F122. SYSTEMATIC STRUCTURAL CHARACTERIZATION OF MISSENSE MUTATIONS IN SCHIZOPHRENIA RISK GENES

Abstract

Background Schizophrenia (SCZ) is a highly heterogeneous psychiatric disorder in which rare missense variants are increasingly recognized as major contributors to disease risk. Recent advances in next-generation sequencing (NGS), AlphaFold-based protein structure prediction, and machine learning provide unprecedented opportunities to investigate how genetic variants disrupt protein structure and function. To address this challenge, we are developing the Platform for Analyzing Mutations associated with Schizophrenia (PAMS), an integrative computational framework that combines structural bioinformatics, computational saturation mutagenesis, protein stability analysis, protein-protein interaction modeling, and pathogenicity prediction to systematically characterize schizophrenia-associated mutations across SCZ risk genes. Methods PAMS integrates AlphaFold structural models with computational saturation mutagenesis and structure-based energy calculations to evaluate the effects of missense mutations on protein stability and protein-protein interactions. Folding free energy changes (ΔΔG) and binding energy changes (ΔΔΔG) were calculated using structure-based modeling approaches, while pathogenicity was assessed using machine learning predictors. As proof-of-concept applications, we analyzed mutations in DISC1, GRIN2A, and CUL1, three schizophrenia-associated genes involved in neuronal signaling, synaptic regulation, and ubiquitin-mediated protein degradation. Results In DISC1, computational saturation mutagenesis generated 16,226 missense mutations and identified highly destabilizing variants, including A697W and D694W, within the NDEL1 interaction region. These mutations were consistently predicted to be pathogenic and deleterious across multiple prediction tools. In GRIN2A, nearly half of all missense variants were predicted to destabilize the NMDA receptor subunit, with strong enrichment of disruptive mutations at the GRIN2A–GRIN1 interaction interface. Glycine residues G458 and G532 were particularly sensitive to bulky substitutions, resulting in extreme destabilization. In CUL1, schizophrenia-associated mutations were predicted to disrupt both protein stability and binding interactions within Cullin-RING ligase complexes, suggesting impaired ubiquitin-proteasome system function. Collectively, these analyses identified structurally disruptive variants and functionally sensitive regions across multiple SCZ risk proteins. Discussion These findings demonstrate the utility of PAMS as a scalable platform for systematic structural characterization of schizophrenia-associated missense mutations. By integrating protein structural modeling, energetic analysis, and machine learning-based pathogenicity prediction, PAMS enables prioritization of rare variants with potential functional and disease relevance. This framework provides mechanistic insight into how missense mutations alter protein stability and interaction networks in schizophrenia and establishes a broadly applicable strategy for investigating disease-associated variants across psychiatric and neurodevelopmental disorders.