Adolescent development of E-I using biophysical network modeling

This repository includes the code associated with the paper "Adolescent maturation of cortical excitation-inhibition ratio based on individualized biophysical network modeling", Saberi et al.

Structure

• scripts: Scripts used for image preprocessing and running biophysical network modeling simulation-optimization jobs. Read the details in ./scripts/.
• tools: Software developed in this project for biophysical network modeling simulation-optimization on GPU and CPU
  • bnm_cuda: Main C++/CUDA used for model simulation-optimization
  • cubnm: A Python toolbox containing a similar C++/CUDA core that was used in some of the analyses
• results: Jupyter notebooks and helper scripts used to run statistical tests and generate the paper main and supplementary figures. Note that these notebooks cannot be run without the input data which is restricted-access and cannot be shared publicly. However, some of the data such as the maps of statistical effects are shared within this folder.

Dependencies

• Data: The scripts require input data of Philadelphia Neurodevelopmental Cohort and the IMAGEN dataset to exist in pre-specified directories defined under $PROJECT_DIR, $PNC_PROJECT_DIR, $IMAGEN_PROJECT_DIR, $INPUT_DIR and $OUTPUT_DIR which are expected to be defined as environment variables. However, data used in this project is restricted-access and cannot be shared publicly. Accordingly we have not shared the scripts used for downloading and handling the input data.
• Python 3 (tested with version 3.10) and packages listed in ./scripts/setup/requirements.txt are needed to run the Python scripts.
• Singularity (tested with version 3.11.4) is needed to build and run the following containerized software:
  • 
  • 
  • : Note that we used an earlier version (0.1.1) which is currently not available on Docker Hub.
• scripts/modeling/run scripts require Nvidia GPUs and a few additional dependencies:
  • scripts/modeling/run/bnm_cuda: See here
  • scripts/modeling/run/cubnm: Additionally requires cubnm to be installed in the Python virtual environment (not installed via ./scripts/setup/requirements.txt). For cubnm requirements see here
• The preprocessing and modeling scripts were run as jobs on our HTC and HPC clusters which use HTCondor and Slurm for job scheduling, respectively.

Support

Feel free to contact Amin Saberi (amnsbr[at]gmail.com) if you have any questions.