Proteomic and Machine Learning Analysis Predicts Treatment Response Signatures in Myasthenia Gravis

Area of Science:

• Immunology
• Proteomics
• Machine Learning

Background:

• Myasthenia gravis (MG) is an antibody-mediated autoimmune disease with varied treatment outcomes.
• There is a need for biomarkers to guide therapeutic decisions in MG.
• Proteomic profiling and machine learning can identify predictive biomarkers for treatment response.

Purpose of the Study:

• To identify serum proteomic signatures that can predict clinical response to treatment in MG.
• To compare predictive biomarker patterns between thymectomy plus prednisone and prednisone alone treatment arms.

Main Methods:

• Serum samples from an MG phase 3 trial were analyzed using liquid chromatography-mass spectrometry.
• Proteomic signatures were derived and associated with 6-month clinical outcomes.
• Machine learning approaches with internal validation were employed.

Main Results:

• Baseline serum proteomes differentiated MG patients from controls, highlighting pathways like complement activation.
• Distinct protein panels predicted clinical improvement in each treatment arm.
• Predictive protein patterns differed between treatment groups, suggesting treatment-specific biology.

Conclusions:

• Baseline serum proteomics can predict short-term clinical response in MG in a treatment-specific manner.
• These findings may enable biomarker-guided treatment selection and improve risk stratification.
• Validated biomarkers could inform future MG clinical trials and patient care.