Skip to content

Prerequisites

This section outlines the computational and data-processing requirements needed to run prismPYP.

⚙️ Computational Prerequisites

Because prismPYP depends on CUDA and PyTorch-GPU, all installation, training, and embedding-generation steps must be performed on a machine with a compatible GPU.

Note

This setup has been tested with PyTorch 2.4.0 and CUDA 12.1 on an NVIDIA A6000 GPU.

🧪 Data Processing Prerequisites

As input, prismPYP expects .mrc files containing both micrographs (i.e., frame averages) and their corresponding power spectra (i.e., output of CTF estimation).

In addition, traditional metrics such as CTF fit, estimated resolution, relative ice thickness, and mean defocus can be used during interactive visualization to help identify high-quality subsets.

Before running prismPYP, micrographs must be pre-processed through CTF estimation to produce the necessary data.

This pre-processing can be performed in either nextPYP or cryoSPARC:

The following steps from the single-particle workflow must be executed:

  1. Import Raw Data
  2. Pre-Processing

The following steps must be executed:

  1. Import Movies/Micrographs
  2. Patch Motion Estimation (if importing movies)
  3. Patch CTF Estimation (optional)
  4. CTFFIND4