Artificial intelligence techniques have been utilized by researchers to significantly speed up the search for Parkinson’s disease treatments.
The team from the University of Cambridge developed and implemented an AI-based approach to pinpoint compounds that can prevent the clumping, or aggregation, of alpha-synuclein, the protein associated with Parkinson’s.
The team utilized machine learning techniques to rapidly analyze a vast chemical library with millions of entries and pinpointed five extremely powerful compounds for further examination. Parkinson’s impacts over six million individuals globally, and that number is anticipated to triple by 2040. Presently, there are no available treatments that modify the disease. The process of screening extensive chemical libraries for potential drugs is very time-consuming and costly, and needs to occur long before potential treatments can be tested on patients.Using machine learning, the researchers were able to accelerate the initial screening process by ten times and lower the cost by a thousand times, potentially leading to faster access to Parkinson’s treatments for patients. The findings are published in the journal Nature Chemical Biology.
Parkinson’s is the most rapidly growing neurological condition globally. In the UK, one in 37 people currently living will receive a Parkinson’s diagnosis in their lifetime. In addition to affecting motor symptoms, Parkinson’s can also impact the gastrointestinal system, nervous system, sleep patterns, mood, and cognition.
Parkinson’s disease can lead to a decreased quality of life and significant disability.
Proteins play a crucial role in cell processes, but in Parkinson’s patients, these proteins become abnormal and cause the death of nerve cells. When proteins misfold, they can form clusters known as Lewy bodies, which accumulate within brain cells and disrupt their normal function.
“One approach to finding potential treatments for Parkinson’s involves identifying small molecules that can prevent the aggregation of alpha-synuclein, a protein closely linked to the disease,” said Pro.Professor Michele Vendruscolo from the Yusuf Hamied Department of Chemistry, who led the research, stated that the process of identifying a lead candidate for further testing can be extremely time-consuming, taking months or even years. Currently, there are ongoing clinical trials for Parkinson’s, but no disease-modifying drug has been approved yet, indicating the challenge of directly targeting the molecular species that cause the disease. This lack of methods to identify the correct molecular targets and engage with them has been a major obstacle in Parkinson’s research, creating a technological gap that has severely hampered progress.
The progress in Alzheimer’s disease research has been hindered by the lack of effective treatments.
Researchers at Cambridge have created a machine learning technique that involves screening chemical libraries with millions of compounds to discover small molecules that can bind to the amyloid aggregates and prevent their growth.
After a limited number of top-performing compounds were tested in experiments to determine the most effective inhibitors of aggregation, the results were used to improve the machine learning model. This iterative process led to the identification of highly potent compounds after a few rounds.
“Instead of conducting experimental screenings, the team utilized a machine learning approach to identify potential inhibitors,”.”We conduct computational screening,” explained Vendruscolo, co-Director of the Centre for Misfolding Diseases. “By utilizing the information we acquired from the initial screening with our machine learning model, we managed to train the model to recognize the specific areas on these small molecules that are responsible for binding. This allows us to re-screen and discover more powerful molecules.”
Using this approach, the Cambridge team created compounds to target pockets on the surfaces of the aggregates, which are accountable for the rapid proliferation of the aggregates themselves. These compounds are significantly more potent and less expensive to develop, being hundreds of times more effective.
Recent discoveries in machine learning have significantly accelerated the drug discovery process, leading to the identification of more promising candidates in a shorter amount of time. Vendruscolo, a researcher involved in the study, expressed excitement about the potential of this advancement, stating that it allows for the initiation of multiple drug discovery programs simultaneously, ultimately reducing both time and cost. The study was carried out at the Chemistry of Health Laboratory in Cambridge, which received support from the UK Research Partnership Investment Fund (UKRPIF) to facilitate the transition of academic research into clinical applications.