Announcement

Collapse
No announcement yet.

When is a variant not a variant?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • When is a variant not a variant?

    If for a particular variant, reference=T and genotype=A; is that really a variant? I mean, if the read had come from the other of the base pair at that same location, it would be recorded as reference=T and genotype=T, right?

    Isn't every location where a C is called, really a C on one side of the strand and a G on the reverse strand?

    Help, I'm definitely missing something here!

  • #2
    The convention is to report results on just one strand of the double helix (called the plus or forward direction). The vast majority of SNPs involve other combinations (e.g. A and G). The small number of SNPs where the two possible alleles are also complementary base pairs (e.g. A and T) can be distinguished because the probes include neighboring bases.

    Comment


    • #3
      So, all calls are from the same strand, and none are from the reverse strand?

      Comment


      • #4
        Yes, that's the intention, anyway. A few SNPs may slip through the cracks.

        Comment


        • #5
          Wow! It amazes me that we can get Y reads from just one side of the helix, while for autosomal reads we can't even separate calls for one chromosome of a pair from calls for the other of the pair. Maybe there will some day be a $500 version of FamilyFinder that can do that.

          Comment


          • #6
            Originally posted by Marinco View Post
            Wow! It amazes me that we can get Y reads from just one side of the helix, while for autosomal reads we can't even separate calls for one chromosome of a pair from calls for the other of the pair. Maybe there will some day be a $500 version of FamilyFinder that can do that.
            That's because there's just one version of the Y chromosome, as opposed to two versions of every autosome. The DNA is chopped into small pieces before analysis. All those pieces are just floating around in the soup. The best way now to separate the calls is to test a father/mother/child trio. Then you can assign the child's pieces to one parent or the other (phasing). Even that isn't perfect -- typically about 5% of the SNPs can't be phased, when all three parties are heterozygous for the SNP. Statistical phasing is another option, based on population frequencies of how adjacent SNPs often travel together, but it's less satisfactory.

            Comment


            • #7
              Thanks Ann.

              So it's impossible to separate a pair of autosomes and analyze each separately, or is it just not done that way?

              The heterozygousity issue (any two people will match at any location where one or the other is heterozygous) that you point out is exactly why I think separating the two would be so valuable. Surely, it would really cut down on the number of false (IBS) matches.

              Comment


              • #8
                Originally posted by Marinco View Post
                Thanks Ann.

                So it's impossible to separate a pair of autosomes and analyze each separately, or is it just not done that way?

                The heterozygousity issue (any two people will match at any location where one or the other is heterozygous) that you point out is exactly why I think separating the two would be so valuable. Surely, it would really cut down on the number of false (IBS) matches.
                Not impossible -- it's been done in a research environment -- but not practical at the present time.

                You are correct about the big reduction in IBS matches if you have phased data. At GEDmatch my son's match list breaks down this way

                unphased: 61,478 (18,125 >= 7 cM)
                maternal haplotype: 3915 (2037 >= 7 cM)
                paternal haplotype: 4867 (2434 >= 7 cM)

                Comment

                Working...
                X