Design and Implementation

This page explains the background of BMTK and BioNet, as well as the design philosophy and implementation details of bionetlite.

About BMTK

BMTK (Brain Modeling Toolkit) is a modeling and simulation framework for the neuroscience community.

Modular Design

BMTK is intentionally divided into the following three modules:

../_images/767339839-bmtk_flow.png

  1. The Network Builder: Network construction

  2. The Simulation Engines: Simulation execution

  3. Analysis & Visualization: Analysis and visualization of results

This separation makes it easier to run large-scale simulations.

Data Standardization

BMTK supports data standardization formats in the neuroscience community:

About BioNet

BioNet is one of the modules of BMTK and handles biophysical neuron networks.

BioNet Builder

This module handles only the network construction part.

  • Users write code to add neurons and synapses to the network

  • When executed, network information is output in formats such as H5 files or CSV files

  • The bionet simulator executes using these files

For large-scale networks, even network construction alone has high computational cost, and operating independently from the simulator part is useful from an error handling perspective.

Usage example:

from bmtk.builder.networks import NetworkBuilder

# Create network
net = NetworkBuilder('mcortex')

# Add nodes
net.add_nodes(...)

# Add edges
net.add_edges(...)

# Build and save
net.build()
net.save_nodes(output_dir='network')
net.save_edges(output_dir='network')

BioNet Simulator

Reads files generated by bionet builder and runs simulations using the NEURON simulator.

Design of bionetlite

bionetlite is designed as an extension of bionet. It generates files for the Neulite kernel with minimal changes to existing code.

NetworkBuilder Override

bionetlite extends the NetworkBuilder from bmtk.builder.networks:

# bionet case
from bmtk.builder.networks import NetworkBuilder

# bionetlite case
from bionetlite import NeuliteBuilder as NetworkBuilder

This simple import change realizes the following:

  • All bionet features remain available as is

  • NeuliteBuilder-specific processing is added

  • Minimal user code changes

Processing Flow

bionetlite processing flow:

../_images/767369966-bionetlite_flow.png

  1. NeuliteBuilder initialization - Sets up preprocessing for morphology files and ion channel configuration

  2. add_nodes() / add_edges() - Executes bionet processing as is and records only biophysical nodes

  3. save_nodes() - In addition to bionet processing, generates <network_name>_population.csv

  4. save_edges() - In addition to bionet processing, determines synapse connection positions and generates <src>_<trg>_connection.csv

Parallel Execution Support

bionetlite supports MPI parallel execution:

# Parallel execution with 4 processes
mpirun -n 4 python build_network.py

Processing during parallel execution:

  • Each MPI process independently executes network construction

  • File output is properly synchronized

  • Significantly reduces build time for large-scale networks

Next Steps