Artificial intelligence (AI) is opening new frontiers in medical research and treatments, particularly for tough-to-treat diseases. This is the fourth feature in a six-part series exploring how AI is transforming healthcare.
AI-Driven Innovation in Drug Discovery
During a video call, Alex Zhavoronkov displayed a small, green, diamond-shaped pill. His company developed it to address idiopathic pulmonary fibrosis (IPF), a rare and progressive lung disease without a known cure or cause. Early clinical trials of the drug have shown promising results, although it has yet to receive approval.
This drug represents a wave of innovations where AI plays a pivotal role in discovery. “We can’t yet claim to have the first AI-designed molecule approved,” says Dr. Zhavoronkov, co-founder and CEO of Insilico Medicine. “But we are likely the most advanced in this space.”
The “AI drug race” features both specialist AI biotech startups and established pharmaceutical firms. Some, like Google’s parent company Alphabet, have created dedicated AI-focused subsidiaries like Isomorphic Labs. Demis Hassabis, its CEO and recent Nobel laureate, has contributed to AI models poised to transform drug design.
Chris Meier of Boston Consulting Group (BCG) believes AI’s role in drug discovery could significantly benefit patients. Traditional drug development averages 10–15 years and costs over $2 billion. Approximately 90% of drugs fail during clinical trials. AI could reduce costs, save time, and improve success rates.
Charlotte Deane, a professor at Oxford University, describes this as the dawn of an era where AI takes center stage in drug development. AI won’t replace scientists, but it will work alongside them to improve outcomes. Recent analyses have identified at least 75 AI-discovered molecules entering clinical trials, signaling a significant milestone. However, experts like Prof. Deane stress that human input remains integral to the process.
Two Key Applications of AI in Drug Discovery
Dr. Meier highlights two critical areas where AI is making a significant impact. First, it identifies therapeutic targets at the molecular level. AI analyzes large databases to find links between disease mechanisms and biological molecules, accelerating discovery compared to traditional lab methods.
Second, AI aids in designing drugs to act on identified targets. Generative AI, similar to ChatGPT, suggests molecules likely to bind effectively to the target. This method streamlines the process by replacing labor-intensive synthesis and testing of numerous molecular variations.
Insilico Medicine employs AI across these steps and more. Its AI predicts clinical trial success probabilities and integrates findings to refine drug discovery. The company currently has six molecules in trials, with more in development. Its IPF drug exemplifies AI’s potential: generative AI designed the molecule to inhibit a protein called TNIK, identified by another AI as key to the disease. The process took 18 months and involved synthesizing 79 molecules—far fewer than the industry standard of 500 over four years.
Despite advancements, data scarcity presents a significant hurdle for AI in drug discovery. Limited data can lead to biases, affecting both target identification and molecule design. Recursion Pharmaceuticals addresses this issue by generating vast datasets through automated experiments. It uses these to train AI tools, uncovering unexpected molecular relationships. The company’s supercomputer, the fastest owned by a pharmaceutical firm, aids this work. One of its AI-developed molecules, designed to treat lymphoma and solid tumors, is now in early-stage clinical trials.
Recursion’s CEO, Chris Gibson, underscores the ultimate goal: proving AI-discovered molecules can succeed in trials and outperform traditional methods. “When that happens,” says Dr. Gibson, “it will be clear that AI is the future of drug development.”