<b>Title: Meta-sGWAS: Integrating brain spatial interactions to uncover genetic variants in Bipolar Disorder</b>

Bipolar disorder (BD) is a severe, highly heritable psychiatric disorder driven by the interplay of genetic liability, brain structural and functional abnormalities, and environmental influences. Conventional genome-wide association studies (GWAS) have identified BD-associated risk loci but typically neglect the spatial heterogeneity and regional interaction architecture of the human brain, resulting in limited statistical power, inflated confounding bias, and weak biological interpretability. Here we present Meta-sGWAS, a spatial-informed GWAS framework that integrates individualized brain regional interaction patterns into genetic association modeling to dissect the genetic basis of BD with improved accuracy and mechanistic depth. Using multimodal brain MRI and genomic data from the Adolescent Brain Cognitive Development (ABCD) study and UK Biobank (UKB), Meta-sGWAS first identifies BD-relevant brain regions of interest (ROIs) and their functional connectivity, notably the right subcallosal gyrus (P = 9.9 x 10-9) and left paracentral gyrus/sulcus (P = 3.6 x 10-8), as well as dysregulated connectivity in the right orbitofrontal cortex underlying emotional dysregulation. Heritability estimation and Mendelian randomization confirm robust genetic contributions and causal links between these ROIs and BD susceptibility. By embedding brain spatial interaction kernels as random effects in generalized linear mixed models, Meta-sGWAS substantially outperforms standard GWAS in detecting low-effect yet biologically critical variants. Single-nucleus RNA-seq validation refines BD genetic risk signals specifically to cortical inhibitory interneuron subtypes (Pvalb, Vip, Lamp5 Lhx6) and highlights CACNA1C as a core BD driver gene with cross-brain regulatory dysregulation: upregulated in prefrontal inhibitory neurons and downregulated in the cerebellum via the functional variant rs1006737. Meta-sGWAS also enhances association signals for established BD risk loci including TRANK1. Collectively, this study establishes a spatially integrated imaging genetics paradigm that uncovers replicable BD risk variants, defines disease-relevant cell types, and provides a mechanistic model for BD pathogenesis. The Meta-sGWAS framework is generalizable to other brain-related complex disorders and accelerates the translation from GWAS loci to biological mechanisms and clinical biomarkers. Meta-sGWAS can refine GWAS results and identify SNPs with significant biological relevance. Before and after performing Meta-sGWAS, we obtained GWAS results for bipolar disorder (BD) with and without incorporating brain structural traits using the ABCD and UKB datasets, namely ABCD-GWAS (without brain structural traits), ABCD-Meta-sGWAS (with brain structural traits), UKB-GWAS (without brain structural traits), and UKB-Meta-sGWAS (with brain structural traits).