Rebuilding Drug Discovery on AI and Data

Also available on:

Drug development has always run on trial and error, organized around blockbuster drugs meant for the largest possible population. David Harel and Shai Shen-Orr see medicine moving decisively off that model, in a clear direction: more personalized, more precise, and steadily converting conditions from terminal to chronic, chronic to curable, and curable to preventable.

What makes the shift possible is an explosion of data. Where a clinical trial once captured five or ten parameters per patient, a single tube of blood can now yield tens of thousands of measurements describing a person’s biology.

But collecting the data is the easy part. We’re moving from trial-and-error to data-driven decision-making that learns formally from every past success and failure. Shen-Orr describes the goal as a decision-support model that works like a map: measure a patient at each point, and let the model choose the next turn in treatment. The payoff is a model that can learn why a drug works, why it fails, and which patients it actually fits.

Key takeaways:

• Cancer is the proving ground. The economics of high-cost cancer drugs justify deep measurement and modeling, so data-driven precision has moved fastest there, with chronic conditions like inflammatory bowel disease, rheumatoid arthritis, and psoriasis next in line.
• Collecting data is only step one. Pharma is increasingly willing to bank rich biological data before it knows the use, but data alone isn’t the breakthrough. The harder, more valuable work is modeling it into decisions.
• The doctor’s visit becomes quantified. Instead of “trust me, I’m a doctor,” patients will ask the odds a drug helps them, and a data-literate generation of physicians will answer with probabilities tuned to a person’s genetic profile.
• Openness is the real unlock. Shen-Orr expects a steep, S-curve jump in data-driven R&D, noting that even in CytoReason’s first few years pharma grew markedly more willing to incorporate and share data.