On the core segmentation algorithms of copy number variation detection tools.

Area of Science:

• Genomics
• Bioinformatics

Background:

• Shotgun sequencing is crucial for high-throughput copy number variant (CNV) detection.
• Existing CNV detection tools exhibit performance variability, necessitating comparative analysis.

Purpose of the Study:

• To comprehensively analyze next-generation sequencing-based CNV detection tools.
• To compare the performance of Circular Binary Segmentation (CBS) and Hidden Markov Models (HMM) algorithms for CNV detection.

Main Methods:

• Developed a parametric model with statistical distributions for unbiased comparison.
• Incorporated bias correction for factors like guanine-cytosine (GC) content.
• Evaluated CBS and HMM performance across varying sequencing depths, segment lengths, and CNV complexities.

Main Results:

• CBS shows high precision; HMM demonstrates high recall under ideal conditions.
• HMM is advantageous at lower sequencing depths; CBS is better for small variant segments.
• HMM is more robust for complex CNVs and more time-efficient for large datasets.

Conclusions:

• HMM provides a more robust and efficient approach for CNV detection in complex scenarios and large-scale data.
• Findings offer guidance for developing improved CNV detection tools.