CNV-ECOD: A copy number variation detection method based on ECOD algorithm using next-generation sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Human Genetics

Background:

• Copy number variations (CNVs) are significant DNA structural variations (SVs) implicated in human diseases and genetic diversity.
• Accurate CNV identification is crucial for disease mechanism analysis, personalized medicine, and drug development.
• Existing next-generation sequencing (NGS) based CNV detection methods often suffer from high false positives and imprecise boundaries.

Purpose of the Study:

• To propose a novel, accurate, and robust method for detecting CNVs from single-sample NGS data.
• To address limitations of current CNV detection tools, specifically false positives and boundary inaccuracies.
• To enhance the reliability of CNV detection for downstream applications in clinical genetics and research.

Main Methods:

• Developed CNV-ECOD, a method integrating empirical-cumulative-distribution-based outlier detection (ECOD) for read depth (RD) signal analysis.
• Incorporated paired-end mapping (PEM) and split read (SR) strategies to refine CNV detection and correct false positives.
• Utilized a hierarchical progressive framework combining RD, PEM, and SR with an ECOD-based anomaly scoring mechanism.

Main Results:

• CNV-ECOD demonstrated superior performance compared to four existing methods in simulation experiments, achieving an optimal balance between precision and sensitivity.
• The method achieved the highest F1-scores and overlap density scores (ODSs) in real-sample experiments.
• Results indicate significant improvements in accuracy and robustness for CNV detection.

Conclusions:

• CNV-ECOD offers an effective solution for accurate CNV detection from NGS data.
• The integrated RD-PEM-SR framework with ECOD enhances CNV identification accuracy and boundary refinement.
• CNV-ECOD is poised to become a valuable tool for genetic variation analysis in research and clinical settings.