AI-powered study finds multiple Parkinson’s types
Dr Natalie Kaempf and Prof Patrik Verstreken.
Researchers have used a machine learning algorithm to try to work out why Parkinson's disease patients can respond so differently to treatment, and guide the development of personalised medicines.
The team, led by scientists at VIB and KU Leuven in Belgium, have identified two types of Parkinson's – and five subtypes – that can be differentiated by their molecular signatures. They contend that Parkinson's should now be thought of as a collection of similar conditions with diverse underlying biological mechanisms.
Writing in the journal Nature Communications, the researchers used fruit fly models of various forms of Parkinson's with selected mutations, to observe directly how the variant and their specific upstream molecular causes affected behaviour, rather than making assumptions about their effects.
"We came in without any preconceived notions of how a specific mutation would affect our animal model," said Dr Natalie Kaempf of KU Leuven, the first author of the paper. "We took animals with mutations in any of those 24 different genes that are causing the disease, and we just monitored their behaviour over periods of time."
That allowed the scientists to identify clusters of mutant genes that separated into two main camps: one group affecting mitochondrial function with two subtypes, and a second with three subtypes involving a mechanism by which molecular cargo is transported across cellular membranes, helping to maintain the integrity of the proteome.
They were also able to demonstrate varying responses to treatment between the various types, with medicines correcting movement disorder symptoms in some, but having no activity in another.
"When we took a first compound that cured subgroup A and tested it in subgroup B, the latter wasn't rescued," said corresponding author Prof Patrik Verstreken of VIB-KU Leuven Centre for Neuroscience. "Our study shows that you can make subgroup-specific drugs that have positive effects and are really specific to that subgroup."
The paper also points out that one group of flies responded to coenzyme Q10, an approach that was found to be ineffective in previous phase 3 trials in early Parkinson's that included all-comer patient populations.
The findings are the first step in an effort to translate these findings to human patients, starting with familial forms of the disease and expanding to idiopathic cases later on, working towards targeted biomarker discovery, stratification in clinical trial design, and subtype-specific therapeutic development.
"The same principle can be applied to other types of diseases," added Verstreken. "Diseases that are caused by mutations in a variety of different genes or environmental factors could be classified according to this principle."
