Request for help: Floating point exception when running GISTIC 2.0.23 (MATLAB R2014a)

[Kun Liu]

Hello Gene Pattern team,

I am seeking help with a reproducible Floating point exception when running GISTIC 2.0.23 to identify significant copy-number alteration (CNA) peaks from a large WES cohort.

---

Dataset

• Data type: Whole-exome sequencing (WES)

• Cohort size: >600 tumor samples

• CNA segmentation was performed prior to GISTIC, and all samples are included in a single seg file.

---

Software & command

• GISTIC version: 2.0.23

• Reference genome: hg38

• Command used:./gistic2 -b /WorkSpace/liukun/Test/Test1 -seg /WorkSpace/liukun/Test/Test1/GISTIC.input.txt -refgene /XSoftware/GISTIC2.0/refgenefiles/hg38.UCSC.add_miR.160920.refgene.mat -genegistic 1 -gcm extreme -rx 0 -brlen 0.5 -broad 1 -conf 0.95 -armpeel 1 -smallmem 0 -savegene 1 -qvt 0.1

---

Error

GISTIC terminates with a Floating point exception during execution.
The detailed runtime output and stack trace are provided in output.log (https://drive.google.com/file/d/1E9e9u_EYuuNA3NnzkHKDXOi75bMrzICO/view?usp=sharing).

Finding peaks on chromosome 21

------------------------------------------------------------------------
      Floating point exception detected at Wed Feb  4 09:32:45 2026
------------------------------------------------------------------------

Configuration:
  Crash Decoding     : Disabled
  Current Visual     : None
  Default Encoding   : UTF-8
  GNU C Library      : 2.35 stable
  MATLAB Architecture: glnxa64
  MATLAB Root        : /WorkSpace/liukun/Software/GISTIC2.0/MATLAB_Compiler_Runtime/v83
  MATLAB Version     : 8.3.0.532 (R2014a)
  Operating System   : Linux 5.15.0-164-generic #174-Ubuntu SMP Fri Nov 14 20:25:16 UTC 2025 x86_64
  Processor ID       : x86 Family 6 Model 106 Stepping 6, GenuineIntel
  Virtual Machine    : Java 1.7.0_11-b21 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode
  Window System      : No active display

Fault Count: 1


Abnormal termination:
Floating point exception

The crash consistently occurs during focal peak analysis (e.g., while processing individual chromosomes), suggesting a numerical issue rather than a file I/O or environment problem.

---

Input files

• The seg file (GISTIC.input.txt, https://drive.google.com/file/d/1mF8OEQwQ7iTuHId75nbtRq3MGO7q2odT/view?usp=sharing) used for this run is attached.

• All segment mean values are finite (no Inf, -Inf, or NaN).

• Chromosomes are encoded numerically (1–22).

• Very short segments and obvious outliers have been filtered.

---

Question 1: Possible causes of Floating point exception

Given the large cohort size (>600 samples), I would like to ask:

• In this context, I would appreciate guidance on how to adjust the GISTIC parameter settings to ensure that the analysis can run successfully. Specifically, are there recommended parameter choices for large WES-based cohorts to avoid numerical instability during focal peak calling?  

Any guidance on recommended parameter adjustments or known bugs would be greatly appreciated.

---

Question 2 (important): Num Markers for WES-based seg files

For whole-exome sequencing (WES) data, the definition of Num Markers in the seg file is not entirely clear.

Currently, following the implementation in
https://github.com/sbamin/canine_gistic2/blob/master/make_segments.R,
I set the number of markers as:

Num_Markers = as.integer((endPosition - startPosition + 1) / 100)

I would like to ask whether this setting is reasonable for WES-based CNA segmentation.

Specifically, for WES data, how should Num Markers in the GISTIC seg file be defined in a principled way?

Thank you very much for your time and for maintaining GISTIC.
I would be happy to provide additional information or a minimal reproducible example if needed.

Best regards,
Kun Liu

[Andrew Cherniack]

Hi

1) I am not sure what is causing the floating point error, but your seg file looks weird because there is a limited number of copy number levels, which could be causing the problem since GISTIC is expecting more diversity.   How did you generate your WES CN data before running it through "make segments"? Were the numbers integers? For WES data, segment breakpoints are typically at the beginning or end of exomes, but this doesn't appear to be the case in your data, as every segment endpoint aligns with the start point.

2) The marker number column is no longer used in  GISTIC 2.0, so it doesn't matter and is not causing the floating-point error. However, in WES, you typically generate CN levels for each exon. The number markers should equal the number of exons in a segment.   

Regards,

Andy 

___________________________
Andrew Cherniack, PhD                                               
Senior Group Leader

Broad Institute of Harvard and MIT and

Dana-Farber Cancer Institute

email: ache...@broadinstitute.org
___________________________

--
You received this message because you are subscribed to the Google Groups "GenePattern Help Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to genepattern-he...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/genepattern-help/d79375f1-6ced-4db3-96f5-20adcdfb4bafn%40googlegroups.com.

[Andrew Cherniack]

One more really important thing.

Did you input .seg file include a header row.

If not, that will cause GISTIC to error out.

The headers should be something like this:

ID chrom loc.start loc.end num.mark seg.mean

[Kun Liu]

Thank you very much for your suggestions—they have been extremely helpful to me. I have successfully run GISTIC on the GenePattern website using the exact same input files and parameters without any issues. As other users have pointed out, when I reduced the number of samples to fewer than 400, the local version also worked properly. Additionally, I noticed that some of the code on GitHub (https://github.com/broadinstitute/gistic2) has already been updated compared to the official executable distribution available at ftp://ftp.broadinstitute.org/pub/GISTIC2.0/. I originally intended to rebuild the gp_gistic2_from_seg executable based on the latest GitHub source code. However, as someone with limited programming experience—particularly unfamiliar with MATLAB and the packaging process—I was unable to accomplish this task. In the future, if it’s convenient for you to release an updated version of GISTIC incorporating these improvements, I believe it would be a tremendous benefit to the entire GISTIC 2.0 user community.

[Andrew Cherniack]

Hi Ken,

I successfully ran GISTIC on your seg file.  I have enclosed the execution_log sd_out.txt.

I have run genepattern GISTIC on files with several thousand samples in the past, so 600 should not be an issue.

Regards,

Andy

___________________________
Andrew Cherniack, PhD                                               
Senior Group Leader

Broad Institute of Harvard and MIT and

Dana-Farber Cancer Institute

email: ache...@broadinstitute.org
___________________________

To view this discussion visit https://groups.google.com/d/msgid/genepattern-help/8d41516c-8f8a-4384-8391-2d69ac31fcabn%40googlegroups.com.

[Kun Liu]

Hi Andy,

Thank you so much for running GISTIC on my seg file—I really appreciate your help! I’ve reviewed the execution log (sd_out.txt) and everything looks great.

It’s reassuring to know that the sample size isn’t an issue, especially given your experience with much larger datasets.

I suspect the issue I encountered might be due to the version of the gp_gistic2_from_seg executable I downloaded—possibly the publicly available GISTIC 2.0.23 binary isn’t the exact one used in GenePattern. If it’s not too much trouble, would you be willing to share the version of gp_gistic2_from_seg that GenePattern uses? That would be incredibly helpful for my local runs.

Thanks again for your generous support and expertise!

Best regards,
Kun