Machine learning approaches can benefit Parkinson’s research: Study
1 method found to help ID 'more potent' treatment candidates in lab
Scientists exploring the potential of machine learning approaches in drug discovery for Parkinson’s disease and other neurodegenerative disorders — focusing on misfolded proteins that are the hallmark of such conditions — found that one such method identified compounds “two orders of magnitude more potent” than ones previously reported, per a new study.
Using this method allowed the researchers, from the U.K. and the U.S., to identify compounds that can effectively block the clumping, or aggregation, of alpha-synuclein protein, an underlying cause of Parkinson’s, the study reported.
“We anticipate that using machine learning approaches of the type described here could be of considerable benefit to researchers working in the field of protein misfolding diseases [such as Parkinson’s], and indeed early-stage drug discovery research in general,” the researchers wrote.
Their study, “Discovery of potent inhibitors of α-synuclein aggregation using structure-based iterative learning” was published in the journal Nature Chemical Biology.
Machine learning approaches ‘speeding up the whole process’ of drug discovery
Parkinson’s disease is marked by the toxic accumulation of misfolded forms of the alpha-synuclein protein within dopamine-producing nerve cells — those responsible for releasing the neurotransmitter dopamine. Dopamine is a signaling molecule that plays a role in controlling movement; Parkinson’s results from the progressive loss of these cells.
Despite efforts to identify compounds that stop this toxic accumulation, there are, to date, no disease-modifying treatments available for Parkinson’s.
Traditional strategies to identify novel therapies — which involve screening large chemical libraries looking for potential candidates prior to any testing in humans — are time-consuming, expensive, and often unsuccessful.
In the case of Parkinson’s, the development of effective therapies has been hampered by the lack of methods to identify the right molecular targets.
“One route to search for potential treatments for Parkinson’s requires the identification of small molecules that can inhibit the aggregation of alpha-synuclein. … But this is an extremely time-consuming process — just identifying a lead candidate for further testing can take months or even years,” Michele Vendruscolo, a professor at the University of Cambridge and the study’s lead author, said in a university press release.
Now, the researchers developed a method that was able to use machine learning to quickly screen chemical libraries containing literally millions of compounds. The goal was to identify small molecules able to block the clumping of alpha-synuclein.
From a list of small molecules predicted to have a good binding to the alpha-synuclein aggregates, the researchers chose a small number of the top-ranking compounds to test experimentally as potent inhibitors of aggregation.
The results from these experimental assays were then fed to the machine learning model, which identified those with the most promising effects. This process was repeated a few times, so that highly potent compounds were identified.
“Instead of screening experimentally, we screen computationally,” Vendruscolo said.
Machine learning is having a real impact on the drug discovery process — it’s speeding up the whole process of identifying the most promising candidates. For us this means we can start work on multiple drug discovery programs — instead of just one.
“By using the knowledge we gained from the initial screening with our machine learning model, we were able to train the model to identify the specific regions on these small molecules responsible for binding, then we can re-screen and find more potent molecules,” Vendruscolo said.
Using this method, the researchers optimized the initial compounds to target pockets on the surfaces of the alpha-synuclein clumps.
In lab tests using brain tissue samples from patients with Lewy body dementia (LBD) and multiple system atrophy (MSA), two forms of atypical parkinsonism, the compounds effectively blocked aggregation of alpha-synuclein.
“Machine learning is having a real impact on the drug discovery process — it’s speeding up the whole process of identifying the most promising candidates,” Vendruscolo said. “For us this means we can start work on multiple drug discovery programs — instead of just one.”
According to Vendruscolo, “so much is possible due to the massive reduction in both time and cost – it’s an exciting time.”
About the Author
