Introme

Introme predicts the impact of coding and noncoding genetic variants on gene splicing, including non-canonical splice sites, by integrating multiple splice prediction outputs with splicing rules and gene architecture features.

Key Features:

• Prediction scope: Predicts effects of both coding and noncoding variants on gene splicing, including non-canonical splice sites.
• Machine learning integration: Employs a machine learning framework to combine predictions from multiple splice-detection tools with additional splicing rules and gene architecture features.
• Input aggregation: Integrates outputs from multiple existing splice prediction algorithms to inform a unified score.
• Benchmarking: Evaluated on a dataset of 21,000 splice-altering variants and achieved an area under the precision-recall curve (auPRC) of 0.98.
• Improved accuracy: Provides more accurate, holistic assessments of variant impacts on splicing compared to individual tools used in isolation.

Scientific Applications:

• Clinical diagnostics: Identify pathogenic splice variants that may be missed by single-tool analyses to reduce diagnostic failures in clinical settings.
• Variant interpretation: Prioritize and interpret coding and noncoding variants for pathogenicity assessment in genetic studies.
• Gene regulation research: Support studies of splicing mechanisms and gene architecture by quantifying variant effects on splicing.

Methodology:

Integrates outputs from multiple splice prediction tools, splicing rules, and gene architecture features within a machine learning framework to evaluate variants' potential effects on splicing.