Yeah, pretty much that. But also the complete absence of anything resembling useful training data in vast areas of cell biology. Which is really all you need to know about AI and drug discovery.

I'll provide 🥕

That is the heart of the matter. Even the best examples don't give a "theory" of what is happening and thus don't really further understanding. We have great protein structure prediction, but when you look at small molecule binding, AF doesn't seem to "learn" the free energy landscape -

Um... no, I am not saying that. That entire "thought experiment" that is making the rounds is, to put it politely, not helpful. Drugs are formally tested, but for efficacy, not to figure out a molecular mode of action.

True for many things, btw. For the vast majority of eukaryotic cell biological processes we do not have enough knowledge to build a somewhat meaningful training dataset. No training data = problem for AI, methinks. There are areas in biology were AI will help, but they're limited.

That's not how it works in drug discovery... unless you mean to encourage the AI crowd to race further down the rabbit hole...

Congratulations Volker! I wish you all the best!

Research Associate and Postdoc Opportunities! I am happy to announce that I will establish a new laboratory at Hong Kong University of Science and Technology (HKUST) to start a major research program supported by a multi-year FTRSS grant.

“This guy’s cursing me up. I better reconsider my decision”, said no editor ever…

Now I cannot unsee it.

Perhaps it's best to think of all these think tank panels with former generals or spies as a form of critical infrastructure. If our grandparents can "make a short comment" rather than ask a question, they don't have to sell secrets to foreign intelligence agencies to stay relevant