Improving detection of copy-number variation by simultaneous bias correction and read-depth segmentation

Area of Science:

• Genomics
• Bioinformatics
• Mammalian Genetics

Background:

• Structural variation, including copy-number variation (CNV), is crucial in mammalian genetics.
• High-throughput sequencing (HTS) offers potential for CNV detection but faces analytical hurdles.
• Existing methods often struggle with experimental biases in HTS data, particularly affecting read-depth estimation.

Purpose of the Study:

• To develop a novel, bias-aware computational method for accurate CNV detection from HTS data.
• To address the challenge of experimental biases confounding read-depth measurements in HTS.
• To improve the accuracy of CNV detection compared to existing read-depth-based algorithms.

Main Methods:

• Developed GENSENG, a read-depth-based CNV detection method.
• Employed a hidden Markov model and negative binomial regression framework.
• Integrated simultaneous bias correction with discrete copy-number change identification.

Main Results:

• GENSENG accurately identifies regions of discrete copy-number changes.
• The method effectively accounts for multiple experimental confounders.
• Extensive calibration demonstrated superior performance over existing read-depth-based CNV detection algorithms.

Conclusions:

• GENSENG provides a robust and accurate approach for CNV detection from HTS data.
• The framework supports combining read-depth data with other information types (read-pair, split-read) for enhanced analysis.
• A user-friendly, efficient implementation of GENSENG is publicly available.