📄️ 1. Denoise
Denoising can vastly improve particle picking by helping to increase signal to noise in low-dose micrographs. Different denoising algorithms exist, include deep denoisers Topaz (Bepler et al., 2020) and Janni (Wagner and Raunser., 2020) and various Gaussian and fourier space denoisers. PartiNet implements a modified heuristic Wiener filter denoiser based on the method from CryoSegNet (Gyawali et al., 2024). PartiNet's implementation introduces multiprocessing, allowing for high-throughput denoising of large datasets, as well as saving in .mrc format if you prefer to perform picking on denoised micrographs in RELION or CryoSPARC.
📄️ 2. Detect
PartiNet uses a modified version of an adaptive YOLO architecture called DynamicDet (Lin et al., 2023) to identify particles in cryo-EM micrographs. This stage provides highly accurate particle detection with customizable confidence and overlap thresholds.
📄️ 3. STAR
The partinet star command is the final step in your particle picking pipeline. It converts the particle coordinates detected in the previous stage into a standardized STAR file format that can be used with downstream cryo-EM processing software like RELION, cryoSPARC, or other reconstruction programs.