AI could shrink drug discovery from years to months, expert says

Editor's Note

Artificial intelligence (AI) is poised to overhaul drug development, with DeepMind CEO Demis Hassabis predicting that discovery timelines could shrink from more than a decade to just months. In a September 13 article from the Times of India, Hassabis described how AI models can identify drug candidates faster, reduce failure rates, and lower costs, opening the door to earlier access to new treatments and more efficient medical research.

Drug discovery has long relied on slow, trial-and-error processes that can take 10 to 15 years from concept to market, notes the article. Hassabis told Bloomberg Television that AI could compress this cycle dramatically, potentially producing viable candidates within weeks. DeepMind’s subsidiary, Isomorphic Labs, uses AI to model complex biological systems, analyze molecular structures, and predict drug-protein interactions. These tools allow researchers to focus on molecules with the greatest chance of success rather than spending years on compounds likely to fail.

Failure rates remain one of the biggest barriers in drug development, with many compounds faltering during later testing due to lack of efficacy or safety concerns. DeepMind’s AI models reportedly simulate protein folding and chemical interactions to forecast how molecules behave in the body. They can also propose novel molecular structures overlooked by conventional methods, expanding the pipeline of therapeutic options while minimizing costly setbacks.

Hassabis emphasized these advances could reshape public health by enabling faster responses to pandemics and emerging diseases, while also supporting personalized medicine tailored to individual genetics and physiology. Lower development costs could make treatments more affordable globally, broadening access in low-resource regions where cutting-edge therapies are scarce.

Although Hassabis did not cite specific drugs, the article noted AI models are already being applied to conditions such as Alzheimer’s disease, rare cancers, and neurodegenerative disorders. Early findings suggest that computational predictions can generate actionable leads for clinical trials while reducing the experimental workload.

The outlet also detailed hurdles. Regulatory validation will remain essential before AI-generated candidates reach approval. Ethical oversight is needed to ensure equitable, safe outcomes, particularly in personalized medicine. And collaboration among technologists, biologists, and clinicians will be critical to translate predictions into therapies.