BEDTools: The Swiss-Army Tool for Genome Feature Analysis

Abstract

Technological advances have enabled the use of DNA sequencing as a flexible tool to characterize genetic variation and to measure the activity of diverse cellular phenomena such as gene isoform expression and transcription factor binding. Extracting biological insight from the experiments enabled by these advances demands the analysis of large, multi-dimensional datasets. This unit describes the use of the BEDTools toolkit for the exploration of high-throughput genomics datasets. Several protocols are presented for common genomic analyses, demonstrating how simple BEDTools operations may be combined to create bespoke pipelines addressing complex questions.

Keywords: bioinformatics; genome analysis; genome features; genome intervals; genomics.

Figure 1. Examples of genome arithmetic operations

Each tool in the BEDTools suite performs a relatively simple operation on one, a pair, or multiple genome interval datasets. The examples presented here reflect genome arithmetic operations on two genome interval files (green and blue). For example, if blue intervals represent gene annotations and green intervals represent DNA sequence alignments, then the result of the intersect tool (gray interval) represents the genomic interval that is shared between a single gene annotation and sequence alignment.

Figure 2. BEDTools scalability

The runtime in seconds (left panel) and memory usage (right panel) are compared when using either unsorted genome intervals (dashed red) or genome intervals that have been pre-sorted in genome order (solid red). As a basis of comparison, the BEDTools performance is compared to the BEDOPS toolset, both with (solid gray) and without (dashed gray) automatic error checking.

Figure 3

Histogram of genome-wide sequencing coverage for NA19146.

Figure 4

Cumulative distribution of sequencing coverage observed among all exome-targeted bases for NA12891.

Figure 5. Schematic of the BEDTools map tool’s functionality

The map tool summarizes the overlaps observed between two interval files, A and B. The result is a summary of all intervals in B that overlapped each interval in A. The summary is computed based on operations such as mean or max, which compute the average or maximum value for a given column from all of the intersecting B intervals.

Figure 6. Transcription factor binding occupancy at transcription start sites

Depicted are the consensus binding profiles of the Sp1 transcription factor (red) and a reverse cross-linked negative control (gray) as observed among all transcription start sites (TSS) in the human genome.

Figure 7

Histogram of the fraction of Dnase I hypersensitivity sites common to the 20 cell types compared.

Figure 8. Schematic of the BEDTools jaccard tool’s functionality

See main text for details.

Figure 9

Heatmap of Jaccard similarities observed for 20 fetal tissue samples based upon Dnase I hypersensitivity profiles.