NCBoost v2: a classifier for non-coding single-nucleotide variants in Mendelian diseases.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Whole-genome sequencing (WGS) currently diagnoses rare diseases at a ~30% rate, necessitating improved variant pathogenicity prediction.
• Previous NCBoost (2019) identified pathogenic non-coding variants using sequence constraint features.
• Increased variant data and detection of purifying selection signals prompted NCBoost's update.

Purpose of the Study:

• To develop an updated pathogenicity score for non-coding single-nucleotide variants.
• To enhance the identification of pathogenic variants in Mendelian diseases.
• To improve diagnostic rates for rare genetic disorders.

Main Methods:

• Implemented NCBoost v2, a supervised learning model for variant pathogenicity scoring.
• Trained on an extensive dataset of pathogenic variants in monogenic Mendelian diseases.
• Integrated conservation features from Zoonomia and gnomAD, plus splice-altering scores.

Main Results:

• NCBoost v2 demonstrates superior performance compared to existing state-of-the-art methods.
• Achieved more consistent pathogenicity scores across diverse non-coding genomic regions.
• Improved scoring accuracy for splice-altering variants in Mendelian disease genes.

Conclusions:

• NCBoost v2 represents a significant advancement in predicting non-coding variant pathogenicity.
• The updated tool enhances the diagnostic yield of whole-genome sequencing for rare diseases.
• Freely available software and precomputed scores facilitate broader research application.