IntSplice

IntSplice predicts the splicing consequences of single-nucleotide variations (SNVs) located at intron positions -50 to -3 near the 3' end of human introns to assess effects on pre-mRNA splicing regulation.

Key Features:

• Parameter extraction: Uses effect-size analysis on annotated alternative splicing to evaluate each intronic nucleotide and extracted 105 parameters influencing splicing signal strength.
• Model development: Initial support vector regression (SVR) attempts to predict percent-splice-in (PSI) scores in normal tissues were unsuccessful, leading to development of support vector machine (SVM) models using 110 parameters to classify pathogenic SNVs from the Human Gene Mutation Database versus normal SNVs from dbSNP.
• Performance: SVM models achieved sensitivity 0.800±0.041 and specificity 0.849±0.021 and showed better discrimination than Shapiro-Senapathy scores and MaxEntScan::score3ss.
• Validation: Validated against naturally occurring and nine artificial intronic mutations in the RAPSN gene associated with congenital myasthenic syndrome, correctly predicting nine of ten mutants.

Scientific Applications:

• Variant effect prediction: Predicts the impact of intronic SNVs in the -50 to -3 region on pre-mRNA splicing.
• Disease mechanism investigation: Supports study of molecular mechanisms underlying genetic disorders in which splicing defects contribute to pathology.
• Pathogenicity discrimination: Aids differentiation between pathogenic and benign intronic SNVs using comparisons to HGMD and dbSNP annotations.
• Experimental interpretation: Informs interpretation of experimental mutational analyses such as RAPSN variants linked to congenital myasthenic syndrome.

Methodology:

Performs effect-size analysis of annotated alternative splicing to extract 105 parameters, tested support vector regression for PSI prediction, and developed SVM models using 110 parameters trained to distinguish HGMD pathogenic SNVs from dbSNP normal SNVs, with performance compared to Shapiro-Senapathy and MaxEntScan::score3ss and validation against natural and synthetic RAPSN intronic mutations.