Systematic Computational Identification of Variants That Activate Exonic and Intronic Cryptic Splice Sites.

Area of Science:

• Genetics
• Bioinformatics
• Molecular Biology

Background:

• Genetic variants can alter gene splicing, leading to diseases like cystic fibrosis (CF) and X-linked dyskeratosis congenita.
• Accurate prediction of splice-altering variants is crucial for molecular diagnosis and understanding disease mechanisms.

Purpose of the Study:

• To develop and validate a novel computational method (CryptSplice) for identifying splice-altering genetic variants.
• To assess the frequency of cryptic splice-site activation by disease-associated variants.

Main Methods:

• Combined sequence-based machine learning with a context-dependent algorithm to predict splice variant effects.
• Compared the splice potential of variant sequences against reference sequences.
• Trained and validated the method on known splice sites and disease-associated variants in CFTR and DKC1 genes.

Main Results:

• CryptSplice achieved 93.3% accuracy in identifying canonical splice sites.
• Successfully predicted the effects of 86% of known splice-altering variants in CFTR.
• Identified 32 potential exonic and 3 intronic cryptic splice variants in CFTR, aiding diagnosis in 6/14 individuals.
• Identified and verified two splice-altering variants in DKC1.
• Predicted splice-site activation in 28% of pathogenic variants and 22% of variants of uncertain significance (VUSs) from ClinVar.

Conclusions:

• CryptSplice is an effective tool for identifying and predicting the functional impact of splice-altering variants.
• Cryptic splice-site activation is a more frequent mechanism in genetic diseases than previously recognized.
• The findings support routine consideration of cryptic splicing for variants within transcribed gene regions.