TCGA Lancet data has only "calls" with t_alt_count = 0 or NA

[cansavvy]

What data file(s) does this issue pertain to?

The TCGA Lancet data : pbta-tcga-snv-lancet.vep.maf.gz of version 14

What release are you using?

v14

Put your question or report your issue here.

I was confused because Lancet's data for TCGA was not agreeing at all with Mutect or Strelka.
I looked into the VAF distributions and saw Lancet was all zeroes, because t_alt_count is only zeroes and NAs. t_alt_count's of 0 shouldn't be calls. Did something happen with a filtering step? It looks like it may have been filtered by n_alt_count > 0 ?

lancet <- data.table::fread("../../data/pbta-tcga-snv-lancet.vep.maf.gz", data.table = FALSE)
summary(lancet$t_alt_count)

  Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
      0       0       0       0       0       0     622 

As a positive control, and by contrast, Mutect and Strelka show no 0's for t_alt_count

mutect <- data.table::fread("../../data/pbta-tcga-snv-mutect2.vep.maf.gz", data.table = FALSE)
summary(mutect$t_alt_count)
Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   2.00    4.00    8.00   19.09   21.00 3528.00 

[jharenza]

Hi @cansavvy - thanks for noticing this - was this also an issue in V13? I am assuming so, since this file should not have changed. I am cc-ing @tkoganti, @migbro, and @yuankunzhu for this.

[jaclyn-taroni]

I can confirm that it was also an issue for v13:

> lancet <- data.table::fread("data/release-v13-20200116/pbta-tcga-snv-lancet.vep.maf.gz", data.table = FALSE)
> summary(lancet$t_alt_count)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
      0       0       0       0       0       0     622 

[jharenza]

Quick spot check on individual MAFs - seems they are all like this, so not a merge/filter issue (no filtering for these). I asked those above to check out the run, as it appears something went wrong there.

[jharenza]

@cansavvy and @jaclyn-taroni it looks like for some reason, all of the lancet tasks were run with the tumor/normal inputs swapped, and the correct IDs were used in VEP/VCF2MAF, and that is why the strange output. @migbro is queueing these up to rerun tonight and I will ask @tkoganti to pick up VCF2MAF/merge in the morning. The strelka2/mutect2 MAF runs were run correctly. Thank you again for finding this!

[jharenza]

closed via #507