Exploring the cooccurrence patterns of multiple sets of genomic intervals

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Understanding spatial relationships between genomic features is crucial for deciphering biological processes.
• High-throughput technologies like ChIP-seq generate vast amounts of genomic interval data.
• Existing methods for analyzing interval relationships are often limited to pairwise comparisons and do not scale well.

Purpose of the Study:

• To develop a statistical method and software tool for characterizing co-occurrence patterns among multiple sets of genomic intervals.
• To provide a scalable approach for analyzing complex spatial relationships in genomic data.

Main Methods:

• A finite mixture model is employed to describe genomic interval occurrences, with each component representing a distinct co-occurrence pattern.
• The Expectation-Maximization (EM) algorithm is used for model parameter estimation, yielding sufficient statistics for co-occurrence patterns.
• The method is implemented in a freely available R package named giClust.

Main Results:

• The developed statistical model accurately captures co-occurrence patterns within genomic interval data.
• Simulation studies and real-world data analysis demonstrate the method's ability to yield biologically meaningful insights.
• The giClust R package provides an efficient implementation for analyzing multiple sets of genomic intervals.

Conclusions:

• The presented method and giClust software offer a user-friendly approach for biologists to explore co-occurrence patterns.
• This tool facilitates the analysis of spatial relationships among a larger number of genomic interval sets than previously feasible.