Inferring Multiple Sclerosis Stages from the Blood Transcriptome via Machine Learning.

Area of Science:

• Immunology
• Genomics
• Computational Biology

Background:

• Multiple sclerosis (MS) exhibits distinct gene and pathway dysregulations in peripheral blood mononuclear cells (PBMCs) across different disease stages.
• These molecular changes may offer potential for classifying MS patients, identifying early disease onset, and differentiating between MS subtypes.

Purpose of the Study:

• To develop and validate machine learning models for classifying MS stages using PBMC transcriptomic data.
• To assess the utility of PBMC gene expression profiles in distinguishing MS from non-MS individuals and various disease courses.

Main Methods:

• An unbiased machine learning workflow was employed to analyze PBMC transcriptomic profiles from over 300 individuals.
• The pipeline optimized and compared various machine learning algorithms, generating predictive models independent of demographic factors.
• Model performance was evaluated on an independent validation cohort.

Main Results:

• Machine learning models demonstrated high accuracy in classifying MS stages and differentiating between healthy controls and MS patients.
• The developed classifiers were robust and not influenced by demographic variables like age and gender.
• The study successfully identified stage-specific transcriptomic signatures in PBMCs relevant to MS.

Conclusions:

• Machine learning applied to PBMC transcriptomics provides a powerful tool for identifying the disease state and stage in multiple sclerosis.
• This approach holds promise for improving MS diagnosis, patient stratification, and potentially monitoring disease progression.