ADaCGH2

ADaCGH2 performs segmentation and scalable analysis of array Comparative Genomic Hybridization (array CGH) data to detect DNA copy number alterations across large probe counts and many samples.

Key Features:

• Segmentation Algorithms: Incorporates Circular Binary Segmentation (CBS), wavelet-based smoothing (Liu et al.), HaarSeg (Ben-Yaacov and Eldar), Hidden Markov Models (HMM and BioHMM), GLAD, and CGHseg for identifying copy number variation regions.
• Parallelization for Efficiency: Performs parallel processing via forking on multicore CPUs, Message Passing Interface (MPI), and socket clusters for distributed computing.
• Data Management with ff: Uses the 'ff' package for on-disk storage and handling of datasets that exceed RAM, enabling processing of arrays with up to 6 million probes per array.
• Performance Improvements: Parallelization yields reported speedups ranging from 25 to 40 times faster on a 64-core machine compared to non-parallelized methods.

Scientific Applications:

• Large-scale genomic studies: Analysis of datasets with millions of probes and thousands of subjects to identify genome-wide copy number alterations.
• High-throughput copy number analysis: Batch processing of numerous samples for comparative genomic hybridization experiments.
• Cancer genomics: Detection and characterization of somatic DNA copy number changes relevant to tumorigenesis.
• Developmental biology and genomic instability studies: Investigation of copy number variation in contexts where genomic instability affects phenotype or development.

Methodology:

Integrates multiple segmentation algorithms (CBS; wavelet-based smoothing; HaarSeg; HMM and BioHMM; GLAD; CGHseg), executes parallel computation via forking, MPI and socket clusters, and uses the 'ff' package for on-disk data storage to process arrays up to 6 million probes.