Picard CollectHsMetrics PCT_EXC_OFF_TARGET

Follow

Holly Corbitt

• May 07, 2020 04:09

Hi,

I am trying to calculate just the number of bases excluded due to being off target. I thought I could use PCT_EXC_OFF_TARGET, but this actually looks like it is a percent of a percent. All of these percents of different exclusion reasons equal 1 in all of my data sets, which confirms my theory that these are the percents of only the total excluded bases. 

PCT_EXC_BASEQ 0.007394
PCT_EXC_DUPE 0.069799
PCT_EXC_MAPQ 0.053678
PCT_EXC_OFF_TARGET 0.798181
PCT_EXC_OVERLAP 0.073081

My question is... how do I calculate the number of excluded off target bases or at least the percent of the total passing filter bases?

 

Thanks!

1

• 

  Megan Shand

  • May 07, 2020 18:25

  Hi,

  The number of excluded off target bases should be PCT_EXC_OFF_TARGET * PF_BASES_ALIGNED. The divisor for calculating all of the PCT_EXC metrics is PF_BASES_ALIGNED.

  Just for clarity, PF stands for passing filter, but it references the vendor filter specifically. So it's reads that pass the sequencer filter, not any other filters.

  Hope that helps!

  0

  Comment actions Permalink
• 

  Holly Corbitt

  • May 07, 2020 18:48

  Thank you for the quick response...

  So these aren't percents of percents (i.e. why do all the PCT_EXC add up to 1 then)? If you did this sort of calculation for all of them then you would get the total number of bases passing filter.

  My other line of evidence is that if you use PCT_EXC_OFF_TARGET * PF_BASES_ALIGNED to get the total number excluded for off target and add that back to the on target bases, you get more than the total bases.

  Here is my example data:

  PF_BASES_ALIGNED 4,915,335,315

  ON_TARGET_BASES 1727103724

  PCT_EXC_BASEQ 0.007394
  PCT_EXC_DUPE 0.069799
  PCT_EXC_MAPQ 0.053678
  PCT_EXC_OFF_TARGET 0.798181
  PCT_EXC_OVERLAP 0.073081

   

  0

  Comment actions Permalink
• 

  Megan Shand

  • May 08, 2020 01:22

  There have been some changes in that section of the code recently. Which version are you using?

  0

  Comment actions Permalink
• 

  Holly Corbitt

  • May 14, 2020 19:14
  • Edited

  Hi Megan,

  
  Maybe that is the root issue. We are still using v1.14.

  I re-ran it on the new version and the metrics now make sense.

   

  Thanks!

  0

  Comment actions Permalink

Please sign in to leave a comment.