Resolving the SCZ–T2DM Paradox: Convergent Pathways Despite Genome-Wide Antagonism

Background: Schizophrenia (SCZ) and type 2 diabetes mellitus (T2DM) share a well-documented clinical comorbidity, with individuals with SCZ facing a 1.5- to 3-fold increased risk of T2DM even in the absence of antipsychotic treatment. Paradoxically, genome-wide association studies have consistently shown a modest negative genetic correlation (r_g ≈ −0.04 to −0.05) between the disorders, raising questions about how strong epidemiological overlap can coexist with apparent genetic antagonism. Methods: We integrated European-ancestry summary statistics from the Psychiatric Genomics Consortium Wave 3 SCZ GWAS (effective N ≈ 58,749) and a large T2DM meta-analysis (effective N ≈ 751,755). Eight partially overlapping gene sets were curated to capture core domains of cellular resilience: synaptic pruning, glutamatergic signalling, kindling/cellular scarring, and somatic mutation drivers (mCA-related, mitotic stress, clonal haematopoiesis), with housekeeping and monoaminergic sets as controls. Analyses included MAGMA for gene-based and competitive set testing, stratified LD-score regression for partitioned heritability and bivariate genetic correlation, S-PrediXcan transcriptome-wide association in brain tissues for T2DM and peripheral metabolic tissues for SCZ, and both standard GSEA and differential GSEA to compare pathway profiles. Results: MAGMA revealed symmetric enrichment of the same four hypothesis-driven pathways in both disorders, with pruning, kindling, glutamatergic, and mCA-related sets achieving Bonferroni significance. GSEA confirmed coordinated signals (e.g., MHC Class I pruning leading in both), and DGSEA showed no differential enrichment across 28 pairwise comparisons (all p_diff = 1.0). LDSC confirmed the negative genome-wide correlation (r_g = −0.049, p = 0.005). TWAS highlighted tissue-specific convergence: mCA and clonal-haematopoiesis genes were most strongly up-regulated in brain for T2DM, while pruning components dominated peripheral expression in SCZ. Conclusions: These findings support a model of convergent pleiotropy with antagonistic polygenic architecture: shared risk alleles converge on the same cellular resilience pathways despite opposing genome-wide effects, creating a common vulnerability amplified by environmental and somatic factors. The framework reconciles the SCZ–T2DM paradox and points to pathway-specific polygenic scores and targeted therapies (GLP-1 receptor agonists, complement modulators, senolytics) for dual brain–metabolic benefit.