update-sort-bed-slurm1

Overview

update-sort-bed-slurm

citation: http://bioinformatics.oxfordjournals.org/content/28/14/1919.abstract version: 2.4.41 authors: Alex Reynolds and Shane Neph

$ update-sort-bed-slurm [ --slurm-memory <MB> ]

[ --slurm-partition <SLURM partition> ]

[ --slurm-workdir <working directory> ] [ --slurm-output <SLURM output directory> ] [ --slurm-error <SLURM error directory> ] [ --bedextract-path <path to bedextract> ] [ --sort-bed-path <path to sort-bed> ] [ --debug ] --input-original <old-bed-file> --input-backup <renamed-old-bed-file> --output-temp <intermediate-new-bed-file> --output-final <new-bed-file>

The "update-sort-bed-slurm" utility applies an updated sort order on BED files sorted per pre-v2.4.20 sort-bed, using a SLURM job scheduler to coordinate resorting each chromosome in "--input-original" per post-v2.4.20 sort-bed and writing the result to "--output-final".

When migration is finished, "--input-backup" specifies the new name of the original input.

As migration progresses, intermediate results are written to "--output-temp" and then written to "--output-final" upon completion.

Each sort task is given 8 GB of memory and is assigned to the "queue0" partition, unless the "--slurm-memory" and "--slurm-partition" options are set. If your input is larger than 8 GB, you will need to allocate more memory.

Because this launches all work on the specified cluster partition, the paths specified by "--input-original", "--input-backup", "--output-final", and "--output-temp" must be accessible to all computational nodes. For example, using /tmp may fail, as the /tmp path is almost certainly unique to a node; it is necesssary to use a path shared among all nodes.

Note that this utility will not work on entirely unsorted BED files, but only on files with a sort order from pre-v2.4.20 sort-bed, where there are ties on the first three columns.

In fact, until further refinements are made, this convenience utility could fail silently on inputs which are not BED, or which are not sorted per prev2.4.20 order, or which do not follow exact specification, all of which can lead a per-chromosome resort task to fail.