Highlights and key takeaways from Day 2 of DeSci Berlin 2026๐
โ
@vitarnabio targets a mutation tied to both cancer risk and accelerated aging, one drug, two disease mechanisms. Early mouse data already shows a 4x increase in tissue uptake across the heart, brain, and liver.
โ Most AI in drug discovery still needs a human to check the data, submit the order, and read the result.
@peptai_ removes all three, designing candidates, paying a wet lab directly through x402, and reading results back with no human in the loop. Every candidate gets minted onchain as IP through
@Molecule_sci, so the trail survives even if the experiment fails.
โ Open Labs lets a researcher post a single hypothesis and get real feedback from a focused community before any token or funding exists. Stake tokens behind a project, and the yield flows directly to fund the research while the principal stays untouched.
โ On the panel about closing the loop between AI design and the wet lab, one constraint kept coming up. Biology lacks perturbation data, real records of what happens when a specific sequence actually gets changed. Agents can propose endless candidates, but proving any of them still runs on lab time.
โ
@adaptyvbio turned wet lab validation into an API call, send a protein sequence, get real binding data back in days. To test it, they pitted AI agents against human protein designers on the same target at a hackathon. The agents matched the best humans in the room, and the top two performers ran on GPT and Grok, no specialized biology model required.
Full talks here โ