Machine Learning-Based Prediction of Early Relapse in Multiple Sclerosis Using Peripheral Blood Transcriptomic Signatures

Background: Relapses are a major contributor to disability in multiple sclerosis (MS); therefore, early prediction of relapse risk is crucial for personalized therapy. With increasing MS incidence in Saudi Arabia and adoption of precision medicine under Vision 2030, molecular stratification is gaining momentum. Peripheral blood gene expression profiling provides a non-invasive approach to identify transcriptomic signatures predictive of relapse risk.

Objective:
To identify gene expression patterns associated with early relapse in MS and evaluate their predictive value using a machine learning approach.

Methods:
Peripheral blood microarray data obtained from the NCBI GEO repository were analyzed. Only those patients who had a confirmed diagnosis of MS along with relapse information and complete gene expression profile information (n = 94) were considered for the study. “Early relapse” was defined as relapse within 500 days. The genes were selected based on normalized values and differential gene expression analysis with FDR correction. They were then used to develop a supervised machine-learning logistic regression classifier.

Results:
Among 94 patients, 40 (42.6%) experienced early relapse. Clinical factors did not differ significantly between groups, suggesting limited predictive value. Transcriptomic analysis revealed molecular differences with FN1, PRELP, PVR, IGHG1, TNXB, CD9, and GPR143 genes downregulated, and NT5C1B and LSM8 genes upregulated in early relapse cases. Genes such as NRF1, ENPP2, KLF5, GRIK5, GRHPR, MUC7, and SSX1 were associated with high relapse risk in predictive modeling, while SPATA20, IGHG1, and CLCN4 played protective roles. The model yielded a mean cross validation AUC of 0.785, which indicated good performance.

Conclusion: Distinct molecular profiles were identified using peripheral blood transcriptomics that can predict MS relapse at an early stage. The model accuracy is commendable and shows that gene expression–based risk stratification can be used successfully for managing MS according to principles of precision medicine under Saudi Vision 2030.