Hey Friends,
Iām teaching a special AI workshop this weekend in Ibiza for my Ladies Who Claude hosted at Olivia Steeleās Studio.
Itās a 3 day workshop where we get AI working for you. We connect Claude to the tools you use everyday (gmail, calendar, Slack, Notion, etc etc), and you start to build your own automations using Claude Code. No technical experience required!
Itās last minute but better way to spend a weekend than learning how AI can change the way you work, build and live?
Hey Friends,
Last week I wrote about the $8.64B that has been poured into AI Drug Discovery companies, with most of them focusing on the Discovery stage of Drug Development. Read it here.
I covered why investors are drawn to companies solving problems at this stage, as it seems the most solvable by computational chemistry and biochemistry modeling. Although there is a lot more complexity to solve at this stage.
AI companies are doing a decent job here, bringing this stage which typically takes 3-6 years to compress it to 18 months, as evidenced by Insilico Medicineās drug to treat idiopathic pulmonary fibrosis (IPF) called Rentosertib.
Rentosertib which was designed entirely with generative AI and has moved through successful Phase IIa clinical trials, making their process 2-4x faster than the industry standard.
Even for Insilico Medicine, which is moving very quickly with Rentosertib, the process took 4-5 years for them to get from Phase IĀ to completion of the Phase IIa trial.
This is a great success story, and one many points to when they say āAI Drug Discovery works!ā
However, a BCG/Wellcome analysis (Jayatunga et al., Drug Discovery Today, 2024) found that while AI-discovered molecules having an unusually high Phase I success of around 80ā90%, Ā they have only a 40% success rate in Phase II trials, which is the same as the historical average for traditional drug discovery.
I got a few comments on last weekās article about this point (thanks for the comments!).
Some people suggested that the clinical trial issue is a consequence of not getting the biology right in the first place, and I agree.
What makes drugs fail in clinical trials?
Drug failure is mostly a biology problem, not a chemistry problem.
And, weāre not getting the biology right because we do not know all of the biology.
We are still making discoveries all of the time. Finally identifying the functions of proteins and enzymes in the human body that we didnāt fully characterize or understand.
And while AI can find molecules that fit into their targets like a lock and a key, it can also mean that the key can fit into other locks as well. Resulting in what is called off-target effects, which can lead to the side effects.
Until we have a complete model of all of the biological systems at the cellular level, tissue level, organ level and whole body level, we canāt offload this entire process to AI with a 100% success rate.
This means there are still many opportunities for companies to improve in target validation, and to make improvements to preclinical drug testing. And many companies are focusing on these areas.
For example companies like Emulate, MIMETAS, and Hesperos are aiming to replace animal studies by creating āorgan-on-a-chipā whey they can work with human tissue to test the drugs as the intermediate step.
But beyond getting the biology right, and preclinical testing we still need to test the efficacy of the drugs in humans in order for the drug to be approved, and that process is painfully slow, insanely expensive and takes years.
The reasons drugs fail in clinical trials deserves its own article. But today I want to focus on other areas in clinical trials which slow down the whole process of getting a drug to market.
What can speed up clinical trials?
The speed of the drug development process limits who gets access to the drugs when they may really benefit from them, and one of the areas that can really impact how fast trials are done is bureaucracy.