T-CNV

T-CNV detects and visualizes copy number variants from targeted hybrid-capture high-throughput DNA sequencing data (whole-exome and custom target panels) generated on short-read platforms such as Illumina and Ion Torrent to support CNV analysis in studies of genetic traits and disease.

Key Features:

• Robust CNV detection: Implements a two-step approach that computes log2 values of normalized read depth ratios between tumor and normal/control samples to identify CNV candidates.
• Bins method: Segments the genome into bins for detailed read-depth analysis of candidate regions.
• Gaussian Mixture Model (GMM) clustering: Uses GMM statistical modeling to classify read-depth signals and refine CNV region calls.
• Window-sliding method: Applies overlapping sliding windows across the genome to detect local variations in read depth.
• Performance metrics: Reported accuracy includes 95.42% sensitivity, 99.93% specificity, and 93.63% positive predictive value versus an MLPA-validated dataset, and simulation results at 100X coverage show 65.95% sensitivity and 88.71% positive predictive value.

Scientific Applications:

• Mendelian and sporadic trait studies: Enables identification of CNVs that influence Mendelian or sporadic phenotypes.
• Complex disease research: Supports investigation of CNV contributions to complex disease etiology.
• Targeted NGS panel analysis and visualization: Provides detection and visualization for hybrid-capture targeted panels, including whole-exome and custom panels from Illumina and Ion Torrent data.
• Method benchmarking and validation: Facilitates comparison with MLPA-validated datasets and performance evaluation in simulation studies.

Methodology:

Computes log2 normalized read depth ratios between tumor and normal/control samples, then confirms candidate CNVs using genome binning, Gaussian mixture model clustering, and a window-sliding read-depth method.