (Wray et al 2013). An alternative to AUC would be to use a better R 2 on the liability scale (Lee et al 2012;Allegrini et al 2019).…”