AI could accelerate drug development for Parkinson's

Researchers at the University of Cambridge used AI (artificial intelligence ) to accelerate the development of treatments for Parkinson's disease . The results of the study were published in the scientific journal Nature Chemical Biologythis Wednesday (17).

At search, scientists developed and used an AI-based strategy to identify compounds that block the clumping of alpha-synuclein, a protein that, when aggregated in the brain, characterizes Parkinson's disease. The technique used was the machine learning who quickly screened a chemical library and identified five potent compounds for further study.

Currently, still There are no treatments for Parkinson's , which affects more than six million people around the world. The process of screening large chemical libraries for compounds that can serve as medicines is time-consuming, expensive, and often unsuccessful.

Therefore, Cambridge researchers decided to use AI to speed up this screening process and managed to optimize it 10 times, reducing the cost by a thousand times . This makes it possible for potential treatments for Parkinson's disease to reach patients much more quickly.

The importance of the study

Parkinson's is a neurological disease that mainly affects the patient's movements. The disorder is usually characterized by difficulty walking and speaking, slow movement, loss of balance and slurred speech, in addition to non-motor symptoms, such as changes in the gastrointestinal system, changes in sleep, mood and cognition.

In Parkinson's, the Accumulation of certain proteins can cause nerve cell death as is the case with the agglomeration of alpha-synuclein.

“A path towards finding potential treatments for Parkinson's requires the identification of small molecules that can inhibit the aggregation of alpha-synuclein, which is a protein closely associated with the disease,” explains Michele Vendruscolo, researcher at the Department of Chemistry Yusuf Hamied and research leader, in a press release. “But this is an extremely time-consuming process – just identifying a lead candidate for further testing can take months or even years.”

Although there are already drugs in clinical trials for Parkinson's, no medicine has yet been approved which reflects the difficulty of directly targeting the molecules that cause the disease.

In the current study, researchers developed a machine learning method to screen chemical libraries containing millions of compounds and identify small molecules that bind to aggregated proteins and block their proliferation in the brain.

Therefore, a small number of compounds were experimentally tested to select the most potent protein aggregation inhibitors. The information obtained from these experimental tests was fed back to the AI ​​in an interactive way, so that, after some interactions, highly potent compounds were identified.

“Instead of doing experimental screening, we do computational screening,” says Vendruscolo, who is co-director of the Center for Misfolding Diseases. “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, and then we can rescreen and find more potent molecules. .”

With this method, researchers developed compounds to target pockets on the surfaces of aggregated proteins, which are responsible for their dissemination in the brain. According to research, These compounds are hundreds of times more potent and cheaper to develop than previously developed compounds.

“Machine learning is having a real impact on the drug discovery process – it is speeding up the entire candidate identification process [os compostos com potencial para o tratamento de Parkinson] most promising,” said Vendruscolo.

“For us, this means we can start working on multiple drug discovery programs – rather than just one. So much is possible due to the huge reduction in time and cost – it’s an exciting time”, he concludes.