dbCNV: deleteriousness-based model to predict pathogenicity of copy number variations

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Copy number variation (CNV) involves abnormal gains or losses in DNA copy number, impacting clinical phenotypes.
• Predicting CNV pathogenicity is crucial for understanding disease associations.
• Existing methods lack focus on the deleterious significance of CNV features.

Purpose of the Study:

• To develop a novel computational model, dbCNV, for predicting CNV pathogenicity.
• To leverage quantitative features and their deleterious significance for improved classification.
• To establish a robust method for both five-tier and binary classification of CNVs.

Main Methods:

• Collected 79 informative features describing CNVs from SNP, gene, and region dimensions.
• Formulated quantitative methods for feature evaluation based on deleterious significance (variable and attribute types).
• Employed the Gradient Boosted Trees (GBT) algorithm to construct the dbCNV prediction model.

Main Results:

• The dbCNV model demonstrated feature value distributions consistent with deleterious significance.
• Achieved high accuracy in five-tier classification: 0.85 for loss and 0.79 for gain CNVs.
• Attained strong performance in binary classification with AUC values of 0.96 for gain and 0.81 for loss CNVs.

Conclusions:

• The dbCNV model highlights the promise of deleteriousness-based approaches for CNV classification.
• This method offers a valuable tool for supporting CNV pathogenicity prediction.
• dbCNV contributes to a deeper understanding of the pathogenic mechanisms underlying CNVs.