Simulation of electron microscopy tilt- and frame-series, focused on modelling whole samples. This is the code that was used to generate the SHREC'21 dataset.
The code is not well optimized, and uses a lot of CPU memory. It does generate fairly accurate simulations as it also models amplitude contrast from inelastic interactions (where it assumes a zero-loss filtered image, corresponding to a modern post-column energy filter). It can also generate deformed vesicles, in contrast to SHREC'21 where it only generated ellipsoidal vesicles (no membrane protein embedding though).
The simulator is heavily based on InSilicoTEM, by Vulovic et al., (2013): https://doi.org/10.1016/j.jsb.2013.05.008 .
Micrograph-modeller is a python(>=3.8) package that can be installed. The base of the code runs on CPU's due to the large memory requirements for the tomography models. For sampling atoms to electrostatic potential a part of the code can be accelerated with GPU's. You will need to install cupy and CUDA toolkit to access it (see below).
To add hydrogens and remove water molecules from pdb/cif files I call chimerax from the command line. So for the full functionality you need to have chimerax installed.
I assume you have miniconda3 installed.
Two options to create the conda environment:
-
(1) this will later build cupy against local cuda toolkit:
conda create --name micrograph_modeller python=3 -
(2) this gets a prebuild cupy and compatible cuda-toolkit:
conda create --name micrograph_modeller python=3 cupy cuda-version=11.8
Then activate the environment, clone the repository and install:
conda activate micrograph_modeller
git clone https://github.com/McHaillet/micrograph-modeller.git
cd micrograph-modeller
python -m pip install .
You can run the unittests as follows.
cd tests
python -m unittest discover
UNDER CONSTRUCTION