Optimising SRMA for speed in the absence of a cluster

[gavin.oliver]

Hi,

I am trying to optimise the speed of SRMA for realignment. I am wondering a few things:

1) When using the RANGE argument to break up analyses, is there any way of specifying an entire chromosome without knowing its actual length?

2) Multithreading doesn't appear to speed things up so I am wondering if it is possible/acceptable to string together multiple srma commands on the command line with the "&" symbol in order to force different ranges to work on different cores.

3) Would it be possible/faster to use info from a BCF/VCF file to only run the relaignment over specific mutations rather than for the whole genome/exome.

Best,

Gavin
(This is a re-post as it was originally in the wrong forum).

[nilshomer]

Hi,

I am trying to optimise the speed of SRMA for realignment. I am wondering a few things:

1) When using the RANGE argument to break up analyses, is there any way of specifying an entire chromosome without knowing its actual length?

Yes, you should be able to just specify the chromosome. If not, let me know and I will add that for you.

2) Multithreading doesn't appear to speed things up so I am wondering if it is possible/acceptable to string together multiple srma commands on the command line with the "&" symbol in order to force different ranges to work on different cores.

Yes, there is a warning that multi-threading may not help. You can definitely have multiple SRMA processes, given enough memory.

3) Would it be possible/faster to use info from a BCF/VCF file to only run the relaignment over specific mutations rather than for the whole genome/exome.

Yes, it would be faster, though there may be variants that are not in the BCF/VCF that could be found after re-alignment. Use the RANGES option to specify a file with a list of ranges to re-align.

[gavin.oliver]

Yes, you should be able to just specify the chromosome. If not, let me know and I will add that for you.

I tried specifying with the format "RANGE=chr1" but it the program dies as follows Nils:

java.lang.Exception: BAM files and BAM indexes when using the RANGE or RANGES option
at srma.SAMRecordIO.<init>(SAMRecordIO.java:53)
at srma.SRMA.doWork(SRMA.java:159)
at net.sf.picard.cmdline.CommandLineProgram.instanceMain(CommandLineProgram.java:156)
at srma.SRMA.main(SRMA.java:98)