AI Is Changing Drug Discovery Faster Than Anyone Expected
Drug discovery has historically been brutally slow and expensive. Getting from target identification to approved drug takes 10-15 years and over a billion dollars, with roughly 90% of candidates failing somewhere in the process. AI is beginning to change that equation in ways that could matter enormously for human health.
AlphaFold changed everything
In 2020, DeepMind's AlphaFold solved a problem that had stumped biologists for 50 years: predicting the 3D structure of proteins from their amino acid sequences. Within two years, it had predicted the structures of nearly every known protein. The practical impact was immediate: researchers could now understand drug targets at a structural level that previously required years of expensive laboratory work.
AlphaFold was the proof of concept. What followed was an acceleration of AI application across the entire drug development pipeline. Generative AI models can now propose novel molecular structures optimized for specific binding targets. Machine learning models predict which candidates will fail in clinical trials based on patterns invisible to human researchers. The process that took a decade is beginning to compress.
Early results are cautiously promising
Insilico Medicine became the first company to advance an AI-designed drug candidate into Phase II clinical trials, for idiopathic pulmonary fibrosis. The company claims the target identification and molecule design that would have taken four to five years took 18 months. Recursion Pharmaceuticals, Exscientia, and Schrodinger have drug candidates in various trial stages.
It's worth calibrating expectations carefully. Phase II success doesn't mean Phase III success, and clinical trial failure rates remain high even for AI-designed molecules. The current AI contribution is primarily to the early-stage discovery and optimization process. The later stages of clinical development remain stubbornly human-paced.
The systemic opportunity
The most significant potential isn't in any single drug. It's in what AI makes economically viable to pursue. Traditional drug development economics forced companies to focus on large patient populations where the math works. Rare diseases, personalized therapies, and treatments for neglected tropical diseases were systematically underfunded because the development cost couldn't be recovered.
If AI genuinely compresses discovery timelines and costs, the economic case for developing drugs for smaller populations becomes viable. That's a structural change in what the pharmaceutical industry can pursue. The patients who stand to benefit most from AI in drug discovery aren't the ones with common diseases. They're the ones with conditions that nobody currently has the economics to solve.