Early Prediction of Spinal Muscular Atrophy (SMA) Using Machine Learning and Genomic Variant Analysis

Spinal Muscular Atrophy (SMA) is a severe neuromuscular disease primarily caused by mutations in the Survival Motor Neuron 1 (SMN1) gene, leading to progressive motor neuron degeneration. Early diagnosis is critical as advanced therapies like Nusinersen are most effective before irreversible damage occurs. Current diagnostic practices, relying on symptom onset and manual genetic testing, often face delays. This paper proposes a machine learning based system for the early prediction of SMA by analyzing genomic variants. The system integrates clinical annotations from ClinVar with genetic sequence features extracted from FASTA files. We employ a hybrid feature extraction strategy, capturing sequence composition, splicing regulatory motifs, and SMN1/SMN2 paralog signatures. Ensemble learning algorithms, specifically Random Forest and XGBoost, are trained to classify variants as pathogenic or benign. The results demonstrate that the system effectively automates variant classification, offering a scalable and interpretable solution to support proactive healthcare decisions and early therapeutic intervention.