10 July 2026

Amsterdam UMC

AI improves genetic risk score for atrial fibrillation detection across diverse populations

Health & Care

Atrial fibrillation is a common heart condition affecting more than half a million people in the Netherlands, increasing the risk of heart failure and stroke. Researchers at Amsterdam UMC have developed a new polygenic risk score (PRS) using artificial intelligence and genetic data from multiple ethnic backgrounds, with the aim of reducing healthcare inequality and improving risk prediction.

Atrial fibrillation has a strong genetic component, yet many patients and doctors remain unaware of its hereditary nature. A PRS calculates an individual’s inherited risk based on many DNA variations, but previous scores were developed exclusively using data from European populations, leaving other ancestry groups underserved. “Our findings highlight the important contribution of genetics to atrial fibrillation risk,” says Dr. Sean Jurgens of Amsterdam UMC.

The new PRS was developed using genetic information from European, Latin American, African, South Asian, and East Asian populations. Compared to previous scores, the novel score markedly improved risk prediction accuracy for people of African, Latin American, and South Asian descent. For Europeans and East Asians, the score was able to identify approximately ten percent of individuals with a more than fourfold increased risk of atrial fibrillation. “By including diverse genetic backgrounds, we significantly improved risk prediction for everyone,” explains Poeya Haydarlou.

Earlier identification of high-risk individuals could enable targeted screening and more intensive follow-up. Dr. Jurgens notes that genetic diversity is key to reducing inequality in healthcare, and that the method developed ensures that available data serve people from all backgrounds more equally. He adds, however, that prediction remains strongest for Europeans and East Asians, and that additional work is still needed to reduce the remaining disparities further.

A logical next step is implementing PRS in clinical settings. In the United States, PRS tests are already being made available to clinicians and patients. Because genetic risk is largely fixed at conception, PRS offers the prospect of early identification and personalised care for high-risk patients. Haydarlou notes that additional work is still needed to demonstrate how and where PRS can change clinical decision-making or preventive counselling. The research was published in Nature Communications and funded by a Dekker Grant from the Dutch Heart Foundation.

Read also

UvA research: Google’s AI results favor GroenLinks-PvdA and D66

25 October 2025

University Pride lecture: creating an AI rainbow code

4 July 2023

Nebius secures $27 billion AI infrastructure deal with Meta

16 March 2026

Subscribe to our newsletter

Would you like to stay informed about what is happening within the Amsterdam AI ecosystem?