Welcome to MetaWorm, an integrative data-driven model of C. elegans, linking brain, body and environment to faithfully replicate C. elegans locomotion behavior.
For a comprehensive introduction to MetaWorm, please refer to our paper available at bioRxiv.
eworm: The neural network model
MetaWorm: The body & environment model
neuroXcore: 3D visualization for the neural network model
The behavior of an organism is profoundly influenced by the complex interplay between its brain, body, and environment. Existing data-driven models focusing on either the brain or the body-environment separately. A model that integrates these two components is yet to be developed. Here, we present MetaWorm, an integrative data-driven model of a widely studied organism, C. elegans. This model consists of two sub-models: the brain model and the body & environment model. The brain model was built by multi-compartment models with realistic morphology, connectome, and neural population dynamics based on experimental data. Simultaneously, the body & environment model employed a lifelike body and a 3D physical environment, facilitating easy behavior quantification. Through the closed-loop interaction between two sub-models, MetaWorm faithfully reproduced the realistic zigzag movement towards attractors observed in C. elegans. Notably, MetaWorm is the first model to achieve seamless integration of detailed brain, body, and environment simulations, enabling unprecedented insights into the intricate relationships between neural structures, neural activities, and behaviors. Leveraging this model, we investigated the impact of neural system structure on both neural activities and behaviors. Consequently, MetaWorm can enhance our understanding of how the brain controls the body to interact with its surrounding environment.
To ensure optimal performance and compatibility, we recommend installing and running MetaWorm on Ubuntu. Below are the tested specifications:
- OS: Ubuntu 20.04
- GPU: Nvidia 3090
- CUDA: 11.4
- Python 3.8.x
- Nvidia Optix: 7.0.0
sudo apt install git
git clone https://github.com/Jessie940611/MetaWorm.git
sudo apt update
sudo apt install cmake libpython3-dev python3-pip libeigen3-dev libgl-dev libxrandr-dev libxinerama-dev libxcursor-dev libxi-dev freeglut3-dev libglew-dev
get Boost
wget https://boostorg.jfrog.io/artifactory/main/release/1.79.0/source/boost_1_79_0.tar.gz
build and install boost to path you want
tar -xzf boost_1_79_0.tar.gz
cd boost_1_79_0
./bootstrap.sh --with-python=python3
./b2 --with-python --with-filesystem --with-system --prefix="YOUR_BOOST_PATH" install
cd PROJECT_ROOT_DIR
mkdir build
cd build
cmake ../neuronXcore -G"Unix Makefiles" -DCUDA_TOOLKIT_ROOT_DIR="path to cuda root" -DOptiX_INCLUDE="path to optix library/include" -DBoost_DIR="path to BoostConfig.cmake"
make -j8
The folder build2 is a reference for your build folder.
cd MetaWorm/eworm/ghost_in_mesh_sim/
# change line 3 in worm_in_env.py to add the `build` dir in your PC
sys.path.append('[directory of build]')
pip install -r requirements.txt
cd MetaWorm/eworm/components/mechanism
nrnivmodl modfile
If you get "Command 'nrnivmodl' not found", find it in /home/[username]/.local/bin
Experience the simulation of C. elegans movement with our demo.
Execute the open-loop simulation of C. elegans movement.
cd MetaWorm/build
./neuronXcore -data ../eworm/ghost_in_mesh_sim/data/tuned/video_offline/video_offline_neuronX
It you get "No module named xxx", try
export PYTHONPATH=~/MetaWorm/
Press space to play or pause the simulation.
Adjust the parameters of the neural network model by modifying the respective files.
├── components
│ ├── mechanism # ion channel models
│ ├── model # cell models, including morphology and locations of cells
│ ├── param
│ │ ├── cell # biophysical parameters of cells
│ │ ├── connection # adjacency matrix
├── network
│ ├── config.json # network config
Run the script to generate a neural network model
python test.py
The eworm_learn file contains code to training the neural network model to fit the target data. The target data can be Preason Correlation Matrix of neurons' membrane potentials, or the Calcium signals of neurons. This training algorithm requires GPUs to run and supports multiple GPUs.
(If you add or change an ion channel model X, you need to make the correspoding X_lr.mod file in /MetaWorm/eworm_learn/components/mechanism/modfile/ for training)
Create an NMODL mechanism library
cd /MetaWorm/eworm_learn
nrnivmodl components/mechanism/modfile
Train the model
./x86_64/special run_eworm_v4.py
Before training, configure the parameters in run_eworm_v4.py
TARGET_MODE: 'corr' - fit the correlation matrix of neurons' membrane potentials; 'traces' - fit neurons' membrane potentials
PERCISE: True - more precise, but costs more memory and time; False - less precise, but save memory and time
ADAM: True - used Adam as optimizer，False - use SGD
ngpu: number of GPUs
K_nblock: Transfer impedance matrix division block number; larger values save more memory, with a maximum equal to the number of neurons.
cd MetaWorm/eworm/ghost_in_mesh_sim
# set parameters in 02_train_cnn.py
python 02_train_cnn.py
To run the simulation in GUI and render the neural network model, you need to export the morphological data of your neural network model.
cd MetaWorm/eworm/ghost_in_mesh_sim
# Set parameters in 03_make_morph.py
python 03_make_morph.py
Set parameters in worm_in_env.py, like group_name and parameters in config.
If you want to do perturbation experiment, replace the corresponding function in line 58 of worm_neural_network.py.
Run the simulation of C. elegans movement.
./neuronXcore -data [directory of your neuron morphology data]
rotate with right button
zoom with middle button
rotate with left button
strafe with middle button
zoom with right button
space : Play & Pause.
'r': reset worm.
't': tetrahedron mesh of worm body(FEM).
'w': worm body.
'm': worm muscles(96 muscles).
'p': path of swimming trajectory.
'a': arrow of worm swimming direction (yellow arrow)
'x': coordinate of world. x-red arrow. y-green arrow. z-blue arrow.
'c': coordinate of worm body & head. x-red arrow. y-green arrow. z-blue arrow.
'q': exit
'v': switch main view to neuron and vice versa
'1': take a screenshot
'2': hide or show subwiew
usage: neuronXcore -param PARAM
-data DIR DIR is the folder of neuron mophology data, must contain a .swc file
-ss DIR DIR is the folder you want screenshots to save at, default: ./screenshots
--worm-auto-screenshot automatically taking screenshots when playing
-win W H adjust start window size by W * H (int)
-spp SPP adjust samples per pixels in neuron rendering (int)
This project is covered under the Apache 2.0 License.
BAAI