Reproducible supervised learning-assisted classification of spontaneous synaptic waveforms with Eventer

Area of Science:

• Neuroscience
• Computational Neuroscience
• Data Analysis

Background:

• Manual detection of spontaneous synaptic events is time-consuming and prone to bias.
• Existing methods for synaptic event detection lack consistency and reproducibility.
• Automated tools are needed to streamline the analysis of electrophysiology and imaging data.

Purpose of the Study:

• To develop an open-source application, Eventer, for automated detection and analysis of spontaneous synaptic events.
• To provide a reproducible and consistent method for analyzing large datasets.
• To reduce subjectivity and improve the throughput of synaptic event analysis.

Main Methods:

• Eventer utilizes a machine learning approach, specifically Random Forests, trained on user-defined criteria.
• The application employs Fast Fourier Transform (FFT)-based deconvolution for initial event candidate identification.
• It is a standalone application compatible with Mac, Windows, and Linux, using the MATLAB Runtime.

Main Results:

• Eventer successfully learns user-defined criteria for classifying synaptic events.
• The application demonstrates high consistency in analyzing large datasets compared to manual selection.
• An associated online repository facilitates sharing of machine learning models to enhance reproducibility.

Conclusions:

• Eventer offers a reproducible and efficient solution for analyzing spontaneous synaptic events.
• The open-source nature and model-sharing repository promote collaboration and standardization in neuroscience research.
• This tool addresses critical needs for increased throughput and reproducibility in synaptic transmission studies.