cobind

cobind quantifies collocation strength between two sets of genomic intervals using six statistical measures to provide a nuanced and reproducible assessment of overlaps for genomic and epigenomic analyses.

Key Features:

• Statistical measures: Implements the Jaccard coefficient, Sørensen–Dice coefficient, Szymkiewicz–Simpson coefficient, collocation coefficient, pointwise mutual information (PMI), and normalized PMI for quantitative overlap assessment.
• Quantitative collocation assessment: Provides a continuous evaluation of overlap strength instead of relying on binary overlap thresholds.
• Reproducibility and comparability: Uses standardized statistical metrics to enable reproducible and comparable results across datasets and studies.
• Application to genomic datasets: Applied to analyses of CTCF binding sites from ChIP-seq, identification of cancer-specific open-chromatin regions (OCRs) using ATAC-seq across 17 cancer types, and examination of oligodendrocyte-specific OCRs from single-cell ATAC-seq (scATAC-seq).
• Discovery of regulatory elements: Facilitates re-discovery of CTCF cofactors and master regulators specific to cancers and oligodendrocytes.

Scientific Applications:

• Quantitative genomic overlap analysis: Enables detailed measurement of interval collocation strength in genomic and epigenomic studies.
• CTCF binding analysis: Supports analysis of CTCF binding sites derived from ChIP-seq data.
• Cancer OCR identification: Identifies cancer-specific open-chromatin regions using ATAC-seq across 17 cancer types.
• Oligodendrocyte chromatin analysis: Examines oligodendrocyte-specific OCRs identified from single-cell ATAC-seq (scATAC-seq).
• Regulatory factor discovery: Aids in re-discovering CTCF cofactors and master regulators relevant to cancer and oligodendrocyte biology.

Methodology:

Computes six statistical measures—Jaccard, Sørensen–Dice, Szymkiewicz–Simpson, collocation coefficient, pointwise mutual information (PMI), and normalized PMI—to quantify collocation strength between two genomic interval sets and avoid binary overlap thresholds.