STTCPy 1.0.0

Calculate STTC for large datasets of spiking time-series.

STTCPy Documentation

Overview

STTCPy is a Python package designed to efficiently compute the Spike Time Tiling Coefficient (STTC), a metric introduced by Cutts and Eglen et al. to assess synchrony between neuronal spike trains (Cutts and Eglen, 2014). While existing STTC implementations in both Python and MATLAB are limited to one-to-one calculations (i.e., STTC between a single pair of signals), STTCPy vectorizes these computations, enabling much faster processing of large datasets.

This approach significantly accelerates STTC calculations, making it suitable for high-volume data analysis. The package also includes functionality for calculating null distributions for significance testing and offers GPU acceleration using PyTorch backend for even greater performance on large-scale analyses.

Key features:

• Efficient, vectorized STTC computation for large datasets.
• Scalable GPU-based functionality with PyTorch and CUDA support.
• Helper functions for pre-processing, analysis, and significance testing via null calculations.

---

Installation

Requirements:

• Python 3.8 or newer.
• Dependencies: numpy, scipy

Note: To enable GPU functionality, PyTorch must be installed with the appropriate CUDA version for your system. For installation instructions, visit the PyTorch installation guide.

Installation:

Install the package via pip:

pip install STTCPy

Alternatively, clone the repository and install manually:

git clone https://github.com/aditiaravind/STTCPy.git
cd STTCPy
pip install .

---

Usage

Import the Package:

from STTCPy import STTC, null

Using CPU-only version:

Compute STTC:

from STTCPy import STTC, null, generate_spike_train

N = 3 # Number of spike trains
L = 1000 # Length of each spike train
sparsity = 0.2 # Density of spikes within the train

spikes = generate_spike_train(N, L, sparsity)

dt = 2  # Number of frames to pad when calculating STTC

spikes_STTC = STTC(spikes, dt)

You can then plot the STTC values as follows

plt.imshow(spikes_STTC, vmax=1, vmin=-1, cmap='bwr')
plt.colorbar()
plt.xlabel('STTC Value')
plt.ylabel('STTC Value')
plt.title(f'STTC across {N} spike trains')
plt.show()

Null Model for Statistical Testing:

Calculate STTC when the signals are shuffled or shifted 'n' times. Average this to compare against the actualy STTC values to check for significance.

n_shifts = 5

spikes_null = null(spikes, n_shifts, dt)
print(f"Null STTC Shape: {spikes_null.shape}")

# Output : (5, 10, 10)

Using GPU:

Here, you have to have torch installed.

import torch
from STTCPy import STTC_gpu as STTC, null_gpu as null

N = 3 # Number of spike trains
L = 1000 $ Length of each spike train
sparsity = 0.2 # Density of spikes within the train

spikes_numpy = generate_spike_train(N, L, sparsity)
spikes = torch.tensor(spikes_numpy, device='cuda', dtype=torch.float32) #Specify dtype as float for best results.

dt = 2  # Number of frames to pad when calculating STTC
n_shifts = 5

spikes_STTC = STTC(spikes, dt)
null_STTC = null(spikes, n_shifts, dt)

For full documentation visit the Github repository

---

Acknowledgments

I would like to acknowledge the following paper for introducing the Spike Time Tiling Coefficient (STTC) metric:

Cutts CS, Eglen SJ. Detecting pairwise correlations in spike trains: an objective comparison of methods and application to the study of retinal waves. J Neurosci. 2014 Oct 22;34(43):14288-303. https://doi.org/10.1523/jneurosci.2767-14.2014. PMID: 25339742; PMCID: PMC4205553.

---

License

STTCPy is distributed under the GNU General Public License v3 (GPLv3). For more details, refer to the GNU GPLv3 License.

---

Contributing

I welcome contributions! Here’s how you can help:

1. Fork the repository.
2. Create a new branch for your feature or bug fix:

   git checkout -b feature-name
   

3. Commit your changes and push to your branch:

   git commit -m "Description of changes"
   git push origin feature-name
   

4. Submit a pull request with a detailed description of your changes.

For issues or feature requests, please use the GitHub issues page.

---