An alternative strategy for drug discovery may finally make it possible to target proteins that have long frustrated scientists. Transcription factors, which control gene expression, have earned the label “undruggable” because their irregular, dynamic shapes resist conventional structural modeling techniques.
Talus Bio, a Seattle-based company, has developed a solution to this challenge. Rather than attempting to predict protein shapes, its platform, Ptarmigan-1, examines mass spectrometry data to detect compounds that interact with proteins without requiring a rigid three-dimensional structure. “If you try to fold a transcription factor, you just get a plate of spaghetti,” Pino said. “You can’t do drug discovery on a plate of spaghetti.”
This approach also addresses a second major obstacle: expense. Traditional structural prediction tools, such as Boltz-2, demand costly, repeated computations when screening large compound libraries. Talus estimates that screening one million compounds against a single protein using those methods would cost between $100,000 and $1 million and take months. Ptarmigan-1 eliminates these inefficiencies.
Performance tests demonstrate that Ptarmigan-1 processes each compound in 10 milliseconds on a single Nvidia H100 GPU, while Boltz-2 requires 54 seconds per compound—a difference of over 5,000 times faster. This speed reduction could shrink a two-year screening process to fewer than three hours. In a test against STAT6, a transcription factor, Ptarmigan-1 identified 40 inhibitors from Pfizer patents published after Boltz-2’s training cutoff with an area under the curve (AUC) of 0.94. Boltz-2 and a docking baseline both scored 0.58, close to chance. On well-folded targets, Boltz-2 still ranked actives better, and the authors pitch the two as complementary: Ptarmigan-1 narrows billions of compounds cheaply, and a structural model refines the survivors.
For proteins with stable structures, Boltz-2 remains more precise, but Talus positions Ptarmigan-1 as an initial screening tool. The system rapidly narrows down billions of compounds, allowing structural models to later refine the most promising candidates. The company recently screened 3.4 billion compounds against all 20,431 human proteins in under a day, using roughly 20 GPU-hours.
About half of human proteins lack a stable structure that folding models can use. Talus CEO Alex Federation points out that many of these targets are linked to diseases like cancer, yet drug developers have avoided them due to technical limitations. “We’ve been sitting on these targets we’ve known about for decades, that we know are good targets,” Pino said.
Leave a Reply