Overlapping samples in MutSig2CV

[Diogo Pellegrina]

Hi,

I'm using MutSig2CV (on my linux server) and I got the following arror that I can't understand:

"

5 patients involved in an overlap.

2 cliques of overlapping samples.

12 unique samples.

Removing the following 3 duplicate patients:

    '30T'

    '50T'

    '56T'

"

I have 15 samples, and already tried to sort my input by the 'patient' column. Why is mutsig considering my samples to be duplicates?

Thanks

Diogo

[Barbara Hill]

Hello Diogo, 

I reached out to one of the MutSig maintainers and they had this to say:

MutSig considers patients to be duplicates when they share a significant amount of mutations.

The exact criteria are if two patients share 10 or more mutations comprising at least 10% of the union of their mutations, or if two patients share 3 or more mutations comprising ≥30% of the union of their mutations.

Sorting the input by the "patient" column (or any reordering of the input) has no effect on the duplicate patient filter (or indeed, on any MutSig results).

Generally, when the duplicate patient filter is tripped, it indicates something is seriously wrong with the input data.  A few common problems include:

• The somatic mutation calls contain a considerable amount of germline events, usually due to no-matched-normal calling or poor paired calling
• The somatic mutation calls contain many recurrent sequencing artifacts
• The "patients" comprise multiple biopsies from the same individual, thereby sharing truncal mutations

For either case, MutSig results will likely be poor, since its background model assumes that all samples are completely independent.

Could any of the above issues be relevant to your callset?

All the best,

Barbara

[Diogo Pellegrina]

Hi Barbara

I checked the union% of the mutations, most share more than 10 mutations (considering two mutations are equal if the chromosome and start position were the same), but the from all possible pairs the maximum I got was 7%.

As for the common problems, they are all made from different individuals, each with its own blood sample control.

Is there a workaround for this problem? I'll look for recurrent mutations that could be considered sequencing artifacts, but it might be challenging to make the distinction.

I don't know if this is relevant, but everything works perfectly if I also add to the .maf file lines from the .vcf files that weren't valid (for example those flagged "t_lod_fstar"). They enlarge the union sets, but they would include many false positives to the input.

Thanks for the quick response.

[Diogo Pellegrina]

Just noticed, those 3 samples have union% below 10% but they are the only ones with union% larger than 5%, maybe this is the actual threshold?

[Diogo Pellegrina]

Hi again,

From the README.txt I found the following option to be placed in a parameter file:

remove_duplicate_patients: Boolean.  Finds and removes duplicate patients in the

cohort by comparing mutation overlap between patients.  This should be

disabled when high levels of overlap are expected between samples (e.g.

primary/met combined cohorts).  Default: true

But MutSig2Cv is behaving very strangely. I tries using "remove_duplicate_patients false" and it didn't appear to change anything, and then I used the following parameter_file:

are_you_reading_me true

remove_duplicate_patients false

remove_duplicate_patients 0

remove_duplicate_patients False

remove_duplicate_patients false

what_happens_if_I_only_use_1_column

No error appeared, and it ran as if the parameters file wasn't being read.

Then I added another line:

remove_duplicate_patients false

remove_duplicate_patients 0

remove_duplicate_patients False

remove_duplicate_patients false

remove_duplicate_patients=false

And it printed this error:

Error setting field "remove_duplicate_patients=false"

  MException with properties:

    identifier: 'MATLAB:AddField:InvalidFieldName'

       message: 'Invalid field name: 'remove_duplicate_patients=false'.'

         cause: {}

         stack: [5x1 struct]

Invalid field name: 'remove_duplicate_patients=false'.

I also tried each of the above lines separately, but it always ran an as if the parameters file wasn't being read.

Thanks again,

Diogo

[Diogo Pellegrina]

Since there was a certain development in this thread I thought it would be better to creat a new thread: https://groups.google.com/forum/#!topic/genepattern-help/uGEWez0NYF4