X-CNV

X-CNV predicts the pathogenicity of copy number variations (CNVs) using an XGBoost classifier and a meta-voting prediction (MVP) score to quantify pathogenic effects.

Key Features:

• XGBoost classifier: An XGBoost classifier generates probabilistic pathogenicity scores for CNVs.
• Integrated feature set: Integrates over 30 informative features, including allele frequency (AF), CNV length, CNV type, and various deleterious scores.
• Large cross-ethnic dataset: Trained on a dataset of more than 14 million CNVs from diverse ethnic groups covering nearly 93% of the human genome.
• Allele frequency calculation: Calculates AF values across populations to inform pathogenicity estimates.
• Meta-voting prediction (MVP) score: Implements an MVP score that quantitatively measures pathogenic effect based on probabilistic values generated by XGBoost.
• Performance metrics: Reports area under the curve (AUC) values of 0.96 in the training set and 0.94 in the validation set.
• Genome-wide prioritization: Prioritizes functional, deleterious, and disease-causing CNVs on a genome-wide scale.

Scientific Applications:

• Population genetics research: Enables population-level assessment of CNV pathogenicity using AF and cross-ethnic data.
• Disease-association studies: Identifies pathogenic CNVs for association with inherited traits and diseases.
• Diagnostic screening: Prioritizes deleterious CNVs for diagnostic and screening workflows.
• Cross-ethnic pathogenicity assessment: Evaluates CNV pathogenicity across diverse ethnic populations.

Methodology:

Integrates over 30 features (including AF, CNV length, CNV type, and deleterious scores) from a dataset of more than 14 million CNVs covering nearly 93% of the genome, trains and validates an XGBoost classifier to produce probabilistic pathogenicity scores, and computes a meta-voting prediction (MVP) score from those probabilistic values.