It's like saying that everything in the Mona Lisa depends on the paint.

That's my favorite quote from Derek Lowe's Science opinion, "The End of Disease". And Derek was kind enough to collect links to all his AI-based drug discovery opinions from the past few years, a treasure trove of biting criticism -no need to agree with all, but you find yourself wanting to address his questions.

The latest installment is a response to Demis Hassabis's statement during his 60 MINUTES interview:

"Scott Pelley: The end of disease?

Demis Hassabis: I think that's within reach. Maybe within the next decade or so, I don't see why not."

😱

Skip Derek's section on the fundamental reasons for his skepticism, they are weak and highly questionable (it's the paragraph that begins with "First Off"). But from then on, his line of reasoning is unimpeachable and it all boils down to the infinite complexity and richness of biology 🧬 , the layers upon layers of levels of description, myriad upon myriad interactions between components .

The way that drugs are developed today and in the coming decade for sure relies on throwing things at a putative target and see if it works in humans, without necessarily knowing the mechanism of action: if you have a large RCT with positive results, it doesn't matter if you don't know how it works. It is an example of Competence without Comprehension, a concept that originated from Darwin and revived by the later Daniel Dennett.

The complexity and diversity of biology prevents a full "understanding" (a loaded word I could spend my days refining) of what happens. Digital twins are more like rough drafts of drafts of drafts of [...] of drafts of the real thing. It makes sense: in modeling, the best models are not the most detailed ones but the ones with the right amount of detail to address a specific question. But in biology, the complexity of the system may well require a level of detail beyond reach.

The idea that knowing "how proteins fold" (a model that is itself a simplification of conformational dynamics) can solve everything because proteins are the basis of everything is what provoked the Mona Lisa quote: "[...] no, "everything in your body depends on proteins" (or any one class of molecules!) is such a reductionist take that it really doesn't get you anywhere useful. It's like saying that everything in the Mona Lisa depends on the paint. We have systems built on top of systems that build up other systems past them, and our knowledge of such things is completely inadequate to cure disease within ten years."

That doesn't mean we shouldn't try to improve and accelerate drug discovery and development with AI and any other approach. It's just that the end of disease is not within reach, not in a decade. I wonder what David Baker at the Institute for Protein Design, University of Washington and a co-laureate of the Nobel Prize with Demis thinks about this.

Andrii Buvailo, Ph.D., Alex Zhavoronkov, Kristen Fortney