Early diagnosis of Parkinson’s disease is important because many commonly used treatments for the condition are more effective when administered early on. New research suggests that a new artificial intelligence tool can help diagnose Parkinson’s disease well before symptoms manifest, with 96 percent accuracy.

Researchers used machine learning to analyze metabolites – byproducts of metabolism – in the blood of 78 people, half of whom had Parkinson’s disease.

They found that the machine learning tool was approximately 96 percent accurate when differentiating between healthy people and those with Parkinson’s disease, based on their blood metabolites. They also found that people with Parkinson’s disease were more likely to have high levels of a poly-fluoroalkyl substance in their blood and low levels of triterpenoids, cholestane steroids, and diacylglycerol.

Polyfluoroalkyl substances, also known as PFAS, are man-made chemicals used in food packaging, household products, and drinking water. PFAS are not excreted in bodily fluids like sweat or urine; rather, they persist in the body for indefinite periods and are often referred to as “forever chemicals.” Triterpenoids, cholestane, and diacylglycerol are plant-based bioactive dietary compounds that exert antioxidant and neuroprotective effects.

These findings suggest that machine learning is a useful tool in detecting Parkinson’s disease early based on metabolites in blood. They also highlight possible interventions to reduce the risk of developing the disease, such as reducing exposure to PFAS and including plant-based bioactive compounds in the diet. Learn more about Parkinson’s disease in this episode featuring Dr. Giselle Petzinger.

• Read full publication.