These projects are open-source research systems and infrastructure for cryo-EM particle processing, representation learning, automated reconstruction, and heterogeneous structure analysis.
Designed and released the Cryo-IEF ecosystem, pretrained on approximately 65 million cryo-EM particle images. The project supports representation learning and downstream tools for CryoRanker and CryoClustering, enabling structural classification, pose clustering, particle-quality assessment, and automated reconstruction workflows.
Developed a prior-guided heterogeneous reconstruction framework that uses Cryo-IEF representations to reduce random ab initio initialization and disentangle compositional classification from 3D reconstruction.
Created the open-source data-processing layer used by Cryo-IEF, CryoDECO, and CryoWizard. The package converts CryoSPARC particle jobs into PyTorch-ready datasets and supports MRC/MRCS preprocessing, LMDB-backed datasets, Fourier/Hartley feature generation, balanced sampling, data loading, and CryoSPARC-to-RELION metadata conversion.
Built and extended an end-to-end computational pipeline integrating CryoRanker with CryoSPARC. CryoWizard streamlines processing from raw movies, micrographs, or particles to high-resolution 3D volumes through command-line, web, and browser-extension interfaces.
For additional public repositories and contributions, see github.com/yanyang1998.