iSPHYNCS: Unsupervised Clustering in Questionnaires and Metadata Reveals Distinct Subtypes in the Narcolepsy...

Area of Science:

• Neurology
• Sleep Medicine
• Biomarker Discovery

Background:

• Central disorders of hypersomnolence (CDH) exhibit significant phenotypic heterogeneity.
• Identifying distinct patient subgroups is crucial for understanding CDH and developing targeted therapies.

Purpose of the Study:

• To uncover distinct participant clusters within CDH using a robust clustering pipeline.
• To explore phenotypic variability and identify potential biomarkers for CDH.
• To investigate the narcolepsy borderland group (NBL) for potential subtypes.

Main Methods:

• Utilized data from the Swiss Primary Hypersomnolence and Narcolepsy Cohort Study (iSPHYNCS) including 227 CDH patients and 33 controls.
• Employed a robust clustering pipeline involving random sampling, UMAP dimensionality reduction, K-means, and consensus clustering.
• Imputed missing data using a multiple random forest algorithm and performed post hoc biomarker analysis.

Main Results:

• Identified four distinct clusters: one healthy control group, one narcolepsy type 1 (NT1) group, and two NBL groups.
• One NBL cluster was characterized by higher symptom severity and psychiatric comorbidities, suggesting a potential subtype.
• The clustering pipeline demonstrated reproducible results, validating the NT1 and healthy control clusters.

Conclusions:

• CDH is characterized by significant phenotypic heterogeneity, necessitating personalized management approaches.
• The identified NBL clusters suggest a potential subtype associated with increased fatigue and psychiatric comorbidities.
• Cluster-based approaches show promise for advancing the understanding and management of CDH.