Picard

[gwilymh]

Can anyone please tell me how to actually run Picard? I have downloaded and unzipped all the files, but do not know how to run them. Is there any software that you recommend for running a java based program out of a Windows 7 system? Are there any tutorials that you recommend?

Please let me know! I am keen to use this program, but don't know where to begin.

[pbluescript]

Picard is a set of command line tools, so you would need a Linux or Mac system to use a Unix command line. You could try running it on Cygwin.

[kmcarr]

Picard is a set of command line tools, so you would need a Linux or Mac system to use a Unix command line. You could try running it on Cygwin.

Picard tools are java programs; so long as you have the appropriate Java VM installed you can run them on the Windows command line.

@gwilymh, the standard procedure for invoking tools packaged as java jar files is:

C:\> java [options for java] -jar <full path to jar file> [options for jar]

For example:

C:\Documents and Settings\myHome> java -Xmx2g -jar C:\Program Files\picard-tools-1.61\MarkDuplicates.jar I=myInput.bam O=myDupMarked.bam

See the Picard tools website for detailed documentation of general options and options specific to each of the command line tools.

[gwilymh]

Thank you kmcarr and pblue script. Your help is much appreciated.

[aggp11]

Hi,

I am using CalculateHsMetrics in picard-tools to get an idea about the target region sequenced. However, could anybody tell me what is the normalized coverage in the PER_TARGET_COVERAGE file?

Thanks

[j.burke]

It looks to be mean_coverage/MEAN_TARGET_COVERAGE (from the metrics file)
The description of MEAN_TARGET_COVERAGE is "The mean coverage of targets that recieved at least coverage depth = 2 at one base." - http://http://picard.sourceforge.net...html#HsMetrics

[golharam]

How exactly are you supposed to interpret normalized coverage? For instance, I have a bunch of samples that got sequenced. For 1 specific target the length of the target is approximately 3000bp. Most of the samples have a low number of reads that map to the target, less than 500 reads. 2 samples have a more than 2000 reads that map to the same target. When I'm comparing the normalized coverage value, some of the samples with very low overall mapped read counts but have a high normalized coverage value for this target. The samples with a large overall mapped read counts have a lower normalized_coverage value for the same target.

That makes comparing the normalized_coverage value across samples difficult and probably not appropriate. Should I only look at the normalized_coverage values for within the sample comparisons? I'm reverting back to comparing mean coverage per target instead.