Vega

Vega segments chromosomal copy-number aberrations in array comparative genomic hybridization (aCGH) data by adapting the Mumford and Shah variational model for copy number (CN) segmentation.

Key Features:

• Variational Model Adaptation: Adapts the Mumford and Shah variational model to genomic CN data, minimizing an energy functional that balances data interpolation fidelity with solution complexity measured by boundary lengths between segmented regions.
• Region Growing Process: Employs a region growing technique with a termination condition driven entirely by the input data to produce adaptive segmentation boundaries.
• Robustness and Accuracy: Demonstrates robust performance across simulated noise conditions on synthetic datasets, reporting high recall and precision for aberration detection.
• Comparative Performance: Benchmarked against three state-of-the-art CN segmentation algorithms and shows consistent performance on both synthetic noise scenarios and real biological datasets.

Scientific Applications:

• Disease-associated aberration mapping: Produces detailed maps of chromosomal aberrations for studying genetic traits associated with diseases using aCGH-derived CN profiles.
• Cancer sample analysis: Applied to eight mantle cell lymphoma cell lines and two glioblastoma multiforme samples to detect known chromosomal aberrations in real-world biological research.

Methodology:

Models optimal segmentation as minimization of an energy functional integrating data fidelity and structural simplicity, adapts the Mumford and Shah variational model to CN data, and performs a data-driven region growing process with termination determined by the input data.