Numerous experimental methodologies have been developed in the past decade to study 3D configurations of the human genome, including Hi-C and ChIA-PET [1, 2]. These “genomic interaction” data have provided key insights into the regulation of gene expression, and suggest that chromatin interactions are driven by discrete, yet spatially-associated, epigenetic features [3, 4]. File standards and tool suites have become essential to conduct efficient bioinformatics analyses; for example, single locus information can be encoded in the BED file format and manipulated using bedtools, enabling a wide variety of bioinformatics inquiries . However, it is currently challenging to fully interpret the biological impact of genomic interactions as tools do not yet exist to quickly and iteratively interrogate the extent to which both regions of paired loci are conserved across genomic datasets from diverse cell-types and contexts. While paired-genomic-loci data generated from these methodologies are widely available, the bioinformatics field has not yet developed either a file standard or analysis tools for their efficient manipulation.
There are currently several file formats for paired-genomic-loci data, however, none of these file formats were designed to enable efficient annotation and data manipulation. Existing file formats include those that encode read count information such as the matrix and the triplet sparse matrix formats , and others that encode the locations of paired segments and specialized metadata for particular pipelines, such as the HiFive ChromatinInteraction format . Although the matrix and triplet sparse matrix formats effectively communicate coverage depth across bins of the genome, they are restricted to fixed locus bin sizes, are not human-readable, and are cumbersome for genomic arithmetic. Additionally, while the ChromatinInteraction format, and the similarly structured bedtools bedpe format , may appear to be suitable storage formats for integration into a genomic arithmetic pipeline, as the two loci can be written in any order within the file, programmatic manipulation is unnecessarily complicated. Finally, the triplet sparse matrix and ChromatinInteraction formats are both specialized for the specific programs for which they were designed. Thus, to facilitate genomic interaction data manipulation, allow for variable locus bin sizes within a single data set, and allow for flexible metadata important to paired-genomic-loci, a new file standard is needed.
Numerous analysis tools exist to process, normalize, or call peaks from raw reads of paired-genomic-loci data [3, 6–9], yet there is no software that performs efficient manipulation and genomic arithmetic, analogous to bedtools, for single locus data, hindering the process of annotating and comparing chromatin interactions. For example, bedtools does not provide operations for bedpe that analyze both loci simultaneously, and there are no tools for genomic arithmetic within HiFive. Furthermore, a tool for converting to the ChromatinInteraction format, or for converting from the triplet sparse matrix format to visualization formats, does not currently exist. An analysis tool suite that performs efficient manipulation and genomic arithmetic of paired-genomic-loci data would allow for more complete analyses of these datasets, and thus the potential to gain deeper biological insights about the 3D conformation of the human genome.
Here we describe a new file standard for paired-genomic-loci data, the PGL format, and an analysis tool suite, pgltools, for genomic interaction data storage and manipulation. The PGL format supports genomic interaction data, allows for appropriate metadata, and enables efficient data manipulation. Pgltools performs genomic arithmetic on PGL files such as comparing, merging, and intersecting two sets of paired-genomic-loci, as well as integrates BED files with PGL files. Finally, we provide functions to convert other genomic interaction file formats to PGL files, and convert PGL files to multiple different visualization formats. This analysis tool suite will allow for iterative bioinformatics analyses and visualization of genomic interaction data, facilitating discovery and collaboration within the genomic interaction field.