Identification of loci involved in childhood visual acuity and associations with cognitive skills and educational attainment

The study investigates how genetic and behavioral factors underlying childhood visual acuity relate to cognitive skills, neurodevelopmental conditions, and later educational outcomes. Visual acuity, often reduced by refractive error, affects quality of life and is linked to reading and academic performance. Prior research shows mixed behavioral associations between visual acuity and reading ability, and large GWAS in adults identified many loci for refractive error and myopia, with age-dependent genetic effects and strong gene–environment interactions. Education has been shown to causally increase myopia risk in adults, yet visual and cognitive function appear positively linked in childhood. The authors hypothesize that EA-associated genetics positively relate to childhood visual acuity (via better cognition), whereas myopia biology should negatively relate to visual acuity. They test these hypotheses by examining phenotypic associations, conducting a GWAS of visual acuity in children, and using PGS and genetic correlation analyses in the ALSPAC cohort.

Previous studies linking visual acuity to neurodevelopmental traits such as ADHD and ASD have yielded varied results. Research on reading skill shows mixed evidence: some studies reported no differences in reading across visual acuity groups, whereas others found that bilateral deficits in near visual acuity related to underachievement in reading even after adjusting for SES and IQ, and dyslexia groups sometimes showed poorer near and distance acuity than controls. Early literacy has been associated with visual acuity, though associations with school grades may diminish after extensive adjustment. Genetic studies of refractive error and myopia (European and Asian cohorts) have identified hundreds of loci, with evidence that SNP effects differ by age and environment, including substantial amplification of myopia risk with higher education. Mendelian randomization suggests that higher educational attainment causally increases myopia risk rather than vice versa. These findings motivate examining genetic links between visual acuity and cognition in childhood, where the direction of association may differ from adulthood due to gene–environment correlation.

Cohort: Data were from the UK ALSPAC longitudinal cohort. Initial recruitment yielded 14,062 live births; after a later recruitment wave, 15,454 pregnancies were included. Ethical approvals were in place and informed consent obtained. Phenotypes: Visual acuity was assessed at mean age 11.81 years using logMAR charts under standardized conditions, with testing randomized between eyes and both without and with pinhole; best-corrected acuity for each eye was derived. The smaller logMAR value from the better eye was used, then transformed to Visual Acuity Rating (VAR) using visual acuity = 100 − 50(logMAR), where higher values indicate better acuity. Exclusions: children with sensory impairments (n=26), visual acuity <85 (n=1450), or interocular difference >10 (n=5350) were excluded, yielding n=6807 for phenotypic analyses (~13% with corrected vision). Cognitive measures included reading skill (WORD basic reading at age 7), short-term memory (Nonword Repetition at age 8), listening comprehension (WOLD at age 8), WISC-III verbal IQ, performance IQ, and total IQ at age 8, communication skills (CCC at age 9), and capped GCSE scores as EA. SES was maternal highest education during pregnancy, grouped into five levels. Hearing thresholds were available for a subset (n=4929) for correlation with visual acuity. Neurodevelopmental subgroup assignment (reading difficulty, language impairment, ADHD, ASD, comorbidities, and unaffected) followed predefined criteria based on performance and reports; for analyses focusing on reading difficulty, children with comorbid reading difficulty were assigned to that group. A sex-matched control group (n=2238) was created. Genotyping and QC: Genome-wide genotypes were generated on Illumina HumanHap550-quad arrays. Standard QC removed non-European ancestry individuals. After QC, 9115 subjects and 500,527 SNPs remained. Phasing with ShapeIT v2; imputation with IMPUTE v3 using the HRC 1.1 reference panel. SNPs with Info<0.8 or MAF<0.05 were excluded. A total of 5,305,352 autosomal SNPs (genotyped or imputed; MAF>0.05) were tested. Genomic analyses: Visual acuity was inverse rank-transformed for genomic analyses. GWAS used BOLT-LMM v2.3.4 with sex, age, and first two PCs as covariates in n=5571. SNP heritability (SNP-h²) and within-sample genetic correlations were estimated in unrelated individuals (IBD<0.05) with GCTA-GREML, adjusting for sex, age, and two PCs; PCs were derived from LD-pruned genotyped SNPs. Replication targeted previously reported genome-wide significant SNPs for refractive error and myopia: 1,033 lead SNPs were reduced to 661 independent SNPs, defining 94 loci; 743 SNPs from these loci were available; Bonferroni threshold 0.05/94. LD Score Regression (CTG-VL) assessed across-sample SNP-h², genomic inflation, and genetic correlations with cognition and visual traits. Post-GWAS: FUMA v1.3.7 (MAGMA v1.08, ANNOVAR) for gene-based tests (p<2.7×10⁻⁶), gene-set enrichment (GO; Bonferroni p<6.8×10⁻⁶), mapping to functional annotations and GWAS Catalog, and eQTLs; FINDOR reweighted p-values based on functional enrichment. Polygenic score analyses: PRSice v2.3.3 constructed PGS for ASD, bipolar disorder, ADHD, schizophrenia, dyslexia, intelligence, educational attainment (EA), refractive error, and myopia. LD-clumping used r²<0.1 within 250 kb. Linear regressions predicted visual acuity from each PGS with covariates sex, age, two PCs, and SES (maternal education); n=5160. Multiple p-value thresholds were tested; the best-performing threshold and eight additional thresholds (0.001–1) were considered. Bonferroni correction across nine base GWAS and nine thresholds set significance at 0.00062. Data processing used R v4.1.2; scripts are available on GitHub.

- Visual acuity distribution: VAR range 85–120 (mean 107.38, SD 3.93). Boys had slightly higher acuity than girls (means 107.52 vs 107.24; d=0.07; p=0.004). Age showed a positive effect on visual acuity (B=0.02 per week; p=6.3×10⁻⁶). - Behavioural correlations: Visual acuity positively correlated with multiple cognitive measures including GCSE scores (all p<3.0×10⁻⁷). After adjusting for sex, age, reading skill, verbal IQ, performance IQ, and SES, the association with GCSE remained significant though attenuated (partial r=0.05; 95% CI [0.02, 0.08]; p=0.001). - Hearing independence: In n=4929, partial correlation (sex, age adjusted) between visual acuity and hearing threshold was not significant (r=−0.02; p=0.104). - SES: Significant main effect on visual acuity (F(4,6230)=17.78; p=1.6×10⁻14), with higher SES associated with better acuity. - Neurodevelopmental groups: Significant main effect (F(4,2666)=6.54; p=3.1×10⁻5); children with reading difficulty had reduced visual acuity compared to controls (means 106.71 vs 107.89; adjusted p=0.0001). Results robust after excluding outliers. - SNP heritability and within-sample genetic correlations: SNP-h² for visual acuity was 0.26 (SE 0.07; p=4.5×10⁻5). Significant positive genetic correlations were observed with listening comprehension, short-term memory, and GCSE scores. No significant genetic correlation with hearing threshold (rg=−0.56; SE 0.73; p=0.181). - GWAS: Twelve SNPs at the NPLOC4 locus on chromosome 17 reached genome-wide significance. Lead SNP rs11656126 (MAF=0.35) had p=2.1×10⁻8, with each G allele associated with +0.11 SD better visual acuity. A top suggestive marker on chromosome 5 (rs159195) lies intronic to PDE4D. Targeted replication identified two loci at Bonferroni significance: a region at 17:79.53–79.63 Mb including NPLOC4 and PDE6G (e.g., rs6420484 A allele associated with reduced acuity, β=−0.11; p=9.3×10⁻8), and an intergenic region upstream of PRSS56 on chromosome 2 previously associated with myopia/refractive error. - Gene-based and functional follow-up: NPLOC4 was the top gene (p=1.2×10⁻7). FUMA linked the NPLOC4 locus to prior sensory phenotypes (age-related macular degeneration, refractive error, astigmatism). eQTLs in the locus affect NPLOC4, PDE6G, and TSPAN10. FINDOR boosted signals at chromosomes 17 and 14. Gene-set enrichment identified “sensory perception of taste” (GO:0050909; p=3.2×10⁻9). - Across-sample LDSC: SNP-h² for visual acuity=0.18 (SE 0.09). Significant genetic correlations with cognitive performance (rg=0.32; SE 0.11; p=0.005) and educational attainment (rg=0.25; SE 0.09; p=0.006). No significant genetic correlations with UK Biobank visual measures, likely due to small SNP-h² and sample sizes for those traits. - PGS: EA PGS showed a robust positive association with visual acuity across thresholds; at the best threshold (p=0.0276), standardised β=0.05; PGS R²=0.28%; p=0.0001; effect remained after adjusting for myopia, refractive error, or intelligence PGS. Myopia PGS showed a negative association (best threshold p=0.0062), standardised β=−0.05; PGS R²=0.26%; p=0.0002. No PGS for neurodevelopmental disorders survived Bonferroni correction.

Findings demonstrate that better childhood visual acuity is associated with enhanced cognitive performance (including language-related skills) and higher academic achievement (GCSE), even after adjustment for multiple confounders. Behavioural and genetic correlations suggest shared biological underpinnings between visual function and cognition. The GWAS identified genome-wide significant variants at the NPLOC4 locus and implicated genes central to retinal function (e.g., PDE6G), aligning with prior adult vision GWAS and supporting biological relevance. The positive association of EA PGS with childhood visual acuity, contrasted with the established causal effect of longer education on increased myopia risk in adults, supports a model of gene–environment correlation: genetic propensity for higher education may initially confer cognitive advantages associated with slightly better visual acuity in childhood, but later environmental exposures linked to educational trajectories (e.g., near work, indoor time) may lead to deteriorating vision and myopia, thus reversing the early advantage. The higher SNP-heritability in children compared to adults further suggests that environmental influences on vision accumulate with age, diminishing the proportion of variance attributable to common genetic variation. The lack of association between visual acuity and hearing measures indicates domain-specific sensory contributions. Overall, results clarify developmental timing and mechanisms linking visual function with cognitive and educational outcomes.

The study confirms a positive association between childhood visual acuity and cognitive development and shows reduced acuity in children with reading difficulty. A GWAS identified genome-wide significant associations at the NPLOC4 locus, with gene-based analyses corroborating these findings and implicating retinal pathways (including PDE6G). Polygenic analyses revealed that EA PGS is associated with slightly higher childhood visual acuity, while myopia PGS relates to lower acuity, supporting a gene–environment correlation model in which educationally linked environments later increase myopia risk. Future research should employ longitudinal designs, more granular visual phenotypes (including uncorrected refractive status and device use), and direct measures of environmental exposures (e.g., near work and outdoor time) to disentangle causal pathways and developmental dynamics.

Key limitations include: use of maternal education as a single proxy for SES, which may not capture all socioeconomic dimensions (e.g., income, wealth, access to healthcare); single time-point assessment of vision and educational outcomes limiting causal inference and developmental trajectories; best-corrected visual acuity measures prevent distinguishing children wearing correction during assessment or attributing reduced acuity to uncorrected refractive error; potential residual confounding from unmeasured variables; and a comparatively modest sample size for GWAS relative to adult studies, which may limit discovery power despite successful replication of key loci.