6 ms·

99.6% is still awfully dissimilar, depending on what that number means. Thinking about single polymorphisms, of ca. 3 billion bases, that is still 120,000,000 S

99.6% is still awfully dissimilar, depending on what that number means. Thinking about single polymorphisms, of ca. 3 billion bases, that is still 120,000,000 SNPs between twins. Of course, they aren't doing full genome sequencing, and they are probably targeting loci that are know to be more divergent. Still... that similarity seems awfully low for twins. I wonder what they typically find for similarity between any two individuals. I also wonder if their lab is a little disorganized....

klmr 8y ago
Current estimates are that there’s a roughly 1 in 1000 difference between two individuals purely due to random mutations in the germline (or, put differently, the DNA replication error rate in human cells is about 1 in 1000). This would imply 99.9% expected similarity between two identical twins. However, this is in the germline. Somatic cells accumulate mutations over their life time (since they, too, divide and therefore need to replicate their DNA, and are furthermore exposed to environmental stress). In addition, the tests work off saliva samples which are mixed with all kinds of stuff in the mouth, and degrade slightly during transport. Lastly, there’s an inherent replication variability in the genotyping process (though this is very small). This back of the envelope estimate suggests that 99.6% similarity for somatic samples of identical twins is not unexpected.
- opportune 8y ago
  Something tells me that the 1/1000 error rate you cite isn't truly accurate, or at least doesn't apply to all portions of DNA equally, because in almost every individual that would result in many SNPs with bad consequences since we have so many subsequences coding essential proteins
- adenadel 8y ago
  This is not accurate. Coarsely, there are roughly 3 million SNPs (out of 3 billion total haploid bases) per individual. When we talk about individuals "sharing" DNA we mean these variants and not the 99.X% of DNA that we all have in common. When we refer to random mutations in the germline (i.e. de novo variants) current estimates of de novo variants per individual are in the low hundreds. The other factors you discuss are more likely to cause issues. One major one is the error rate of the microarray because of the high number of genotypes being assayed. If the chip has an error rate of 0.1% and 500k tests are performed you would expect 500 mis-calls. I used to work for Illumina, who makes these chips, but I worked on the DNA sequencing side of things so I don't know the actual error rate off the top of my head.
  - klmr 8y ago
    Right, de novo mutation rate is much lower (even lower than what you wrote, just slightly above 1e-8 according to deCODE). The 1/1000 rate is the average difference between two random individuals (not twins), which corresponds to your 3M SNPs per 3B bases.
    - adenadel 8y ago
      Yes, but you referred to them as random mutations, which is not the case. They are inherited variants. The DNA replication error rate is nowhere near 1 in 1000.
      - klmr 8y ago
        Hence my correction. Without looking it up now, the actual replication error rate is probably close to the deCODE rate for de novo mutations (modulo gamete selection, so probably higher than that).
SubiculumCode 8y ago
Yeah Im a bit surprised about that, even given de novo mutations.
adenadel 8y ago
It represents the underlying error rate of the assay.
- glup 8y ago
  Which seems pretty reasonable for a direct-to-consumer, cheap test, right? Noisy sensors = variability in reports.
  - adenadel 8y ago
    It's totally reasonable (and in line with my knowledge of a similar technology, DNA sequencing). Often the stochasticity comes from biomolecular processes rather than from the optics of the instrument.
pesfandiar 8y ago
Yes, especially since we estimate humans and apes share near 99% of their DNA sequence.
- klmr 8y ago
  They don’t, this number is extremely misleading. Roughly 97% (I think is the current estimate) of their protein-coding DNA is equivalent, but not 1:1 identical.
- Tor3 8y ago
  Not the same kind of 99%.. For (fraternal) siblings you'll get around 50% or thereabouts.
  - klmr 8y ago
    No, that’s also incorrect. The real number is way higher. They share 50% of alleles, and hence 50% of variants. But a lot of the DNA within two distinct alleles will be identical, except for (say) a single nucleotide difference.
    - Tor3 8y ago
      Yes, that's actually what I meant to say.. couldn't remember the right terminology at the time. The point stands though, I think.. the number "99.6%" for identical twins isn't measuring the same as "98.5%" (or so) for humans vs chimps.
peter303 8y ago
Article said identical twins. So I'd expect around a couple thousand SNP differences. Each cell devision has about 30 mutations. Embroyo bifurcated within a dozen or so cell divisions. Dr. Hayflict found human soma cells live aboutb50 divisions.

klmr 8y ago

Current estimates are that there’s a roughly 1 in 1000 difference between two individuals purely due to random mutations in the germline (or, put differently, the DNA replication error rate in human cells is about 1 in 1000). This would imply 99.9% expected similarity between two identical twins. However, this is in the germline. Somatic cells accumulate mutations over their life time (since they, too, divide and therefore need to replicate their DNA, and are furthermore exposed to environmental stress). In addition, the tests work off saliva samples which are mixed with all kinds of stuff in the mouth, and degrade slightly during transport. Lastly, there’s an inherent replication variability in the genotyping process (though this is very small). This back of the envelope estimate suggests that 99.6% similarity for somatic samples of identical twins is not unexpected.

opportune 8y ago
Something tells me that the 1/1000 error rate you cite isn't truly accurate, or at least doesn't apply to all portions of DNA equally, because in almost every individual that would result in many SNPs with bad consequences since we have so many subsequences coding essential proteins
adenadel 8y ago
This is not accurate. Coarsely, there are roughly 3 million SNPs (out of 3 billion total haploid bases) per individual. When we talk about individuals "sharing" DNA we mean these variants and not the 99.X% of DNA that we all have in common. When we refer to random mutations in the germline (i.e. de novo variants) current estimates of de novo variants per individual are in the low hundreds. The other factors you discuss are more likely to cause issues. One major one is the error rate of the microarray because of the high number of genotypes being assayed. If the chip has an error rate of 0.1% and 500k tests are performed you would expect 500 mis-calls. I used to work for Illumina, who makes these chips, but I worked on the DNA sequencing side of things so I don't know the actual error rate off the top of my head.
- klmr 8y ago
  Right, de novo mutation rate is much lower (even lower than what you wrote, just slightly above 1e-8 according to deCODE). The 1/1000 rate is the average difference between two random individuals (not twins), which corresponds to your 3M SNPs per 3B bases.
  - adenadel 8y ago
    Yes, but you referred to them as random mutations, which is not the case. They are inherited variants. The DNA replication error rate is nowhere near 1 in 1000.
    - klmr 8y ago
      Hence my correction. Without looking it up now, the actual replication error rate is probably close to the deCODE rate for de novo mutations (modulo gamete selection, so probably higher than that).

SubiculumCode 8y ago

Yeah Im a bit surprised about that, even given de novo mutations.

adenadel 8y ago

It represents the underlying error rate of the assay.

glup 8y ago
Which seems pretty reasonable for a direct-to-consumer, cheap test, right? Noisy sensors = variability in reports.
- adenadel 8y ago
  It's totally reasonable (and in line with my knowledge of a similar technology, DNA sequencing). Often the stochasticity comes from biomolecular processes rather than from the optics of the instrument.

pesfandiar 8y ago

Yes, especially since we estimate humans and apes share near 99% of their DNA sequence.

klmr 8y ago
They don’t, this number is extremely misleading. Roughly 97% (I think is the current estimate) of their protein-coding DNA is equivalent, but not 1:1 identical.
Tor3 8y ago
Not the same kind of 99%.. For (fraternal) siblings you'll get around 50% or thereabouts.
- klmr 8y ago
  No, that’s also incorrect. The real number is way higher. They share 50% of alleles, and hence 50% of variants. But a lot of the DNA within two distinct alleles will be identical, except for (say) a single nucleotide difference.
  - Tor3 8y ago
    Yes, that's actually what I meant to say.. couldn't remember the right terminology at the time. The point stands though, I think.. the number "99.6%" for identical twins isn't measuring the same as "98.5%" (or so) for humans vs chimps.

peter303 8y ago

Article said identical twins. So I'd expect around a couple thousand SNP differences. Each cell devision has about 30 mutations. Embroyo bifurcated within a dozen or so cell divisions. Dr. Hayflict found human soma cells live aboutb50 divisions.