Arch Med Res. 2026 Oct 07. pii: S0188-4409(26)00140-2. [Epub ahead of print]57(8):
103518
Rare diseases collectively affect between 3.5 and 5.9% of the global population. Yet, patients still endure an average diagnostic interval of five to seven years before receiving an accurate explanation of their condition. Approximately 80% of these disorders have a genetic origin, and nearly 95% lack approved disease-modifying therapies that alter the disease course. This combination places enormous pressure on healthcare systems and affected families. In the last decade, artificial intelligence (AI), particularly machine learning, deep learning, and natural language processing, has emerged as a credible response to these gaps. This narrative review summarizes recent AI contributions across four interrelated dimensions: differential diagnosis and pattern recognition; prognosis and risk stratification; therapeutic discovery and drug repositioning; and the operational organization of rare disease care. Available evidence indicates that AI-based approaches can shorten diagnostic intervals, reduce unnecessary testing, and increase therapeutic options for orphan conditions. Despite these global advances, implementation remains highly unequal. This review foregrounds the Latin American perspective, setting itself apart from broader overviews. The region presents a scenario in which rich genetic diversity and growing collaborative networks coexist with infrastructural bottlenecks and underrepresentation in genomic databases. The integration of AI into rare disease care will depend on federated data infrastructures, transparent and interpretable models, harmonized regulatory frameworks, and sustained investment in workforce training. Patient organizations are essential partners, particularly in emerging economies.
Keywords: Artificial intelligence; Deep learning; Diagnosis; Machine learning; Precision medicine; Rare diseases