38. INTEGRATIVE MULTI-OMICS AND DEEP LEARNING ANALYSIS IDENTIFIES CTNND2 AND NETO1 AS KEY NEURODEVELOPMENTAL DISORDER RISK GENES IN A PEDIATRIC COHORT OF 68,975 INDIVIDUALS

Background Neurodevelopmental disorders (NDDs) are highly heterogeneous psychiatric conditions with complex genetic architectures and limited mechanistic understanding. Although genome-wide association studies (GWAS) have identified multiple susceptibility loci, translating these findings into biological insight and therapeutic opportunities remains challenging. We performed an integrative genomic and transcriptomic analysis in one of the large pediatric cohorts to identify risk variants, regulatory mechanisms, and biologically relevant pathways underlying NDD susceptibility. Methods We analyzed genotype and phenotype data from 68,975 individuals enrolled through the Center for Applied Genomics at Children’s Hospital of Philadelphia, including 30,972 neurodevelopmental disorder cases and 38,003 controls. Genotypes were imputed to the TOPMed reference panel, followed by stringent quality control, retaining 4.2 million high-quality variants. Genome-wide association analyses were performed using logistic regression in PLINK2, adjusting for sex, ancestry principal components, and genotyping batch effects. Population-specific analyses were conducted in European and African ancestry cohorts. Polygenic risk scores were generated using PRS-CS. Functional annotation and gene-mapping analyses were performed using FUMA and MAGMA with GTEx brain transcriptomic datasets. Transcriptome-wide association studies were conducted using S-MultiXcan with GTEx v8 brain tissue models. Protein interaction and pathway analyses were explored using STRINGDB. To prioritize potential therapeutic targets, we applied a graph neural network-based deep learning framework integrating genetic associations, transcriptomic enrichment, and protein interaction networks for drug repurposing analysis. Results We identified a genome-wide significant locus within CTNND2 on chromosome 5 (chr5:10986596:C; P = 2.2 × 10^-30^, OR = 1.89), a gene implicated in dendritic spine development and synaptic signaling. Ancestry-stratified analyses identified NETO1 (chr18:72819414:G) as a shared cross-population risk locus in both European and African ancestry cohorts (P = 3.05 × 10^-11^, OR = 0.91), supporting a conserved neurodevelopmental role across ancestries. PRS analyses demonstrated significantly elevated polygenic burden in NDD cases compared with controls. TWAS identified multiple brain-associated genes including GIT1, CHD1L, and FAM234B, which are involved in synaptic organization, neuronal signaling, and chromatin regulation. Network analyses revealed functional interactions connecting CTNND2 and NETO1 with FAM234B, suggesting convergence within neurodevelopmental signaling pathways. MAGMA tissue enrichment analyses demonstrated significant expression enrichment in prenatal and early postnatal brain regions including the prefrontal cortex, hippocampus, basal ganglia, and cerebellum. Discussion Our integrative multi-omics framework identified CTNND2 and NETO1 as robust contributors to NDD susceptibility while highlighting convergent synaptic and neurodevelopmental pathways across diverse ancestries. By combining GWAS, TWAS, PRS, tissue-specific enrichment, and network biology with deep learning-based drug repurposing, this study advances the biological interpretation of psychiatric genetic risk and provides a scalable framework for precision psychiatry and therapeutic target discovery in neurodevelopmental disorders.