Machine-learning based lipid mediator serum concentration patterns allow identification of multiple sclerosis...

Area of Science:

• Biochemistry
• Neuroscience
• Computational Biology

Background:

• Multiple sclerosis (MS) pathology is increasingly linked to altered bioactive lipid metabolism.
• Existing diagnostic methods for MS can be invasive or lack specificity.

Purpose of the Study:

• To develop a complex serum lipid biomarker for the diagnosis of multiple sclerosis (MS).
• To identify specific lipid markers in serum that can differentiate MS patients from healthy individuals.

Main Methods:

• Utilized unsupervised machine learning (self-organizing maps, swarm intelligence, Minimum Curvilinear Embedding) to analyze serum lipid concentrations.
• Employed supervised machine learning (random forests, Bayesian statistics) for biomarker creation and validation.
• Analyzed serum concentrations of 43 different lipid markers in 102 MS patients and 301 healthy subjects.

Main Results:

• A distinct cluster structure was identified in serum lipid profiles, correlating with MS diagnosis.
• Eight key lipid markers (GluCerC16, LPA20:4, HETE15S, LacCerC24:1, C16Sphinganine, biopterin, PEA, OEA) were selected for the biomarker.
• The developed complex biomarker achieved approximately 95% sensitivity, specificity, and accuracy in predicting MS.

Conclusions:

• Serum lipid profiles contain patterns indicative of MS pathology.
• The developed lipidomic biomarker demonstrates high accuracy for MS diagnosis.
• This study supports the potential of serum lipidomics for establishing a non-invasive MS diagnostic biomarker.