University of Cambridge scientists have employed artificial intelligence (AI) to expedite the quest for novel Parkinson’s disease therapies. Through machine learning techniques, they efficiently screened millions of potential drug compounds, identifying promising candidates at a rate ten times faster and at a cost that is 1000 times cheaper than traditional methods.
Parkinson’s disease is a progressive neurodegenerative disorder affecting around 6 million individuals globally, with projections indicating a tripling of cases by 2040. Presently, available treatments do not reliably impede the disease’s progression.
Conventional screening of vast chemical libraries to discover potential drug candidates is slow, costly, and often ineffective. Lead researcher Professor Michele Vendruscolo explains that identifying suitable candidates for further testing can take months or even years.
How They Used AI to Speed Up Parkinson’s Treatment Search
To address this challenge, Vendruscolo and colleagues devised a five-step machine learning approach, detailed in a publication in Nature Chemical Biology:
- Start with a small set of compounds showing the potential to inhibit alpha-synuclein protein clumping, a key feature of Parkinson’s.
- Train a machine learning model using experimental results to predict effective molecular structures and properties.
- Utilize the trained model to swiftly screen millions of compounds and identify top contenders.
- Experimentally validate AI-selected candidates and refine the model based on results.
- Iterate the process, with the AI model becoming progressively more adept at pinpointing potent compounds.
This iterative approach significantly increased the optimization rate from 4% to over 20% across multiple cycles. Moreover, AI-discovered compounds exhibited greater potency and chemical diversity compared to previously identified ones.
Vendruscolo highlights the impact of machine learning on drug discovery, emphasizing the accelerated identification of promising candidates. The newfound ability to discern specific molecular regions responsible for binding allows for the discovery of more potent molecules.
The researchers anticipate continued advancements in AI-driven drug discovery, particularly for diseases involving protein misfolding and aggregation. While challenges remain in translating AI-identified candidates into approved treatments, this study underscores the potential of combining machine learning with experimental biology to expedite drug discovery processes.
This work aligns with other research endeavours leveraging AI in drug discovery, underscoring the transformative potential of AI methodologies in healthcare and medicine.
See also: Baylor University Explores ChatGPT’s Forecasting Power
