GISTIC input parameters.

Question

[rin]

Hi everyone

I am running GISTIC 2.0 in segmentation data from TCGA and I was wondering whether it is advisable to set the input parameters as default or change the different thresholds. Running the analysis with everything on default and by only changing the confidence level to 0.99, I got many regions amplified/deleted so I thought of making the significance criteria more strigent. What would be your advice? Is this kind of plot considered "normal" output?

Here is how a plot I get looks like for example.

Thank you in advance!

UPDATE: Followed some recommendations on the GISTIC forum, but still I get many amplified and deleted regions. Any help on how I could extract any meaningful results from such noisy plots?

[2953531955]

Hi, did you fix the problem? I had the same problem. could you please share some solutions? Thank you!

[Kevin Blighe]

Hi 2953531955. Which specific problem are you having?

[zizigolu]

Hello

I have one specific question about GISTIC

Is this possible from GISTIC results that a peak being both amplified and deleted in a group of samples?

I have a set of samples in which 8p23.1 is being amplified in a range of chr8:7154386-7222701 and chr8:10360002-12034355 (2 times repeated in GISTIC amplified peaks) and deleted in chr8:11850002-12604291 and chr8:7220002-8178222

I really can not imagine how a peak could both amplified and deleted

Can somebody give me some intuition?

Thank you

Answer 1

[achernia]

Hi,

It looks like the file you are using has not been filtered for germline CNVs. If anyone of you out there have questions about GISTIC, write to us directly at the Broad instead of posting on Biostars!

http://software.broadinstitute.org/cancer/software/genepattern/contact

Answer 2

[Amitm]

Hi, Did you provide the 'control' CNV data using the -cnv param? That param can be used provide segments that are seen in healthy pop. and/ or control samples. The GISTIC repo. has a file provided for that param, under here - ftp://ftp.broadinstitute.org/pub/GISTIC2.0/hg19_support/

Pick the file CNV.hg19.bypos.111213.txt See if that helps.

Also, I am not sure if you have access to segments seen in the controls samples for the TCGA cancer type you are looking at. If you have, then it is probably a good idea to subtract such segments from the tum. sample segments first.

Segments that have reverse overlap satisfying >=50% are good candidates to filter out. Try

[rin]

Thanks for your answer! I was not aware for that option, but I cannot use it as my data are mapped on hg38. TCGA masked copy number data might be an alternative I guess? (They include only somatic variations while germline variations are filtered out)

[always_learning]

Sorry for naive questions but what do you mean by reverse overlap?