Machine Learning for Separating Dopamine and Octopamine Electrochemical Signals in Drosophila

Area of Science:

• Neuroscience
• Computational Neuroscience
• Analytical Chemistry

Background:

• Dopamine and octopamine are crucial neurotransmitters for learning in *Drosophila melanogaster*.
• Studying their individual roles is difficult due to colocalization and overlapping electrochemical signals.
• Existing signal separation methods fail with time-varying voltammograms, especially for octopamine.

Purpose of the Study:

• To develop a novel method for simultaneous detection of dopamine and octopamine.
• To overcome limitations of traditional signal separation techniques in electrochemical recordings.
• To enable accurate quantification of individual neurotransmitter contributions in complex mixtures.

Main Methods:

• A deep learning-based regression approach using a modified U-Net architecture was employed.
• The model analyzed color plots of electrochemical data to separate dopamine and octopamine signals.
• It specifically used octopamine's primary oxidation peak to predict and subtract its secondary oxidation peak.

Main Results:

• The deep learning method achieved high accuracy, with normalized root-mean-square errors of 0.06 for dopamine and 0.08 for octopamine.
• Estimation errors were consistently below 10%, demonstrating reliable signal separation.
• Experimental data showed strong agreement between predicted and actual concentrations (r = 0.93, CCC = 0.93).

Conclusions:

• Machine learning, specifically deep learning, provides a robust framework for deconvoluting overlapping electrochemical signals.
• This technique facilitates simultaneous neurochemical detection of dopamine and octopamine in *Drosophila*.
• The developed method enhances the ability to study the distinct contributions of these neurotransmitters to behavior.