Phenotypic Diversity in Multiple Sclerosis Can Be Represented by Four Additive Symptom Modules

Area of Science:

• Neurology
• Computational Biology
• Data Science

Background:

• Multiple sclerosis (MS) presents with heterogeneous combinations of sensory, motor, cognitive, and autonomic impairments.
• Current disability scales like the Expanded Disability Status Scale (EDSS) do not fully capture the modular nature of MS phenotypes.
• Lesions disseminated in time and space contribute to the complex and varied clinical presentations of MS.

Purpose of the Study:

• To develop a data-driven approach to represent the phenotypic diversity of multiple sclerosis.
• To identify latent symptom modules that characterize MS patient phenotypes.
• To quantify the degree of symptom admixture in individual MS patients.

Main Methods:

• Analysis of 4617 de-identified neurology progress notes from 577 MS patients.
• Utilized a large language model (GPT-5.2) to categorize notes based on 17 neurological features.
• Applied Non-negative Matrix Factorization (NMF) to identify latent symptom modules, with a preferred four-module solution.

Main Results:

• A four-module solution provided the clearest clinical interpretation, aligning with established MS functional systems: sensory-visual-pain, ataxic-spastic-falls, cognitive-psychologic-fatigue, and autonomic-bladder-bowel.
• Most patients exhibited admixed phenotypes, with varying degrees of module dominance and symptom overlap.
• A minority of patients (26.6%) showed highly dominant single-module profiles, predominantly in the sensory-visual-pain module.

Conclusions:

• MS phenotypic diversity can be effectively represented by mixtures of four latent symptom modules, rather than a single severity axis.
• This modular framework offers an interpretable method for quantifying MS phenotypes and understanding patient heterogeneity.
• Further research is needed to establish the biological relevance of these identified MS subtypes.