Researchers at the University of Cambridge have accelerated the screening process for novel Parkinson’s disease treatments by leveraging the potential of artificial intelligence (AI).
The approach has the potential to accelerate the drug development process tenfold, as it revealed five very potent prospective therapeutic candidates for further research.
Nature Chemical Biology publishes the research.
Creating therapies that alter disease
Worldwide, more than 6 million people suffer from Parkinson’s disease. It can produce a wide range of symptoms, from the typical movement signs to issues with the gastrointestinal, sleep, mood, and cognitive function.
By 2040, the number of Parkinson’s patients is predicted to triple, making it the neurological disorder with the fastest rate of growth in the world. Parkinson’s disease still has no authorized disease-modifying therapies, despite the condition’s increasing prevalence. These therapies try to ameliorate the disease’s symptoms by specifically targeting the mechanisms that cause the illness.
It is believed that renegade proteins, which misfold and aggregate to create Lewy bodies, are the origin of Parkinson’s disease. These aggregates eventually accumulate within nerve cells, impairing function or even causing cell death.
Trials for possible Parkinson’s disease-modifying medications are underway, but the absence of experimental techniques to pinpoint the right molecular targets has hindered the development of these treatments due to a technological gap.
Prof. Michele Vendruscolo, the study’s lead author and a professor of biophysics in Cambridge’s Yusuf Hamied Department of Chemistry, stated that finding small molecules that can prevent alpha-synuclein from aggregating is one way to look for potential Parkinson’s disease treatments. However, this is a very time-consuming process; it may take months or even years to identify a lead candidate for additional testing.
In a recent study, Vendruscolo and colleagues at the University of Cambridge used artificial intelligence (AI) to reduce and expedite the expenses related to Parkinson’s disease treatment development.
AI-powered iterative screening
The scientists used a method based on machine learning to filter libraries of millions of chemical compounds in order to find potential candidates that attach to the protein aggregates and stop them from growing.
The most potent compounds that suppressed protein aggregation were identified by experimental testing of the top-ranking compounds. The machine learning model was then iterated with this data in order to determine which candidate compounds were the best.
“We screen computationally, as opposed to experimentally,” Vendruscolo said. “We were able to train our machine learning model to identify the specific regions on these small molecules responsible for binding by using the knowledge we gained from the initial screening. This allows us to rescreen and find more potent molecules.”
Using this technique, the scientists were able to create chemicals that target the pockets on the aggregate surface that facilitate their growth. Compared to earlier examples, these chemicals are both considerably more potent and less expensive to create.
“The process of finding the most promising candidates is being sped up by machine learning, which is having a significant impact on the drug discovery process,” Vendruscolo stated. This allows us to begin working on several drug development projects rather than just one. It’s an exciting time since so much is possible because of the significant reduction in both time and expense.
