Odor space analysis, synthetic ORN dataset, and odor classification

This repository contains scripts and data to reproduce Figures 4,5,6 in Synthetic Biology Meets Neuromorphic Computing: Towards a Bio-Inspired Olfactory Perception System by Max, Sames, Ye, Steinkühler, and Corradi (2025).

Electrophysiological recordings in 20170717_cell4.csv from Cryo-EM structure of the insect olfactory receptor Orco, Butterwick et al. 2018, provided by Josefina del Mármol.

Installation

Clone all contents of this folder, cd into it, make new python environment and activate:

git clone https://github.com/kma-code/SYNCH_perspective.git
cd SYNCH_perspective
python3 -m venv SYNCHenv
source SYNCHenv/bin/activate
pip3 install -r requirements.txt
python -m ipykernel install --user --name=SYNCHenv 

What this repository does

This repo generates a synthetic (fake!) dataset of voltage traces using data extracted from Cryo-EM structure of the insect olfactory receptor Orco by Butterwick et al., 2018.

To understand the steps of generating the dataset, look at the jupyter notebook.

The actual dataloaders are generated using the following scripts:

Defining a parameter file

You need to define a parameter file containing the following items. A template can be found in saved_datasets/dataset_template.json (make sure that you activate SYNCHenv).

• odorant_names: list of odorants. To see all available odors, run python odor_space_analysis.py --list. Include the none odor for better results, see paper.
• N_OR: number of receptors to be reconstituted
• N_ORCOs: how many Orcos to simulate per ORN
• output: voltage or mean, i.e. whether a trace should be generated or only the mean voltage
• output_dt: e.g. 0.005, how big one time step (in seconds) in the generated traces should be
• total_steps: how many steps to simulate per trace
• data_steps: how many of total_steps should be
• dataset_size: how many traces to generate per odorant

Odor space analysis

To perform the analysis, run python odor_space_analysis.py --params 'saved_datasets/dataset_template.json', using your parameter file. The selected OR types will be saved back into the parameter file.

To see some examples, see the jupyter notebook (make sure that you activate SYNCHenv).

Generating a dataset

To generate a dataset, run python generate_dataset.py --params 'saved_datasets/dataset_template.json' after running odor_space_analysis.py. Simulation parameters will be saved back into the parameter file.

The dataset will be saved in saved_datasets as .pkl files. To import, use:

import dill
import json
# load dataset
with open('saved_datasets/dataset_size1_Nodor3_NOR7_voltage.pkl', 'rb') as f:
    dataset = dill.load(f)
# load corresponding dict
with open('saved_datasets/dataset_template.json') as f:
    dataset_dict = json.load(f)

Running the classifier

To run the classification, check out the jupyter notebook (make sure that you activate SYNCHenv).