SpliceAI-visual: a free online tool to improve SpliceAI splicing variant interpretation

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• SpliceAI is a deep learning tool for predicting splicing defects from DNA variations.
• Traditional SpliceAI outputs (delta scores) can be difficult to interpret and may mask severe consequences.
• Complex deletion-insertion variants (delins) are often not adequately handled by existing SpliceAI implementations.

Purpose of the Study:

• To introduce SpliceAI-visual, a free online tool that complements SpliceAI analysis.
• To address the limitations of SpliceAI's numerical delta scores and graphical output.
• To provide a user-friendly interface for interpreting splicing defects, including complex variants.

Main Methods:

• SpliceAI-visual utilizes raw SpliceAI scores instead of delta scores for more accurate interpretation.
• The tool offers a graphical presentation of results for enhanced user-friendliness.
• SpliceAI-visual supports the annotation of complex variants, such as complex delins.

Main Results:

• SpliceAI-visual provides a more intuitive interpretation of splicing defect predictions.
• The tool demonstrates relevance in assessing and modulating PVS1 classification criteria.
• SpliceAI-visual successfully elucidates complex splicing defects from both literature and unpublished cases.

Conclusions:

• SpliceAI-visual offers significant advantages over traditional SpliceAI analysis for variant interpretation.
• The tool's graphical output and handling of complex variants improve the assessment of splicing defects.
• SpliceAI-visual is accessible via Google Colab and integrated into the MobiDetails tool.