BayesCNV: A Bayesian Hierarchical Model for Sensitive and Specific Copy Number Estimation in Cell Free DNA.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Detecting copy number variations (CNVs) in next-generation sequencing (NGS) data is challenging, especially with low-signal cell-free DNA (cfDNA) and targeted panels.
• High noise levels in cfDNA sequencing complicate accurate CNV identification.

Purpose of the Study:

• To develop and validate BayesCNV, a Bayesian hierarchical model for robust gene-level copy ratio estimation from targeted sequencing data.
• To provide accurate CNV calling with uncertainty quantification and an evidence-based quality control (QC) metric for cfDNA analysis.

Main Methods:

• Implemented a Bayesian hierarchical model for gene-level copy ratio estimation using targeted amplicon read depths.
• Utilized thermodynamic integration for reliable estimation of marginal log likelihood for QC.
• Benchmarked BayesCNV against IonCopy and DeviCNV using reference samples with known CNVs on the OncoReveal Core Lbx panel.

Main Results:

• BayesCNV achieved a sensitivity of 0.87 and specificity of 0.996, outperforming competitor methods.
• The marginal log likelihood effectively distinguished degraded from high-quality samples in FFPE datasets, surpassing conventional QC metrics.
• Demonstrated accurate and interpretable gene-level CNV estimates with uncertainty quantification.

Conclusions:

• BayesCNV offers a robust and accurate solution for CNV detection in targeted cfDNA sequencing.
• The integrated QC metric enhances the reliability of CNV calling in challenging low-input samples.
• The method provides interpretable results and uncertainty quantification crucial for clinical applications.