@marcoderossi
"How Machines Learned to Discover Drugs" by D.Khullar on The New Yorker
TL;DR:
- Antibiotics aren’t very profitable for pharma companies: they are expensive to develop, and customers use them for 1-2 weeks
- Bacteria quickly learn to resist antibiotics. By 2050, 10m people per year will die from antibiotic-resistant infections
- We need new generations of antibiotics for those "evolved" bacteria, which are even less profitable (they affect fewer people)
- AI-based drug testing research can help by inventing molecules that pass Phase I (testing on healthy individuals) but so far haven’t passed Phase II (testing on patients) better than human-made molecules
- AI & medicine perform best when the dataset is richer (than in the Antibiotics space). See the recent success of DeepMind’s AlphaFold in designing proteins, a completely different field
- Long story short: let's moderate the enthusiasm and keep on studying.
https://www.newyorker.com/magazine/2024/09/09/how-machines-learned-to-discover-drugs
0 reply
0 recast
1 reaction