CNV-P: a machine-learning framework for predicting high confident copy number variations

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Copy-number variants (CNVs) are a significant cause of genetic disorders.
• Accurate CNV detection from genome sequencing is crucial for disease research.
• Existing CNV detection software suffers from high false-positive rates.

Purpose of the Study:

• To develop a novel post-processing approach, CNV-P, for improving CNV detection accuracy.
• To reduce false-positive fragments identified by existing CNV detection tools.
• To enhance the reliability of CNV analysis in genetic disorder research.

Main Methods:

• Developed a machine-learning framework named CNV-P.
• Utilized features such as read depth (RD), split reads (SR), and read pairs (RP) around putative CNV fragments.
• Trained a classifier to distinguish true CNVs from false positives.

Main Results:

• Achieved over 90% precision and 85% recall rates in classifying CNVs on real biological datasets.
• Demonstrated significant performance improvement compared to state-of-the-art algorithms.
• Showed robustness of CNV-P across different CNV sizes and sequencing platforms.

Conclusions:

• The CNV-P framework effectively classifies high-confidence CNVs.
• This advancement can improve both basic research and clinical diagnosis of genetic diseases.
• Enhanced accuracy in CNV detection facilitates a better understanding of genetic disorder etiology.