Welcome to DeePhys

The package for Deep electrophysiological phenotype characterization:

Created with BioRender

Full documentation

Overview

DeePhys was created to facilitate the analysis of extracellular recordings of neuronal cultures using high-density microelectrode arrays (HD-MEAs). DeePhys allows users to easily:

• Extract electrophysiological features from spikesorted HD-MEA recordings
• Visualize differential developmental trajectories
• Apply machine learning algorithms to classify different conditions
• Obtain biomarkers predictive of the respective condition
• Evaluate the effect of treatments
• Dissect heterogeneous cell populations/cultures on the single-cell level

Requirements

Currently DeePhys is only available on MATLAB, so a recent MATLAB installation (>2019b) is required.

Installation

The package is ready-to-use right after cloning. Run startup.m from the repo root to add all paths.

Usage

Code requires spikesorted data in the phy format. For help with spikesorting check out the SpikeInterface package.

We provide seven tutorials covering the full analysis workflow:

1. Data Processing — load Kilosort output, run QC, extract features
2. Feature Exploration — inspect and prepare feature matrices
3. Phenotype Analysis — classification, dimensionality reduction, regression
4. Cell Type Classification — excitatory/inhibitory labelling
5. E/I Analysis — network burst detection and E/I quantification
6. Transfer Learning — external data interoperability
7. Legacy Migration — migrate old MEArecording objects to the new API

A dataset from our most recent paper will accompany the tutorials (link to be added upon publication). The dataset of the original paper is still available.

Citation

If you find this package helpful or used in your analyses, please cite the DeePhys paper and link to this GitHub repository.

Dependencies

This package uses several packages/toolboxes, all of which are bundled in the repository (under Functions/ and Toolboxes/) and added to the MATLAB path automatically by startup.m. No separate installation is required.

• the readNPY function provided by the npy-matlab package
• the CCG function provided by the FMAToolbox
• the othercolor function
• the catch22 toolbox as published here
• the ISIN burst detection algorithm as published here
• the Brain Connectivity Toolbox
• the UMAP for MATLAB toolbox

Issues?

If you face any problems or bugs, or have ideas for additions to this package please open an issue.