Mendelian randomization (MR) applied to multiple AD GWAS and functional multiomics datasets was used to nominate druggable targets, with functional validation of EPHX2 inhibition in iPSC-derived neurons, cerebral organoids, and a 5xFAD mouse model; separately, electronic health records from 111,680 MCI patients were screened for drugs associated with reduced AD incidence.
19 druggable targets were identified for European ancestry and 7 for African ancestry; the EPHX2 protective missense variant p.Arg287Gln reduced EPHX2 protein and pTau-181 levels in AD patient iPSC-derived neurons; pharmacological EPHX2 inhibition improved cognition in 5xFAD mice; among MCI patients (23.7% developed AD), 12 drugs were associated with reduced AD incidence, including trazodone (targeting ADRA1A) and baclofen (targeting GABBR1).
The alzMR score inflation assumption (≥1% heritability explained by MR instruments) may not hold for ~4–8% of genes with low instrument strength. MR estimates reflect total causal effects only, and compensatory SNP interactions or opposing causal pathways could bias results. The pharmacoepidemiology analysis used observational insurance claims data, so residual confounding cannot be excluded despite propensity score matching.
EPHX2 inhibition is a promising therapeutic direction for AD, supported by genetic, cellular, organoid, and mouse model evidence. Trazodone and baclofen, both already in clinical use, warrant prospective study as potential agents to slow AD progression in MCI patients.