Monday, September 9, 2019

Informal Notes: Collection of References Regarding the Frequency of Pathogenic Mutations in Moderate-to-High Cancer Risk Genes

While I typically prefer to keep Twitter responses less formal than blog posts (which in turn are less formal than pre-prints or peer-reviewed journal articles), I feel kind of bad continuing to add citations to an earlier discussion thread.

So, I thought I would copy the notes here, and provide information for those that are interested (without having all the notices go to the person with the original response):

BRCA Exchange tries to provide some variant information (for BRCA1 / BRCA2) also provides some information about specific variants (but counts are often low).  Color also has this page for Breast Cancer Awareness Month (October).

Ambry provides an Interactive Prevalence Tool (in collaboration with Mayo)

Stanford BRCA Decision Tool provides some statistics to guide discussions about risk prevention options (although that relates more to risk than total frequency; however, this with total numbers is essentially what I want to see)

MSKCC lists total inherited breast cancer rate at 5-10%

There is a link from the CDC website listing the BRCA1+BRCA2 rate for hereditary breast cancer as 3% for breast cancer and 10% for ovarian cancer.  If you sum the frequencies, that is somewhat close to LaDuca et al. 2019.

There is also a different source of information from the CDC, mentioning for all hereditary breast and ovarian cancer (which must be greater than the sum of BRCA1+BRCA2), which was 5-10% and 10-15% respectively.

  • Editorial about BRCA mutation frequency and risk

Hu et al. 2018
  • Table 2 shows some percentages per gene

LaDuca et al. 2019
  • Learned about from @kristinclift
  • GenomeWeb review mentions "13.8 percent of ovarian cancer patients carried germline pathogenic variants in at least one of the 32 genes tested"
  • Supplemental Table S7 also shows a higher BRCA1/2 ovarian cancer percent (3-5%, OC) than breast cancer percent (1-2%, BC)
  • There are all high-risk patients 
  • Total numbers are fairly different: 66,954 BC, 9,106 OC, 1,087 BC+OV.
  • Supplemental Table S6 has gnomAD control frequencies, but I think variant calls may need to be more specialized for targeted gene panels.
    • In other words, those percentages look low compared to that matched control study, but they do provide those details.
  • I also think they used results from the reports from different Ambry panels, rather than obtaining and re-analyzing raw data independently.  So, I wonder how much of an effect that had.

Unpublished (?) CARRIER Study Data
  • In a CARRIERS plot (a PDF from a Clinical Cancer Genomics Conference presentation that I don't know if I have permission to post), getting control variants for BRCA1/2 below 0.4% puts BRCA1/2 case mutation rates at under 1.4%.
    • I think that is for breast cancer.
  • Plots for 14 individual studies (for a total of 10,000s of samples, with matched controls) were shown for 21 genes of with overall case rate of ~6% and control rate of ~2.5%, as well as 17 higher-risk genes with overall case rate of ~5% and control rate of ~2% (which I guess could up to a 30% “false positive” rate either way).

Variability in Variant Calling
  • May need more specialized variant calling for targeted gene panels - Warden et al. 2014
    • For example, notice the difference in scale for the y-axis in Figure 9, where the relative “novel” variants in controls is a proxy for false positives:
  • You may get better concordance between variants with reprocessing (in this case, using a more typical processing strategy for WGS and Exome Data) - blog post on custom scripts and precisionFDA comparisons.
  • In addition to, I think cBioPortal can be a useful research to look at mutation information across different cancer types and studies
    • cBioPortal has some tutorials that can be viewed here
    • For example, I think I could see results more similar to what I read elsewhere (with BRCA2 being found at higher frequencies) if I removed variants of unknown significance
    • I think 47:45 - 51:55 of this webinar also gives some useful caveats to mutation and copy number calling that might be worth taking into consideration (where a region that is supposed to have a homozygous deletion has a variant with an allele frequency that you would normally consider a match for a heterozgyous variant, when a true homozygous deletion call shouldn't have any variants)

Polygenic Risk Scores

These are arguably more for risk assessment rather than frequency (although the risk estimate vary across the population).

Nevertheless, I would also be interested in seeing more results from Myriad myRisk and AmbryScore-Breast Polygenic Risk Scores (PRS).  This is in part because I haven't been highly satisfied with the polygenic scores that I was able to apply to myself, but I should note that clinical features are included in the calculation for these scores (beyond just the genetic information).

For example, I noticed that some more information Myriad myRisk was provided in these slides, which I believe reference this announcement.  I believe that is a reference to the Hughes et al. 2020 and Gallagher et al. 2020 papers.  However, what caught my eye was this figure, which I don't think it is directly from those publications?

I am not sure if I am correctly understanding that both Myriad myRisk and AmbryScore-Breast use "clinical history" information?

As a possibly separate point, I thought Figure 2 of Fahed et al. 2020 was interesting because of the difference in the range of risk estimates by PRS with versus without a BRCA1 or BRCA2 mutation.  The stratification without being a BRCA1/2 carrier was more limited.  However, I would be interested to get a better sense of how PRS stratification compares to variants in or near the BRCA1 or BRCA2 genomic footprint.  I think the later more along the goals of resources like BRCA Exchange.

Those interested in these notes may also be interested in this post with notes on the range of risk estimates.

Change Log:

9/9/2019 - public post date
9/10/2019 - add "informal note" category, along with link to original response
9/13/2019 - add CDC links, as well as some extra information not previously copied over from earlier Twitter discussion
10/10/2019 - add Color link for BRCA1/2 mutation frequencies
11/12/2019 - add link to other blog post
5/8/2020 - add cBioPortal notes
7/6/2020 - add "breast cancer" label
7/7/2020 - add note about interest in hearing about a range of PRS experiences.
8/14/2020 - add additional PRS links
8/16/2020 - minor change
3/9/2021 - add link for Ambry Prevalence Table
4/23/2021 - move link from earlier blog post to this blog post
6/26/2022 - add link to PRS within BRCA1/2 carriers

No comments:

Post a Comment

Creative Commons License
My Biomedical Informatics Blog by Charles Warden is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 United States License.