AI Advances in Predicting RNA Splicing

Researchers continue to grapple with the challenge of predicting how genetic variations influence RNA splicing, a critical process for gene expression and cellular function. While recent advancements in artificial intelligence (AI) have enhanced the ability to identify splice sites—regions where RNA is cut and joined—many models remain limited in their capacity to account for regulatory signals located far from these sites. These distant elements, which can span thousands of DNA bases, play a significant role in modulating splicing outcomes, yet their complex interactions with splice sites are poorly understood, hindering accurate predictions of splicing defects linked to disease.

Current AI models often focus on local DNA sequences surrounding splice sites, overlooking long-range regulatory mechanisms such as enhancers, silencers, or chromatin looping that influence splicing decisions. Studies have shown that mutations in these distant regions can disrupt splicing by altering the three-dimensional structure of DNA, affecting how RNA processing machinery recognizes splice sites. Developing models that integrate chromatin conformation data, epigenetic markers, or RNA secondary structures may improve predictive accuracy. However, such approaches require vast, high-quality datasets and computational methods capable of modeling spatial and functional relationships across the genome.

Addressing this gap is critical for advancing precision medicine, as splicing errors contribute to thousands of genetic disorders. Scientists emphasize the need for interdisciplinary collaboration to refine AI tools, combining genomics, structural biology, and machine learning. While progress is being made, overcoming the limitations of current models to capture long-range regulatory signals remains a key priority for unlocking the full potential of splicing prediction in both research and clinical applications.