Polygenic Score Prediction Within and Between Sibling Pairs for Intelligence, Cognitive Abilities, and Educational Traits From Childhood to Early Adulthood

The prediction of behavioral traits from polygenic score (PGS) in the population is strongest for cognitive and educational traits, which include intelligence, cognitive abilities, and educational attainment and achievement. For educational attainment, this greater predictive power of PGS appears to be due to substantial shared family-level (between-family) genetic effects, such as ancestral, mating, and family processes, that are not reflected within families at the level of the individual. Here, we extend our previous work by systematically investigating PGS prediction within and between sibling pairs for intelligence, cognitive abilities, and educational traits in childhood, adolescence, and early adulthood. We assess within-family PGS prediction from dizygotic twin differences and population-level prediction from unrelated individuals; the extent to which the within-family prediction is less than the population-level prediction indexes between-family effects. We used 6972 unrelated individuals and 3306 dizygotic twin pairs from the UK Twins Early Development Study. We compared the results for cognitive traits to height and body mass index (BMI). Finally, we assessed the extent to which socioeconomic status accounts for between-family effects. Results showed that within- and between-family effects contributed about equally to PGS prediction across cognitive and educational traits. Similar results emerged across development, despite population PGS prediction doubling. In contrast, for height and BMI, the within-family component accounted for most of the population prediction. We also confirmed that socioeconomic status largely accounted for the between-family effects. We discuss the implications of these findings for interpreting PGS prediction of cognitive and educational traits in the population and within and between families. Although between-family effects have been considered as a source of bias, we argue that between-family as well as within-family effects are legitimate sources of genetic variance for predicting individual differences in the population.