Skip to content

Latest commit

 

History

History
87 lines (57 loc) · 3.17 KB

README.md

File metadata and controls

87 lines (57 loc) · 3.17 KB
Raw

simplecryoem

Basic implementation of cryo-EM projection operators and reconstruction algorithms in JAX from scratch, developed mainly for the implementation of ideas in Toader, Brubaker & Lederman, Efficient high-resolution refinement in cryo-EM with stochastic gradient descent.

To reproduce the numerical experiments in the article, see the 1_Preconditioned_SGD.ipynb notebook, as well as the rest of the notebooks/preconditioned_sgd directory.

Reconstruction demo

For an introduction to doing reconstruction using simplecryoem, see the Reconstruction_demo notebook.

Structure of the repository

The main functionality:

  • simplecryoem.forwardmodel : Forward model-related functions and classes (rotation, interpolation, projection).
  • simplecryoem.optimization : Algorithms and other utilities for solving the volume reconstruction problem.
  • simplecryoem.sampling : Unorthodox use of sampling as optimization for the nuisance variables (angles, shifts). Fairly experimental.
  • The other modules under simplecryoem contain basic functions to make the package useful (reading/writing cryo-EM files, estimating noise, preprocessing data, CTF evaluation).

The notebooks directory contains useful examples:

  • Reconstruction_demo.ipynb : Quick demo of how volume reconstruction works.
  • basic_functionality : Notebooks illustrating random bits of functionality in the package.
  • comparisons : Notebooks to compare outputs from simplecryoem with other packages/softare.
  • preconditioned_sgd : My experiments on preconditioned SGD for cryo-EM.
  • sanity_checks : Very simple examples of projections and reconstruction that serve as sanity checks. Some are implemented as tests too.

The scripts directory currently only contains one script for ab-initio reconstruction. It has not been used recently, and may require updating.

Installation

  1. Clone the simplecryoem repository
git clone [email protected]:bogdantoader/simplecryoem.git
cd simplecryoem
  1. Create the conda environment and activate it
conda env create -f environment.yml
conda activate jax_minimal
  1. Install pyem

Clone the pyem repository in a separate directory:

git clone https://github.com/asarnow/pyem.git
cd pyem
conda install numpy scipy matplotlib seaborn numba pandas natsort
conda install -c conda-forge pyfftw healpy pathos
pip install --no-dependencies -e .
  1. Install simplecryoem in development mode

Back in the simplecryoem directory, run:

pip install --editable .
  1. Check tests pass.
python3 -m unittest -v tests/test*

Citation

If you found this code useful in academic work, please cite: (arXiv link)

@article{toader2023efficient,
    title = {Efficient high-resolution refinement in cryo-EM with stochastic gradient descent},
    author = {Toader, Bogdan and Brubaker, Marcus A. and Lederman, Roy R.},
    journal = {arXiv:2311.16100v1},
    year = {2023},
}