This repository contains the code described in the following publication:
O'Neill P.S., Baccino-Calace M., Rupprecht P., Lee, S., Hao, Y.A., Lin, M.Z., Friedrich R.W., Müller, M., and Delvendahl, I.
(2025) A deep learning framework for automated and generalized synaptic event analysis. eLife 13:RP98485 (doi:10.7554/eLife.98485)
miniML is a deep-learning-based tool to detect synaptic events in 1d time-series data. It uses a CNN-LSTM network architecture that was trained using a large dataset of synaptic events (miniature excitatory postsynaptic currents) from cerebellar mossy fiber to granule cell synapses.
In this repository, we provide documentation, pre-trained models, and Python code to run model inference on recorded data. In addition, an application example (cerebellar granule cell mEPSC recording) is included.
- Initial release as package
To use miniML, clone the GitHub Repository and install, e.g. using pip. It is strongly recommended to install miniML into its own virtual environment! miniML requires Python 3.11. The default Python dependencies are: sklearn, matplotlib, h5py, pandas, numpy, scipy, tensorflow, pyabf, ruptures.
Important
The release of TensorFlow 2.16 and Keras 3 introduced breaking changes that raise an error when loading models trained with earlier TensorFlow versions. To avoid this, it is recommended to use TensorFlow 2.14 or 2.15.
Note
miniML can be run on a GPU to speed model inference. Either CUDA or tensorflow-metal are required for GPU use. Installation instructions for these requirements may depend on the specific hardware and OS and can be found online.
For the GUI, additional dependencies are required (PyQt5, qt-material, and pyqtgraph). Install miniML with pip install .[gui] to include these dependencies and get full GUI support. An executable script miniml-gui will be created to start the application.
Detailed documentation for miniML can be found here.
First, a miniML MiniTrace object needs to be created containing 1d timeseries data. Currently, miniML supports direct loading from HEKA .dat files, Axon .abf files as well as HDF .h5 files. The miniml.fileio.TraceLoader object features discrete methods for loading from these file formats (e.g., TraceLoader.from_h5_file()). Data in other file formats need to be imported as Python objects.
Next, a miniML EventDetection object is initiated. Here, one needs to specify a miniML model file to use as well as the Trace object to operate on.
miniML model inference can then be run using the detect_events() method. This method will run miniML over the given data using the specified model. Runtime will depend on data length.
Following event detection, the individual detected events are analyzed and descriptive statistics are calculated for the recording.
miniML includes several plotting methods. They can be found in the Plotter class in plotting.py. A detection object has to be passed as data argument.
Event data and statistics can be saved in different formats (.pickle, .h5, .csv).
Tip
miniML can also be used via a GUI (see Installation with GUI). The GUI allows easy loading of data, pre-processing (filtering, detrending etc.) and model inference. Found events can be inspected and deleted, if desired. The GUI can also be used to save results to a PICKLE, CSV or HDF5 file.
The folder "example_data/" contains an example recording from a cerebellar mossy fiber to granule cell synapse. To use miniML on this data, run the commented example Jupyter Notebook (tutorial) illustrating the use of miniML.
The documentation includes notebooks showing how to train miniML models. You can also use the following links to Kaggle to train mminiML models on the cloud:
1 - Transfer learning
2 - Full training
3 - Training dataset
The repository contains trained models for several event detection scenarios, as outlined in the associated paper. If you have trained a model that could be useful for other researchers, please consider opening a pull request or get in touch with us in order to add it to the repository.
If you use miniML in your work, please cite:
@article{ONeill2025,
article_type = {journal},
title = {A deep learning framework for automated and generalized synaptic event analysis},
author = {O'Neill, Philipp S and Baccino-Calace, Martín and Rupprecht, Peter and Lee, Sungmoo and Hao, Yukun A and Lin, Michael Z and Friedrich, Rainer W and Mueller, Martin and Delvendahl, Igor},
volume = 13,
year = 2025,
month = {mar},
pub_date = {2025-03-05},
pages = {RP98485},
citation = {eLife 2025;13:RP98485},
doi = {10.7554/eLife.98485},
url = {https://doi.org/10.7554/eLife.98485},
}The development of miniML was funded by the Swiss National Science Foundation, the University of Zurich Research Talent Development Fund, and the German Research Foundation.
The GUI uses Material Icons by Google (https://fonts.google.com/icons).
The Python code to read HEKA Patchmaster files was adapted from https://github.com/campagnola/heka_reader
The template matching implementation was adapted from https://github.com/samuroi/SamuROI
Please feel free to contact us in case of questions, either via email, or by opening an Issue here on GitHub.
philipp.oneill@physiologie.uni-freiburg.de or igor.delvendahl@physiologie.uni-freiburg.de