A new study by Demange et al., (2020) brings more information about the genetic basis of educational attainment.
The authors aimed to find which non-cognitive, genetic factors were contributing to individual differences in educational attainment. The result was, of course, psychiatric disorders, behavior, and personality. The method was fairly straight-forward. The team conducted a GWAS on a very large sample and subtracted cognitive factors,
“The objective of our GWAS-by-subtraction analysis was to estimate, for each SNP, the association with educational attainment that was independent of that SNP’s association with cognition (hereafter, the NonCog SNP effect). We used Genomic-SEM22 to analyze GWAS summary statistics for the educational attainment and cognitive performance phenotypes in the SSGAC’s 2018 GWAS (Lee et al. 201823). The model regressed the educational-attainment and cognitive-performance summary statistics on two latent variables, Cog and NonCog (Figure 1). Cog and NonCog were then regressed on each SNP in the genome. This analysis allowed for two paths of association with educational attainment for each SNP. One path was fully mediated by Cog. The other path was independent of Cog and measured the non-cognitive SNP effect, NonCog. To identify independent lead hits with p <5e-8 (the customary p-value threshold to approximate an alpha value of 0.05 in GWAS), we pruned the results using a radius of 250 kb and an LD threshold of r2 <0.1 (Supplementary Tables 1 and 2).”
Their method allowed the genetic variance in cognitive factors to be independent of the genetic variance in non-cognitive factors. They also conducted a number of analyses on the relationships between said cognitive and non-cognitive factors and gray matter, white matter, brain volume, specific tissues, etc.
“The NonCog latent factor accounted for 57% of genetic variance in EA [Educational Attainment]. LD Score regression analysis estimated the NonCog SNP-heritability as h2NonCog=.0637 (SE=.0021).”
The non-cognitive factors they found were associated with decision-making preferences, non-risky behavior, delayed fertility, the Big Five personality traits, professional competency, and higher risk for psychiatric disorders.
The cognitive and non-cognitive factors actually differed in their relationship to brain volume as well as gray matter, however they were enriched in the same tissues, which may still imply some level of pleiotropy between personality/mental health/behavior and IQ, but more research should be done:
“As measured by correlation between Z-statistics, enrichment for Cog was similar to NonCog (r=.85) and there were no differences in cell-type-specific enrichment, suggesting little differentiation between cognitive ability and non-cognitive traits at the level of cell-type (Supplementary Figure 5). Stratified LDSC results were similar to results from MAGMA (Supplementary Note 2, Supplementary Figure 6 and Table 13). While the same gene-sets, based on scRNA-seq expression in neuronal cell-types, are enriched for NonCog and Cog, gene-level analysis69 (Supplementary Note 3) confirms the specific genes driving this enrichment do not necessarily affect the two traits in the same direction.”
Concerning gray matter, white matter, and brain volume, they say,
“For total brain volume, genetic correlation was stronger for Cog as compared to NonCog (Cog rg=.22 (SE=.04), NonCog rg=.07 (SE=.03), pdiff=.005). Total gray matter volume, controlling for total brain volume, was not associated with either NonCog or Cog (NonCog: rg=.07 (SE=.04); Cog: rg=.06 (SE=.04)). For total white matter volume, conditional on total brain volume, genetic correlation was negative and stronger for NonCog as compared to Cog (NonCog rg= −.12 (SE=.04), Cog (rg=−.01 (SE=.04), pdiff=.04).
NonCog was not associated with any of the regional gray-matter volumes after FDR correction. In contrast, Cog was significantly associated with regional gray-matter volumes for the bilateral fusiform, insula and posterior cingulate (rg range .11-.17), as well as left superior temporal (rg=.11 (SE=.04)), left pericalcarine (rg=−.16 (SE=.05)) and right superior parietal volumes (rg=−.22 (SE=.06)) (Figure 5).”
They report the polygenic prediction of IQ, achievement, and SES below:
Following this chart, they say,
“NonCog genetics have weaker associations with cognitive functions as compared to Cog genetics. NonCog and Cog were both genetically correlated with childhood IQ30; however, the magnitude of NonCog rg was less than half the rg for Cog (NonCog rg=0.31 (SE=.06), Cog rg=0.75 (SE=.08), pdiff_fdr<.0001). Of the total genetic correlation between childhood IQ and EA, 31% of the variance was explained by NonCog and 69% by Cog. In PGS analysis in the NTR and Texas Twin cohorts (N=2,815), effect-sizes for associations with IQ were smaller for NonCog as compared to Cog (NonCog β=.13 (SE=.01), Cog β=.25 (SE=.03); pdiff<.0001; Dunedin and E-Risk analysis pending). Sensitivity analyses of tests measuring different dimensions of cognitive function are reported in Supplementary Figure 2. These results confirm that NonCog genetic associations with cognitive test performance, while greater than zero, are of smaller magnitude as compared to EA or Cog genetics.”
They also found a significant genetic correlation between socioeconomic success and cognitive factors, non-cognitive factors, and educational attainment. This was consistent across cohorts:
Overall, a really good study. From a sociological viewpoint, we can probably use these to better tailor school resources, particularly in programs dedicated to working with different behavioral strategies. They generally confirmed what anyone on the hereditarian side of behavior beliefs were saying about outcomes.