bedshift

bedshift perturbs genomic region sets in BED files by applying randomized shifts, drops, additions, cuts, and merges to enable controlled benchmarking of region-set similarity metrics.

Key Features:

• Random perturbations: bedshift applies randomized shifting, dropping, adding, cutting, and merging of regions in BED files.
• Benchmark dataset generation: It generates perturbed BED files with known degrees of similarity relative to a reference to create controlled evaluation datasets.
• Similarity-metric evaluation: It enables benchmarking of region-set similarity metrics including Jaccard score, coverage score, Euclidean distance, and cosine similarity.
• Metric sensitivity insights: Empirical comparisons indicate the Jaccard score is particularly sensitive to added or dropped regions while the coverage score is more sensitive to shifted regions.

Scientific Applications:

• Benchmarking similarity metrics: Use perturbed datasets as ground truth to evaluate and compare region-set similarity metrics.
• Pipeline validation: Test and validate genomic analysis pipelines for ChIP-Seq, ATAC-Seq, and other region-based experiments by measuring performance on perturbed data.
• Algorithm development: Generate controlled evaluation datasets to develop and refine algorithms for comparing genomic region sets.

Methodology:

bedshift generates randomized BED files by applying specified perturbations (shift, drop, add, cut, merge) to a reference BED and systematically assesses the impact of each perturbation on region-set similarity metrics.