Evaluation of CNV detection tools for prenatal diagnosis using amniotic fluid clinical whole-exome sequencing data
1Department of Prenatal Diagnosis Center, Women and Children's Hospital of Chongqing Medical University, China; Department of Prenatal Diagnosis Center, Chongqing Health Center for Women and Children, Chongqing 401147, China.
Background:
Copy number variations (CNVs) are significant contributors to genetic disorders, making their accurate detection crucial for prenatal diagnosis. While clinical whole exome sequencing (cWES) is increasingly used to identify sequence variants in prenatal samples, the feasibility of using the same cWES data to detect CNVs remains underexplored, particularly in amniotic fluid samples.
Methods:
We systematically evaluated seven CNV detection tools (CANOES, CODEX2, CLAMMS, ExomeDepth, XHMM, CNVkit, and GATK/gCNV) across various experimental conditions. Using 132 amniotic fluid (AF) samples, we assessed the impact of control sample type (AF and PB), CNV characteristics (including exon coverage and length), and tool integration strategies on detection accuracy. Performance was evaluated against CNV-seq results as the gold standard, with special focus on detecting pathogenic CNVs.
Results:
The choice of control sample substantially influenced detection performance, with PB generally yielding better results than AF. CNV characteristics, particularly exon coverage, clearly affected detection rates, with multi-exon CNVs being more reliably detected than single-exon variants. Individual tools demonstrated varying capabilities, with CANOES achieving the highest recall rate for pathogenic CNVs (70.3%) but all tools showing limited precision (highest 3.4%). Integration of multiple tools appeared to improve overall detection capability for pathogenic variations. This study suggests that effective optimization of WES-based CNV detection should consider control-sample selection, CNV characteristics, and multi-tool integration. While the current precision limitations indicate that WES-based CNV detection cannot yet replace dedicated CNV-seq testing, the findings may inform strategies to streamline prenatal diagnostic workflows by using WES data as an initial CNV screening step, potentially helping to reduce unnecessary additional testing in selected cases.
Conclusion:
These findings contribute to developing more effective and efficient genetic testing strategies in prenatal clinical settings, offering a foundation for future improvements in WES-based CNV detection methods.
