KCsmart

KCsmart analyzes multi-sample array Comparative Genomic Hybridization (aCGH) data to detect and quantify DNA Copy Number Alterations (CNA) across samples using kernel convolution to preserve continuous probe signal information.

Key Features:

• Kernel convolution-based estimation: Estimates the magnitude of CNAs at specific genomic locations by applying kernel convolution to probe intensities, local genomic context, and cross-sample signal distribution.
• Comparative analysis module: Performs supervised analysis between two user-defined groups of samples to identify regions with differential aberrations.
• Recurrent and class-specific CNA detection: Detects both recurrent CNAs and class-specific alterations at multiple genomic scales within the same analysis framework.
• No segmentation required: Identifies CNA regions without requiring additional segmentation steps on the data.
• Preserves continuous signal: Avoids discretization into gain/loss/no-change states and retains continuous probe intensity information.
• Memory efficiency: Employs computational approaches aimed at reducing memory usage during analysis.

Scientific Applications:

• Lymphoma CNA analysis: Applied to B- and T-cell lymphomas where it identified VDJ positive control regions and additional novel regions.
• Comparison with segmented t-test methods: Revealed regions not detected by t-tests on segmented data and produced less fragmented, more coherent CNA regions.

Methodology:

Applies kernel convolution that integrates probe signal intensity, local genomic information, and cross-sample signal distribution to estimate CNA magnitude without discretization.