Combining Mass Spectrometry with Machine Learning to Identify Novel Protein Signatures: The Example of Multisystem...

Area of Science:

• Biomarker discovery
• Proteomics
• Machine learning in medicine

Background:

• Multisystem Inflammatory Syndrome in Children (MIS-C) is a novel hyperinflammatory illness.
• Accurate diagnosis of MIS-C is crucial for timely intervention.

Purpose of the Study:

• To develop an integrated approach combining biomarker analysis and machine learning for identifying protein signatures.
• To apply this approach to distinguish MIS-C from other conditions.

Main Methods:

• Plasma samples from MIS-C patients and various control groups were analyzed using mass spectrometry.
• Support Vector Machine (SVM) algorithms were employed for protein pathway analysis and classification.
• Diagnostic accuracy was assessed through internal and external cross-validation.

Main Results:

• A three-protein signature (ORM1, AZGP1, SERPINA3) distinguished MIS-C from SARS-CoV-2 controls with high accuracy (93.5% AUC).
• A distinct signature (VWF, SERPINA3, FCGBP) differentiated MIS-C from pneumonia and Kawasaki disease (95.6% AUC).
• Proteomic analysis revealed increased inflammation/coagulation proteins and decreased lipid metabolism proteins in MIS-C.

Conclusions:

• The integration of mass spectrometry and machine learning (SVM) is an effective strategy for identifying disease-specific protein biomarker signatures.
• This approach facilitates accurate classification and diagnosis of complex pediatric inflammatory conditions like MIS-C.