Probability estimation of narcolepsy type 1 in DTA mice using unlabeled EEG and EMG data

Area of Science:

• Neuroscience
• Computational Biology
• Animal Models

Background:

• Manual analysis of mouse sleep studies is time-consuming.
• Automatic sleep stage classification exists, but phenotype detection is lacking.
• Narcolepsy Type 1 (NT1) mouse models require efficient assessment methods.

Purpose of the Study:

• To develop a fully automated pipeline for estimating NT1 probability in hypocretin-tTA;TetO-Diphteria toxin A (DTA) mice.
• To utilize unlabeled electroencephalographic (EEG) and electromyographic (EMG) data for phenotype classification.
• To accelerate the evaluation of NT1 treatments through rapid assessment.

Main Methods:

• Developed a three-module pipeline: automatic sleep stage classification, feature extraction, and phenotype classification.
• Trained sleep classifiers (UsleepEEG, UsleepEMG) on wild-type (WT) mouse data.
• Extracted features (EEG power bands, EMG metrics) and trained an L1-penalized logistic regression classifier using a Leave-One-Subject-Out approach.

Main Results:

• Achieved 97% accuracy in phenotype classification.
• The pipeline successfully captured disease progression in DTA mice across four timepoints.
• Demonstrated good generalization to data from other laboratories, though sensitive to artifacts.

Conclusions:

• The presented pipeline offers a fast and automated method for assessing NT1 probability in the DTA mouse model.
• This tool can significantly accelerate large-scale evaluations of potential NT1 treatments.
• Researchers should be mindful of potential data artifacts when applying the pipeline.