Neuroanatomically Derived Genetic Representations Enhance Detection of Plasma Protein Associations

Abstract

Genetic variation shapes brain structure, yet it remains unclear whether this neuroanatomical expression of genotype is reflected in circulating proteomic profiles, which provide functional molecular readouts of biological processes. Addressing this question requires integrating genetic, neuroimaging, and proteomic data within the same individuals, which is methodologically non-trivial. To address this challenge, we used brain-genotype scores, an approach recently developed in our lab that allows the creation of individual-level, SNP-specific neuroanatomical representations of genotype learned directly from whole-brain structural MRI. Adapting these scores to proteomics, we tested associations in UK Biobank between 120 brain-genotype scores and plasma levels of 2,920 proteins, followed by pathway and tissue-enrichment analyses to assess biological coherence. After FDR correction, brain-genotype scores yielded 116 significant genomic-neuroimaging-proteomic associations across 49 scores and 52 proteins; none were detected using conventional SNP dosage models, and they explained substantially more variance in protein levels than genotype alone. Enrichment analyses identified convergent immune, metabolic, signalling, and cell-cycle pathways, with tissue enrichment spanning brain, liver, pancreas, and hypothalamus. Several identified proteins overlapped with prior imaging-proteomic literature, supporting biological plausibility. These findings demonstrate that brain-genotype scores reveal biologically meaningful proteomic variation beyond conventional genotype analyses, providing a framework for linking genetic variation, brain structure, and the circulating proteome.