Loading Curie Briefs...
Getting the latest healthcare news for you
Getting the latest healthcare news for you
Finding where metals latch onto proteins just got a whole lot faster. Researchers at Hokkaido University developed PRIME, a deep-learning tool that predicts metal-binding sites in proteins in just 11 seconds — about 10 times faster than existing approaches. It outperforms current tools across 14 metal types and could help identify drug targets for diseases linked to metal imbalances like anemia and immune dysfunction.
Finding where metals latch onto proteins just got a whole lot faster. Researchers at Hokkaido University have developed PRIME (Probe-based Identification of Metal-binding sites), a deep-learning tool that predicts metal-binding sites in proteins in just 11 seconds — roughly 10 times faster than existing methods. Published in Nature Communications, PRIME works in two stages: first analyzing a protein's amino acid sequence to flag likely metal-contact regions, then placing virtual "probes" at those sites to evaluate the 3D environment and confirm binding locations.
PRIME was tested across 14 different metal ions and outperformed existing tools for both well-studied metals (zinc, copper, iron) and harder-to-predict, loosely interacting metals like sodium and calcium. When screened across 1,000 random protein families, ~14% were predicted to bind metals — and nearly 8% of those had no existing database annotation for binding sites, pointing to a largely unexplored landscape of metalloproteins.
By the numbers:
Why it matters: Metal imbalances underlie real diseases — zinc deficiency impairs immunity, iron deficiency causes anemia. PRIME's speed and accuracy could accelerate drug discovery targeting metal-dependent proteins and enable large-scale mapping of metalloproteomes across entire organisms.
