1 Comment

Fascinating! I’m curious how you think about the path dependence in this process. Surely the first 1000 sequences were a tiny sampling of the fitness landscape. Yet, those results probably helped design the next library (2100 variants) and so forth. Had you started with a different set of 1000 sequences in Round 0, how close would you have been to solution you ended up with?

Thanks for sharing these!

Expand full comment