Application of machine learning models to identify differences in neural electrophysiological properties across estrous cycle phases
Using whole-cell patch clamped cell recordings from female rat medium spiny neurons across four assessed estrous phases and three recording techniques (action potential bursts, passive membrane recordings, and miniature excitatory postsynaptic currents), seven machine learning models (MLMs) were trained to classify estrous phase based on biologically relevant features extracted from each recording type. The differences in model performance across the feature sets serves as a proxy for the influence of hormones on the specific aspect of neuronal function captured by the given recording type.
See the included requirements.txt file
All data used in this study can be found here: https://datadryad.org/dataset/doi:10.5061/dryad.k0p2ngfgm
We thank Dr. Stephanie Proano for generation of the parent dataset, Eleanor Gayle Brademan, Emily Gaeta, and Sophia Burgess for reviewing earlier drafts and figures, and the NC State University Center for Human Health and the Environment Behavior and Neuroscience group for critical discussion.
This project is licensed under the MIT License, see LICENSE for details.