Allow building "markers" files for use with GISTIC

[lbeltrame]

The SEG files generated by CNVkit could be used with GISTIC to identify recurring regions of gains and/or losses. However, GISTIC has been built with array CGH in mind, and so requires a so called "markers" file:

The markers file identifies the marker names and positions of the markers in the original dataset (before segmentation). It is a three column, tab-delimited file with an optional header. The column headers are:

(1) Marker Name
(2) Chromosome
(3) Marker Position (in bases)

I'm not sure which of the files from cnvkit could be used to generate this (if possible). I have tried *.cnr files but I get errors with non-matching positions (I might be doing something wrong, though).

[etal]

Sure, that sounds useful. Could you link to a spec for the GISTIC input formats?

Or, if you'd like to take a shot at it with cnvlib -- the marker name is probably any unique string, so you could use the cnvlib.cnary.CopyNumArray.labels() method to generate them, and the marker position is probably cnvlib.cnary.CopyNumArray.start + 1. See the function cnvlib.fix.match_ref_to_sample for extracting the bins/markers matching between two tables, if that's the issue you encountered.

[lbeltrame]

Well, the definition is what I posted above (very generic, I know). What I've found "working" was:

1. Get all *.cnr files (1 or more)
2. Create a 3 column file with "generic ID" (as you suggested) + chromosome + start position for each row of all files
3. Concatenate all the end positions from the cnr files in the same file (so we have all start positions on top, then all end positions on bottom)
4. Get rid of duplicate lines

My plan afterwards is to use these marker files + the SEG files to compare how GISTIC runs with cnvkit versus "good" GISTIC results (made with array CGH).

[WJaratlerdsiri-Lab]

I also did the same as you. Let us know the comparison with CGH would be great.

James

[etal]

For reference, GISTIC documentation:

• ftp://ftp.broadinstitute.org/pub/GISTIC2.0/GISTICDocumentation_standalone.htm
• http://genepattern.broadinstitute.org/ftp/distribution/genepattern/modules_public_server_doc/GISTIC2.pdf
• http://gdac.broadinstitute.org/runs/analyses__2013_05_23/reports/cancer/KICH-TP/CopyNumber_Gistic2/nozzle.html

Looks like GISTIC performs best on autosomes alone.

[etal]

Already implemented as export_gistic_markers in cnvlib.export. Closing as answered — please reopen, or open a new issue, if a concrete problem remains. Related ongoing work: #370.

For general usage questions, we now use GitHub Discussions (Q&A): https://github.com/etal/cnvkit/discussions