Simon Haworth: Where the Hype Runs Out in Drug Discovery

<aside> 💬

"Focus on the stuff below the water, the dark biology that we don't know."

</aside>

Thursday October 29, 2026 · 1:45 PM · Talk

Key ideas from the talk

Of 100 drug projects entering phase one, about 42 to 48 reach phase two. About 70 percent fail in phase two, which gives a 92 percent overall failure rate.
He puts the cost at 288 billion US dollars a year.
AI has halved phase one failure. It has not improved the later stages, because the underlying biology is poorly known.
He estimates people know about one percent of essential human biology. He urges AI work on target identification in the unknown part.
Drug development needs wet labs, hospitals and physical kit. He sees wet lab validation as an opportunity for BC.
Digital twins may predict failures in advance. He thinks only about 20 of 100 clinical trials would need to start.
He described a 20:60:20 split for drug discovery. People set the question at the start, AI does the middle 60 percent, and people judge the result at the end.

"The pharmaceutical sector is our patient."
"The net performance of this wonderful industry is a 92 percent failure rate."
"So basically, AI so far has made pharma fail faster. That's literally it."
"This is the whole of human biology, and that little sliver above the waterline is what we know already."
"What proportion of essential human biology do we know? And the estimates, it's about one percent."
"In my space, if you get it wrong, it's curtains."
"I get a call from Penn State saying we're doing a clinical trial GLP-1 on people who just committed to stop fentanyl, and it is improving their journey."