Supervised deep learning for efficient cryo-EM image alignment in drug discovery with cryoPARES
What happened
On November 18, 2026, researchers published details of cryoPARES, a supervised deep learning method for cryo-electron microscopy (cryo-EM) image alignment. The system was developed by teams at Astex Pharmaceuticals, University of Oxford, and IE University. cryoPARES uses prior pose information from earlier refinements to speed up the alignment process and automates the removal of unsuitable particles. The method was tested on seven ligand-bound protein complexes across four different protein targets.
Key details
- cryoPARES was published on November 18, 2026.
- The method was demonstrated on seven ligand-bound complexes involving four protein targets.
- cryoPARES automates particle pruning, removing the need for manual intervention.
- Three fragment-bound cryo-EM datasets were released alongside the research.
- The project involved researchers from Astex Pharmaceuticals, University of Oxford, and IE University.
Why it matters
Drug discovery teams using cryo-EM can now obtain 3D structural information much faster, as cryoPARES enables near real-time reconstructions and reduces manual work. This change allows researchers to get feedback during data collection and accelerates the structural determination of protein-ligand complexes. The automation of particle selection lowers the barrier for high-throughput applications and can reduce the time and labor needed for structure-based drug discovery.
