open-source biomolecular model

An open-source biomolecular model is an AI program that calculates what shape a molecule in the body, like a protein, will fold into and what it might bind to. Because its blueprint is public, anyone can download it, use it, and modify it.

Proteins are molecules made of long chains of amino acids, and the 3D shape this chain folds into determines what the protein actually does. Just as a key has to match the exact shape of a lock to open it, a binding protein designed as a drug has to match the exact surface shape of its target to attach to it. Checking these shapes in a lab one by one takes a lot of time and money, but a biomolecular model lets you test this process in advance through computation. And being open source means it's not a paid tool locked to one company — any university lab or startup can pick it up for free and use it in their experiments.

With a model like this, AI can repeatedly generate candidate proteins, predict their shape, and score how well they fit a target, all on its own. Instead of a human standing by the lab equipment every step of the way, the AI can run through many rounds of design and evaluation by itself.

The article describes how Claude, given only information about a target protein, spent 24 hours on its own repeatedly generating candidate proteins, predicting their folded structure, and scoring them using an open-source biomolecular model. It's worth noting that this model isn't Claude itself — it's a separate computational program that Claude called on and used like a tool.