Putting AI to work: better, faster, more accurate
January 30, 2025
On improving yields across the therapeutic landscape:
Our job is to reduce complexity, reduce cycle times and improve yields, whether those yields are in the drug development pipeline, diagnostics that improve therapeutic response rates, and of course manufacturing yields in drug production. Think about the opportunities that lie ahead to improve those. Clearly artificial intelligence will be key to even more acceleration and improvement of those yields, and the more we enable in silico, the quicker we’ll get to the best possible answer
Rainer Blair, Danaher
On the ultimate proof of AI-driven drug development:
Clearly we have seen the impact of AI on small molecule generation, we do see the first impact of generative AI on large molecules and antibodies, and clearly what we also see is the impact on target identification and validation and repurposing of compounds. All that is making its way into clinical development. But the ultimate proof everyone is still waiting for: where are the first approved drugs? And of course my anxiety is, where are the first drugs which connect all those elements?
Andreas Busch, AbSci
Watch: From science fiction to science facts, to science hype: reality check
On going beyond efficiency gains:
When it comes to molecules, we are entering a world where we can do just about anything we want. Now it’s not a true statement yet, but for the first time, it’s a reasonable approximation of a true statement. So the question is, is that impactful? And the answer lies almost entirely in whether it translates into the clinic–whether you know which question to ask. There’s a lot of biology we don’t understand and a lot of cases where we don’t actually know what question to ask of this magical molecular compiler.
Gevorg Grigorian, Generate:Biomedicines
Watch: Beyond the hype: advancements in protein modeling, digital pathology, and human virtual models
On implementing safety guardrails on new technology:
When you think of these big AI tools, we actually don’t have the capacity to say when the tool is out of its confidence level. These kinds of safety tools that can alert the provider and say, ‘you shouldn’t really be trusting me on this patient,’ the kind of safety mechanisms that exist in any other industry, do not really exist in AI. And it’s particularly troublesome because we’re moving to the next generation of tools where humans cannot validate the predictions.
Regina Barzilay, Massachusetts Institute of Technology
Watch: Lost in translational AI: what is the true impact in human health?
On how large-scale collaborations can speed up science:
The Human Cell Atlas is the Google Maps of health and disease. And that ability to map, to share, to build consortia, will transform science in an accelerated way.
Jose-Carlos Gutierrez-Ramos, Danaher
Watch: All science is data science
These quotes were edited for length and clarity.
