Talus Releases Ptarmigan-1 AI Drug Discovery Model
The structure-free model predicts protein interactions that co-folding models cannot reach, and researchers have opened it to other scientists.
Updated on Nov. 11, 2026 in Biotech
Talus Bioscience released Ptarmigan-1, an AI drug discovery model that does not rely on protein structures. Researchers have also opened the model to other scientists through a public portal.
Why it matters
Co-folding models do not work for about 40 percent of the proteome, leaving many disease-relevant targets hard to address with current tools. A structure-free approach could expand the set of proteins that drug developers can screen.
The developers cite a 40 percent failure rate for co-folding models across the proteome, against which Ptarmigan-1 screened over a billion compounds in a single day rather than the few thousand a co-folding model handles in a couple of days. Independent validation of these throughput and accuracy claims has not been reported.
The players
Talus Bioscience
A biotechnology company that developed the Ptarmigan-1 drug discovery model and released it publicly.
The details
Ptarmigan-1 was trained on data from automated proteomic approaches and retains information about individual amino acids, allowing it to predict interactions involving unstructured proteins. In a proof-of-concept study it identified successful drug candidates that a co-folding model could not, and indicated drug target sites on STAT6.
Timeline
The structure-based drug discovery revolution began in the 1990s.
In the first two weeks after public release, users ran about two million experiments on Ptarmigan-1.
The Big Picture
Ptarmigan-1 departs from co-folding models, which rely on protein databank structures that poorly represent many disease-relevant proteins, by bypassing protein folding to screen compounds directly. This continues the shift that began with the structure-based drug discovery revolution of the 1990s toward models that do not depend on solved structures.
For patients, a faster route to candidate drugs could eventually shorten the time before new treatments reach clinical testing, though that depends on lead candidates advancing. Researchers say they will keep finetuning the model to move promising leads toward the clinic.
The takeaway
Tools that skip protein folding could widen the range of disease targets that drug developers can screen. Whether Ptarmigan-1's candidates hold up in clinical testing remains the key question to watch.
Further reading
Explore more coverage in our Biotech section.
Source note: This article includes information reported by The Scientist.
