Closing the data loop in AI-driven drug discovery
MIT Technology Review AI MIT Technology Review Insights
AI is accelerating drug discovery by screening molecular candidates computationally rather than through physical screening, but labs struggle to validate the increased volume of AI-generated compounds because traditional screening techniques produce low-fidelity data. AI models trained on publicly available datasets hit a 'data wall' because they lack access to negative data (failed experiments), suffer from publication bias showing only successes, and face growing concerns about data fabrication that could compromise model training. The future depends on closed-loop autonomous labs with integrated infrastructure that generates high-quality, structured data flowing between computational and physical systems, though no AI-discovered drug has yet received FDA approval.
Why it matters
Drug discovery is a high-cost, high-risk endeavor that is under growing pressure from a market increasingly defined by first-mover advantage. Since the 1950s, the cost of developing new pharmaceuticals has roughly doubled every nine years—a phenomenon known as Eroom’s Law. Today, bringing a new drug to market takes an average of 10-15 years and costs…
