From GWAS to Mechanism: Synaptic Pruning Emerges as a Key Polygenic Driver of Cognitive Ability

Human intelligence is strongly heritable, but the genes that fine-tune the brain's wiring are still being mapped. We re-examined the large IQ genome-wide association study by Savage et al. (2018; N = 269 867) with three complementary tools—partitioned SNP-heritability, MAGMA gene tests, and a transcriptome-wide association study (TWAS)—to ask whether synaptic pruning makes an independent contribution beyond classic glutamatergic signalling. Seven pre-registered gene sets were contrasted: two glutamate lists, two pruning lists, two negative-control lists (monoamine and housekeeping genes), and a "pruning-only" set that removed every glutamatergic gene. Heritability was significantly enriched in both glutamate and pruning sets, but the pruning-only panel still carried a clear signal (1.10-fold, P ≈ 5 × 10⁻¹⁵), showing that pruning effects are not simply spill-over from glutamate pathways. MAGMA supported this pattern, highlighting SEMA3F, RHOA, MAP1LC3B and TCF4 after Bonferroni correction. TWAS added tissue context: pruning genes showed the strongest over-representation (core set 1.38-fold, P ≈ 2 × 10⁻⁵), with RHOA down-regulated in caudate and SEMA3F up-regulated in anterior cingulate. Together, the results inspire a "Calibrated Pruning Framework." We propose that common variants adjust the timing of a multistep pruning cascade—TCF4 keeps critical periods open a little longer, SEMA3F–RHOA steers branch retraction, MAP1LC3B clears debris, and HLA tags mark synapses for removal—thereby fine-tuning network efficiency and, ultimately, cognitive ability. Limits of the work include the European bias of the base GWAS and reliance on adult-brain expression panels; future longitudinal imaging and multi-ancestry studies will be needed to test the model's predictions.