The ML drug discovery startup trying really, really hard to not cheat (Leash Bio)
owlposting.com · 5,790 words · saved by 3 readers
6k words, 27 minutes reading time
Note: I’ll be Austin until Jan 3rd, and in San Francisco (for JPM) from Jan 3rd-17th, message me on X/email to hang out! Also, thank you to Ian Quigley and Andrew Blevins, the two co-founders of Leash Bio, for answering the many questions that arose while writing this essay. Introduction Some of Leash’s research The BELKA result The Hermes result The train/test split result The ‘Clever Hans’ result Conclusion What I will describe below is a rough first approximation of what it is like to work in the field of machine-learning-assisted small-molecule design. Imagine that you are…
saved by
related reading
- Practical Cheminformatics Index - Practical Cheminformaticspatwalters.github.io
- Five things to keep in mind while reading biology ML papersowlposting.com
- So where are we with deep learning for biochem?ladanuzhna.xyz
- MAI-Thinking-1: Building a Hill-Climbing Machinemicrosoft.ai
- A Conversation With Navvye Anand (Bindwell) | Rowanrowansci.com
- AlphaFold2 @ CASP14: “It feels like one’s child has left home.” << Some Thoughts on a Mysterious Universemoalquraishi.wordpress.com
- Widespread false negatives in DNA-encoded library data: how linker effects impair machine learning-based lead prediction - Chemical Science (RSC Publishing) DOI:10.1039/D5SC00844Apubs.rsc.org
- We Need Better Benchmarks for Machine Learning in Drug Discoverypracticalcheminformatics.blogspot.com
- MAI-Thinking-1: Building a Hill-Climbing Machinemicrosoft.ai
- Protein World Model | Biohubbiohub.ai
- nesso1.pdfvalencelabs.com
- AI After Drug Developmentasteriskmag.substack.com