SnpSift Intersect

This command intersects several interval files and produces a result of all intersections.

A typical usage example is to create a consensus of peaks from several ChIP-Seq experiments.

Supported file formats

Input files can be: BED, VCF, GFF, or TXT (assumed to be chr \t start \t end). Gzipped files (.gz) are also supported.

Usage

java -jar SnpSift.jar intersect [options] file_1.bed file_2.bed ... file_N.bed
Options:
    -minOverlap <num> : Minimum number of bases that two intervals have to overlap. Default: 1
    -cluster <num>    : Total number of intersecting intervals required (including the query interval itself). Default: 2
    -intersect        : Report the intersection (narrowest overlap) of all intervals. Default: true
    -union            : Report the union (widest span) of all intervals. Default: false
    -not <file>       : Filter out results that intersect with any interval in this file. Can be used multiple times.

The command requires at least two input files to produce meaningful results (with a single file, all intervals pass through unchanged).

Algorithm

For every interval in all input files, the command queries all other interval forests for overlapping intervals. Two intervals must share at least minOverlap bases to be considered overlapping. If the total number of overlapping intervals (including the query interval itself) is at least cluster, a consensus interval is produced.

With -intersect (the default), the consensus is the narrowest region shared by all overlapping intervals. With -union, the consensus is the widest span covering all overlapping intervals.

Duplicate consensus intervals (same chromosome, start, and end) are deduplicated in the output.

Output format

Tab-separated BED-like format: chr \t start \t end \t id

Coordinates are 0-based with exclusive end (BED convention). The id field is the concatenation of all contributing interval IDs joined by +.

Results that overlap with intervals from a -not file are filtered out.