CNV ConvertACNVResults: tag doc, give example command, question summary

[sooheelee]

Pending questions

ConvertACNVResults original summary blurb:

Convert files into TITAN and Broad CGA Allelic CapSeg (ACS) formats.  This tool uses spark, though running locally is fine.
As part of this process, this tool generates calls whether a particular segment is balanced (MAF=0.5)
IMPORTANT:  The additional CNLoH calls from this tool should be treated with a lot of skepticism.  Preliminary results indicated very poor performance.

• Eventually we will need a description and example of the TITAN format.
• Eventually we will need a description and example of Broad CGA Allelic CapSeg (ACS) format.
• "This tool uses spark, though running locally is fine." Is using Spark the default? What parameters are involved to use or not use Spark (e.g. --sparkMaster)? Can I instead say "To enable Spark, use the --sparkMaster option."?
• If the CNLoH calls need to be treated with a lot of skepticism and results indicate very poor performance, then shouldn't this tool be under the experimental category? Otherwise, if the tool has improved since this blurb was written, shall we remove this notice?
• What does "poor performance" mean? Unreliable or false? Can we instead say something along the lines of "Treat the additional CNLoH (Copy-Neutral Loss of Heterozygosity) calls with skepticism, as preliminary results using TCGA data have shown a high rate of false-positive calls." Or high rate of false-negative calls, etc.

[LeeTL1220]

@sooheelee Answers to questions...

• For descriptions and examples of ACS and TITAN, can you file issues? Or do those need to be addressed now. Those are not GATK tools.
• I am fine with putting this tool under the experimental category.
• Basically, the CNLoH calls have a high error rate. Both false positive and false negatives.

[sooheelee]

Made some edits based on your answers @LeeTL1220. One unanswered question:

"This tool uses spark, though running locally is fine." Is using Spark the default? What parameters are involved to use or not use Spark (e.g. --sparkMaster)? Can I instead say "To enable Spark, use the --sparkMaster option."?

[LeeTL1220]

You cannot disable spark in this tool. The default value for sparkMaster parameter is "local[*]", which runs the tool locally (using all available cores), rather than on a spark cluster.

[sooheelee]

Hmm, wouldn't it be better to be able to disable Spark altogether, e.g. with --disableSpark, like our other tools do? I think I read somewhere using Spark locally adds inefficient overhead to the compute. @LeeTL1220

I do see the appeal of having only one Spark parameter though.

[sooheelee]

Made changes to rename example command files and also to clarify the tool is non-Spark by default but can use Spark. Once the checks complete, I will just merge this.

[LeeTL1220]

Sounds good.

On Fri, May 26, 2017 at 11:11 AM, sooheelee notifications@github.com wrote:

Made changes to rename example command files and also to clarify the tool is non-Spark by default but can use Spark. Once the checks complete, I will just merge this.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/broadinstitute/gatk-protected/pull/1080#issuecomment-304308506, or mute the thread https://github.com/notifications/unsubscribe-auth/ACDXk4rHJfpJ4nAgdidu1OxwuHACfrOnks5r9uuggaJpZM4Nh0NO .

-- Lee Lichtenstein Broad Institute 75 Ames Street, Room 8011A Cambridge, MA 02142 617 714 8632