Mixed PoN (Blood + FFPE) ?

cbaocbao Member, Broadie

Hi,
I have some FFPE WGS data. I am able to access TCGA bam file. I want to build a PoN for mutation and CNA analysis with 900+ blood tissues (TCGA) and 30+ FFPE tissues (TCGA + my own). Does this PoN make sense? Many thanks!

Or, simply, which is better?
1. Only 900+ blood tissues
2. Only 30+ FFPE tissues
3. Mixed PoN

Best,
Chunyang

Best Answers

  • shleeshlee CambridgeMember, Broadie, Moderator
    Accepted Answer

    Hi @cbao,

    The considerations for a mutation PoN versus CNA PoN would be different. The best thing to do is to test out the suitability of the sets for each workflow.

    Building the mutation PoN with FFPE samples sounds dangerous given these samples will have systematic errors in sequencing. We have a workflow (in beta) to filter out variants from a callset that are likely due to FFPE errors (FilterByOrientationBias) that takes in a Mutect2 callset. I think you should examine the extent of these errors for each of your samples to determine suitability.

    As for CNA analysis, let me refer you to our most recent CNV tutorial. I've attached the worksheet.

Answers

  • shleeshlee CambridgeMember, Broadie, Moderator
    Accepted Answer

    Hi @cbao,

    The considerations for a mutation PoN versus CNA PoN would be different. The best thing to do is to test out the suitability of the sets for each workflow.

    Building the mutation PoN with FFPE samples sounds dangerous given these samples will have systematic errors in sequencing. We have a workflow (in beta) to filter out variants from a callset that are likely due to FFPE errors (FilterByOrientationBias) that takes in a Mutect2 callset. I think you should examine the extent of these errors for each of your samples to determine suitability.

    As for CNA analysis, let me refer you to our most recent CNV tutorial. I've attached the worksheet.

  • cbaocbao Member, Broadie

    Thank you very much! Let me read the tutorial first.

  • cbaocbao Member, Broadie

    @shlee said:
    Hi @cbao,

    The considerations for a mutation PoN versus CNA PoN would be different. The best thing to do is to test out the suitability of the sets for each workflow.

    Building the mutation PoN with FFPE samples sounds dangerous given these samples will have systematic errors in sequencing. We have a workflow (in beta) to filter out variants from a callset that are likely due to FFPE errors (FilterByOrientationBias) that takes in a Mutect2 callset. I think you should examine the extent of these errors for each of your samples to determine suitability.

    As for CNA analysis, let me refer you to our most recent CNV tutorial. I've attached the worksheet.

    Thank you for your kind reply. From this tutorial and #9143, it seems that we should construct the CNV PoNs using the blood normals of the tumor cohort(s) because matched normal (FFPE) brings some noise. Simply, we should just use blood normals but not matched normals for CNV PoN. Is that correct?

    Maybe, I missed something and am not able to find "cohort_C.sam" file.
    So, what is different between cohort-M and cohort-C?

    Could you please help me with checking my answer for these questions? I am not quite sure ...
    → In a somatic analysis, what is better for a PoN: tissue-matched normals or blood normals?
    A: Blood normals only
    → Should we include our particular tumor’s matched normal in the PoN?
    A: No

    Thank you so much!

    Best,
    Chunyang

  • cbaocbao Member, Broadie

    Hi @shlee ,

    Based on the worksheet, I should use SparkGenomeReadCounts instead of CalculateTargetCoverage if I want to build a WGS PoN and perform WGS analysis by that PoN. Is it correct?

    Best,
    Chunyang

  • cbaocbao Member, Broadie
Sign In or Register to comment.