SVDiscovery Queue script
GenomeSTRiP Documentation | Created 2012-09-12 | Last updated 2012-09-21

Comments (0)

1. Introduction

SVDiscovery.q is a sample Queue script that is part of Genome STRiP.

This script runs deletion discovery over an input data set based on a set of bam files. The input bam files must have been previously run through the SVPreprocess pipeline to generate auxilliary metadata.

2. Inputs / Arguments

  • -I <bam-file> : The set of input BAM files.

  • -md <directory> : The metadata directory in which to store computed metadata about the input data set.

  • -R <fasta-file> : Reference sequence. : An indexed fasta file containing the reference sequence that the input BAM files were aligned against. The fasta file must be indexed with 'samtools faidx' or the equivalent.

  • -genomeMaskFile <mask-file> : Mask file that describes the alignability of the reference sequence. : See Genome Mask Files.

  • -genderMapFile <gender-map-file> : A file that contains the expected gender for each sample. : Tab delimited file with sample ID and gender on each line. Gender can be specified as M/F or 1 (male) and 2 (female).

  • -configFile <configuration-file> : This file contains values for specialized settings that do not normally need to be changed. : A default configuration file is provided in conf/genstrip_parameters.txt.

  • -runDirectory <directory> : Directory in which to place output files and intermediate run files.

  • -minimumSize <n> : The minimum size of an event that should be detected.

  • -maximumSize <n> : The maximum size of an event that should be detected. : This parameter also determines the size of the overlap between search windows when partitioning for parallel processing.

  • -windowSize <n> : For parallel processing, the size of each genomic locus to process in parallel. : The maximum size of an event that can be detected is also limited by the window size.

  • -windowPadding : This parameter specifies how far outside the search locus should be searched for informative read pairs. : This should be set based on the maximum insert sizes for read pairs in the input data set. : Ideally, we should estimate this parameter from the input data set.

3. Outputs

  • -O <vcf-file> : The main output is a VCF file containing candidate SV sites. : The output VCF file also contains a variety of metrics in the INFO field that should be used for filtering to select a final set of SV calls.

The SVDiscovery pipeline also produces a number of other intermediate output files, useful mostly for debugging. The content of these files is not documented and is subject to change. If the genome is processed in parallel, there will be output from each parallel partition plus merged genome-wide output.

4. Running

The SVDiscovery.q script is run through Queue.

Because Genome STRiP is a third-party GATK library, the Queue command line must be invoked explicitly, as shown in the example below.

java -Xmx2g -cp Queue.jar:SVToolkit.jar:GenomeAnalysisTK.jar \
    org.broadinstitute.sting.queue.QCommandLine \ 
    -S SVDiscovery.q \ 
    -S SVQScript.q \ 
    -gatk GenomeAnalysisTK.jar \ 
    -cp SVToolkit.jar:GenomeAnalysisTK.jar \ 
    -configFile conf/genstrip_parameters.txt \ 
    -tempDir /path/to/tmp/dir \ 
    -runDirectory run1 \ 
    -md metadata \ 
    -R Homo_sapiens_assembly18.fasta \ 
    -genomeMaskFile Homo_sapiens_assembly18.mask.36.fasta \ 
    -genderMapFile \ 
    -I input1.bam -I input2.bam \ 
    -O output.sites.vcf \ 
    -minimumSize 100 \
    -maximumSize 1000000 \ 
    -windowSize 10000000 \ 
    -windowPadding 10000 \ 
    -run \
    -bsub \ 
    -jobQueue lsf_queue_name \ 
    -jobProject lsf_project \ 
    -jobLogDir logs

5. Parallel Processing

The discovery pipeline is designed to allow parallelism across many processors. Parallelism is achieved by partitioning the problem space and running on overlapping genomic windows, then merging the output from each partition.

One practical strategy, which was employed in the pilot phase of the 1000 Genomes Project and is as shown in the example above, is to search for deletions between 100 bases and 1Mb in 10Mb windows (overlapping by 1Mb) using 10Kb padding, based on having paired-end libraries with relatively small insert sizes (100 bases - 2Kb).

It is also possible to partition the problem space by size of event. For example, you could search for small events (< 1Kb), medium sized events (1Kb - 100Kb) and large events (100Kb - 10Mb) in separate parallel runs, each with suitable window sizes, and then merge the outputs together. This approach is not currently implemented in an automated QScript, but if you are adventurous you could do this fairly easily using the same underlying building blocks as the standard pipeline.

6. Typical Queue Arguments

Queue typically requires the following arguments to run Genome STRiP pipelines.

  • -run : Actually run the pipeline (default is to do a dry run).

  • -S <queue-script> : Script to run. : The base script SVQScript.q from the SVToolkit should also be specified with a separate -S argument.

  • -gatk <jar-file> : The path to the GATK jar file.

  • -cp <classpath> : The java classpath to use for pipeline commands. This must include SVToolkit.jar and GenomeAnalysisTK.jar. : Note: Both -cp arguments are required in the example command. The first -cp argument is for the invocation of Queue itself, the second -cp argument is for the invocation of pipeline processes that will be run by Queue.

  • -tempDir <directory> : Path to a directory to use for temporary files.

7. Queue LSF Arguments

  • -bsub : Use LSF to submit jobs.

  • -jobQueue <queue-name> : LSF queue to use.

  • -jobProject <project-name> : LSF project to use for accounting.

  • -jobLogDir <directory> : Directory for LSF log files.

Return to top Comment on this article in the forum