fastseg

fastseg performs rapid segmentation of genomic data to detect copy-number segments and transcript boundaries from DNA microarrays, next-generation sequencing (NGS), and RNA microarray (tiling array) datasets.

Key Features:

• Versatile Segmentation Capabilities: Segments DNA microarrays, NGS data, RNA microarray (tiling arrays) and other matrix or vector-based genomic data to detect copy number segments and transcripts.
• Efficiency and Speed: Provides faster processing compared with earlier methods such as DNACopy while maintaining analytical accuracy.
• Bayesian Coverage Modeling: Uses a Bayesian framework that models depths of coverage at each genomic position with Poisson distributions and mixture components to decompose coverage into integer copy numbers and noise.
• Noise Filtering: Estimates and filters high-noise detections to reduce false positives in segmentation results.
• Benchmark Performance: Demonstrated higher precision (1−FDR) and recall for gains and losses versus five leading CNV methods on simulated datasets, NGS data from a male HapMap individual with implanted CNVs on the X chromosome, known HapMap CNVs, and high-coverage 1000 Genomes Project data.

Scientific Applications:

• Copy Number Variation (CNV) Detection: Quantitative analysis of NGS and microarray data to identify CNVs across genomes.
• Reduction of False Discoveries: Distinguishes true CNVs from noise using Bayesian Poisson mixture modeling and noise filtering to lower FDR.
• Transcript Detection from Tiling Arrays: Identifies transcript boundaries and expression segments in RNA tiling array data via segmentation.

Methodology:

Applies a Bayesian framework modeling per-position depths of coverage with Poisson distributions and mixture components to decompose coverage into integer copy numbers and noise, and includes estimation and filtering of high-noise detections.