Gen‑cryoEM - Generative priors for data‑efficient 3D cryo‑EM reconstruction at large-scale HPC

Short description of the innovation study

Histopixel’s Gen-cryoEM project aims to accelerate protein structure determination by using large-scale generative vision models to reduce cryo-EM data requirements and improve 3D reconstruction quality. Building on preliminary research, we will integrate generative models into an iterative refinement workflow, using learned priors and constrained searches in model latent space. A feasibility study suggests a potential large reduction in required cryo-EM particle images, which could lower microscope time, sample consumption, and costs while enabling faster R&D decisions. 

Organisations involved: 

Main SME: Histopixel

Technology expert: Atem Structural Discovery