Nature Neuroscience study combines genetics, medical records and disease models to map drug targets for Alzheimer’s disease

Nature Neuroscience study combines genetics, medical records and disease models to map drug targets for Alzheimer’s disease

The authors used genetic data to select candidate proteins, then assessed drugs using de-identified prescription records. For one pathway involving EPHX2, they traced the evidence from a gene variant through human cells to a mouse model.

The study, published on 8 October 2026 in Nature Neuroscience, addresses a practical question: how to identify promising protein targets for new Alzheimer’s drugs before undertaking expensive clinical trials. The authors started with 1 229 proteins already targeted by drugs or investigational compounds and assessed the candidates through a series of checks.

The first step was Mendelian randomization, which uses inherited DNA variants associated with protein levels to assess the direction of the relationship between a protein and disease risk. Using seven genomic studies that included 275 540 Alzheimer’s disease cases and 1,55 million control participants, the authors identified 19 priority drug targets in populations of European ancestry and seven potential targets in populations of African ancestry. A drug target is a protein that a drug can act on; these findings are specific to the ancestry groups studied.

The second step used de-identified insurance records of drug prescriptions. Among 111 680 people with mild cognitive impairment, meaning early difficulties with memory and thinking, the authors assessed 210 commonly prescribed drugs, accounting for age, sex, ancestry and coexisting conditions. Twelve drugs were associated with a lower rate of subsequent Alzheimer’s diagnosis across all four statistical analyses. The authors propose further investigation of these drug and target pairs.

EPHX2 served as an example of the full sequence of checks. This gene encodes an enzyme that breaks down protective fatty acids; the authors link its activity to inflammation and synaptic damage. In human neurons and miniature brain models grown from patient cells, the gene variant p.Arg287Gln was associated with lower levels of EPHX2 and phosphorylated tau, a disease-associated form of tau. In 5xFAD mice, a model with accelerated amyloid deposition, the EPHX2 inhibitor EC5026 reduced amyloid deposits and inflammatory signals associated with microglia, the brain’s immune cells. The animals also performed better on memory tests.

The authors propose testing the selected targets and drugs in clinical studies that use biomarkers and account for genetic differences between patients.