Specifications and Limitations

bionetlite and Neulite currently have minimal implementation and reports targeting Fugaku, but have the following specifications and limitations.

Supported Features

Neuron Models

  • ✅ Biophysical neurons: Morphologically detailed biophysical neurons (Perisomatic model)

  • ❌ Point neurons: Point neurons (LIF, IAF, etc.) - Not supported

  • ❌ Virtual neurons: Virtual neurons (for input) - Not supported

Synapse Models

  • ✅ exp2syn: Double exponential synapse model

  • ❌ Other synapse models - Not supported

However, if using exp2syn, you can use synapse model parameters in the directory specified in config.

Input Methods

  • ✅ Constant current input: Supported by default

  • ❌ Spike input from file: Not supported (requires implementation on Neulite side)

  • ❌ Input via Virtual cell: Not supported

Morphology Models

Currently, bionetlite supports the Perisomatic model.

What is the Perisomatic Model?

Perisomatic model is a type of biophysical neuron model used in Allen Cell Types Database. This model has the following characteristics:

Definition and Structure:

  • Ion channel placement: Major ion channels are placed around the cell body (soma)

  • Detailed morphology: Uses morphological data reconstructed from actual neurons

  • Simplified axon: An artificial simplified axon is given instead of the actual axon morphology

See also

Allen Institute official documentation: Perisomatic Single Neurons

Detailed Limitations

1. Neuron Model Limitations

Possible in bionet but not yet supported in bionetlite

Point neurons and Virtual neurons are not supported

BMTK and bionet can define point neurons without morphology and Virtual neurons that are used as input sources without substance, but bionetlite and Neulite do not support these.

bionetlite supports only the following neuron types:

  • ✅ model_type='biophysical'

  • ❌ model_type='point_neuron'

  • ❌ model_type='virtual'

2. Input Method Limitations

Spike input is not supported

By default, bionetlite only accepts constant current input to neurons.

BMTK can use spike trains saved in formats such as H5 files as simulation inputs, but bionetlite does not support this. This is because Neulite can only define constant current input.

Note

If spike input functionality is needed, implementation on the Neulite side is required.

3. Synapse Model Limitations

Only exp2syn is supported

For synaptic connections, we currently support only the double exponential synapse model (exp2syn).

Synapse models other than exp2syn are not supported, but if using exp2syn, you can use synapse models in the directory specified in config.

4. Synapse Connection Position Limitations

Manual specification of detailed connection positions is not supported

Current bionetlite cannot set detailed synaptic connection positions between cells.

bionet constructs networks by setting synapse connection rules (from specific populations to populations, etc.) during network construction.

However, the detailed connection position (such as distance x in segment b of section a on a neuron) is determined when the bionet.simulator module is executed.

Current bionetlite randomly determines the positions of connections that match the rules set during network construction.

Therefore:

  • ❌ Currently cannot obtain exactly the same synaptic connections as bionet

  • ✅ The output from bionetlite for Neulite generated once includes the detailed position information that would normally be determined at simulation runtime, and can reliably execute simulations on the same network

5. Data Type Limitations

Delay values must be int type only

Warning

BMTK delay values are double type, but Neulite only accepts int type.

If using delay values with decimals, appropriate conversion is needed.

6. Notes on Perisomatic Model

For information on the Perisomatic model used by bionetlite, see the Morphology Models section.

It is important to note that actual axon morphology is replaced with a simplified artificial axon.

Summary of Support Status

Feature

Status

Notes

Biophysical neurons

✅ Supported

Perisomatic model only

Point neurons

❌ Not Supported

Virtual neurons

❌ Not Supported

Constant current input

✅ Supported

Spike train input

❌ Not Supported

Requires Neulite implementation

exp2syn synapses

✅ Supported

Other synapses

❌ Not Supported

Parallel execution

✅ Supported

Allen Cell Types Database models

✅ Supported

SONATA format

✅ Supported

via bionet

Future Development

bionetlite and Neulite currently have minimal implementation and are being reported for Fugaku, but these specifications may change with future development.

Next Steps