A predictive framework for stop-loss variants with C-terminal extensions.

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Stop codons normally terminate protein synthesis.
• Stop-loss variants lead to C-terminal extensions, altering protein function.
• Existing tools struggle to predict the impact of these extended proteins.

Purpose of the Study:

• To develop a novel machine-learning classifier, TAILVAR, for predicting the functional consequences of stop-loss variants.
• To integrate multi-omics data for improved prediction accuracy.
• To provide a framework for interpreting stop-loss variant effects in genetic diagnosis.

Main Methods:

• Developed TAILVAR, a machine-learning classifier.
• Integrated transcript- and protein-level features, and variant effect annotations.
• Analyzed evolutionary constraints and biophysical properties (hydrophobicity, aggregation propensity) of extended peptides.

Main Results:

• Transcripts without downstream stop codons show lower evolutionary constraint.
• Deleterious variants correlate with increased C-terminal hydrophobicity, reduced stability, and higher aggregation.
• TAILVAR outperformed existing methods, showing high correlation with functional experiments.
• TAILVAR reliably distinguishes benign from pathogenic variants, primarily via loss-of-function.

Conclusions:

• TAILVAR offers a systematic framework for interpreting stop-loss variants.
• Accurate prediction of elongated protein effects aids genetic diagnosis.
• This approach may facilitate the discovery of novel disease-associated genes.