optimization, inverse problems, also proteins. ml at escalante. formerly: atomicai, xgenomes, broad, berkeley.
Nick Boyd
Loading...
We present sampleworks, an open, modular platform for generating and evaluating biomolecular conformational ensembles from structure predictors guided by experimental data doi.org/10.82153/jkx....
also, nice they cited us! but they missed the part where we say hallucination is currently only better for minibinders (that comparison is notably missing from this paper)
The DiffUSE Project
Nick Boyd
Another banger
new protein design model from Bowen Jing: seems like boltzgen without the design residue AA encoding scheme (so designed residues are backbone only)
this bit from the ESMFold2 paper stuck out to me: at the high end that's $30k (at market rate) for an scFv design run. that's more expensive than using a llama!
mildly sketchy to compare your model's performance against other models using ... your model's scoring function. probably doesn't change the results much though
the ability to google and run RFDiffusion or whatever can't (and shouldn't) be gated. pure marketing
added JAX translations of @biohub.org's new ESMFold2 and ESMC models to our protein design library; including some experimental scripts for VHH and minibinder design following their paper. great to see more amazing, open-source models!
parents of young children should have access to tax-advantaged 529(b) accounts for the purchase of fresh berries
obviously these aren't comparable. also, not a knock on the paper, which is fantastic. and these models are very nice and open.
Kresten Lindorff-Larsen
Accurate protein stability prediction for small domains using mega-scale experiments https://www.biorxiv.org/content/10.64898/2026.05.19.726285v1
