Multi-ancestry score improves Alzheimer’s risk prediction

A number assembled from DNA markers is being asked to do a job that earlier Alzheimer’s risk scores handled unevenly: estimate susceptibility across populations, not just in people of European ancestry. Researchers at Boston University developed and validated a multi-ancestry polygenic risk score (PRS), a numerical estimate of genetic risk, and found it was more predictive of a clinical Alzheimer’s diagnosis across all populations examined.

The model was built from genetic-marker data covering more than 63,000 Alzheimer’s cases and 484,000 age-matched controls. Researchers tested it separately in 10,612 cases and 16,625 older controls, then validated it in another mixed-ancestry group of 1,500 cases and 75,500 older controls. Compared with earlier European-ancestry scores, the multi-ancestry model was especially predictive among African American, Hispanic and East Asian participants.

The score’s signal also appeared before a diagnosis. Higher genetic risk was associated with poorer memory, executive function and language performance, a smaller hippocampus—the brain region most affected early in Alzheimer’s—on magnetic resonance imaging, and abnormal levels of amyloid-beta and phosphorylated tau, or pTau, in cerebrospinal fluid. People with very high scores experienced the steepest cognitive decline before Alzheimer’s onset, and the study found a greater pTau deviation among women.

The practical shift is straightforward: a risk model that travels better across ancestries could help researchers identify participants for clinical trials and study early biological changes in groups that older models represented less accurately. It could also inform personalized intervention and prevention research. The score remains a research result, however—not a diagnosis or a treatment—and the reported measurements come from study cohorts rather than a clinical service in routine use.

Lindsay A. Farrer, chief of biomedical genetics at Boston University Chobanian & Avedisian School of Medicine, identified the limited representation of diverse ancestries in Alzheimer’s genome-wide association studies as a central obstacle. Xiaoling Zhang, an associate professor of medicine at the school, said the links to early cognitive and biological changes support the score’s use for long-range risk prediction. The findings were published in Nature Genetics.