Jamie Leckie, Sunny Z. Wu, Terryanne Standell, Toshifumi Yokota
Rare neuromuscular disorders impose a significant burden on patients, caregivers, and the health care system, yet, effective disease-modifying therapies remain limited. Antisense oligonucleotides (ASOs) have emerged as a promising therapeutic strategy, enabling targeted modulation of gene expression through mechanisms such as exon skipping, exon inclusion, and transcript degradation. However, the clinical efficacy of currently approved ASO therapies is often suboptimal. This limitation reflects not only poor target tissue uptake and delivery barriers, but also suboptimal design of ASO sequences and chemical modification patterns, which can compromise potency, safety, and translational robustness. Recent advances in machine learning have led to the development of ASO optimization platforms such as eSkipFinder and ASOptimizer, which aim to predict effective ASO sequences and chemistries for specific RNA targets. While these tools show considerable promise, their broader applicability remains limited due to a lack of comprehensive validation and the absence of integrated safety considerations. Further refinement and validation are necessary to improve their translational utility. Nevertheless, such platforms represent a critical advancement toward accelerating ASO development. By improving design precision, reducing reliance on extensive preclinical screening, and enabling researchers with limited ASO experience to generate optimized candidates, machine learning is poised to accelerate the development and clinical translation of ASO therapies for rare neuromuscular disorders.