SpliceSelectNet: a hierarchical Transformer-based deep learning model for splice site prediction

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not reported in abstract — quantitative performance improvements over comparator models are not specified

• Primary Outcome: State-of-the-art splice site prediction performance across multiple benchmark datasets, with demonstrated capacity for aberrant splicing detection

• Nnt Or Sensitivity: Not reported — no sensitivity, specificity, AUROC, AUPRC, or F1-score values are provided in the abstract; full metrics must be extracted from the primary paper

• Confidence Interval: Not reported

Clinical Application

Feasibility in routine clinical genomics pipelines is uncertain. The 100 kb input window and Transformer architecture impose substantial computational requirements. Integration with existing variant interpretation workflows (e.g., as a replacement or complement to SpliceAI) would require benchmarking against clinically validated variant sets and regulatory approval pathways for diagnostic use. No implementation package, API, or clinical-grade software is described in the abstract. In Australia, clinical genomic sequencing is increasingly funded through Medicare (MBS items for germline cancer predisposition and rare disease) and state-based programs such as the Australian Genomics Health Alliance initiatives. Splice site prediction tools are directly relevant to variant classification under ACMG/AMP guidelines, which are adopted by RCPA-accredited laboratories. The TGA does not currently regulate in silico prediction tools as medical devices unless embedded in diagnostic software, but this landscape is evolving. RACGP and RACP guidelines on genomic medicine do not yet specify preferred computational tools for splicing analysis. SSNet could complement existing tools such as SpliceAI (already used in Australian diagnostic laboratories) if validated against Australian-relevant variant databases including those from diverse ancestral backgrounds represented in the Australian population. Patients undergoing clinical genomic sequencing for suspected hereditary cancer syndromes, rare Mendelian disorders, or undiagnosed disease where intronic or synonymous variants of uncertain significance may affect splicing; particularly relevant in paediatric and adult genetics, oncology, and neurology settings

Abstract

Accurate RNA splicing is essential for gene expression and protein function, yet the mechanisms governing splice site recognition remain incompletely understood. Aberrant splicing caused by mutations can lead to severe diseases, including cancer and genetic disorders, underscoring the need for accurate computational tools to predict splice sites and detect disruptions. Existing methods have made significant advances in splice site prediction but are often limited in handling long-range dependencies due to high computational costs, a factor critical to splicing regulation. Moreover, many models lack interpretability, hindering efforts to elucidate the underlying biological mechanisms. Here, we present SpliceSelectNet (SSNet), a hierarchical Transformer-based deep learning model that predicts splice sites from DNA sequences spanning up to 100 kb. By integrating local and global attention mechanisms, SSNet efficiently captures both proximal and distal regulatory signals while maintaining single-nucleotide resolution. Across multiple benchmark datasets, SSNet achieves state-of-the-art performance in splice site prediction and aberrant splicing detection. Systematic in silico mutagenesis demonstrates that attention scores reflect functional sequence importance, supporting their biological relevance. Long-range sequence perturbation experiments further show that SSNet captures distal regulatory effects beyond conventional receptive fields. Together, these results establish SSNet as a biologically interpretable framework for modeling long-range splicing regulation from genomic sequence.

References

1. 1.Miyachi, Y., & Nakai, K. (2026). SpliceSelectNet: a hierarchical Transformer-based deep learning model for splice site prediction. Nucleic Acids Research. https://doi.org/10.1093/nar/gkag625