> "Foetal DNA makes up about 10% of a pregnant woman's blood plasma." <- I'd be quite surprised to learn that DNA makes up 10% of anyone's plasma, so I think they're saying that 10% of the DNA isolated from plasma is found to be fetal.
> "The latest research isolated foetal genetic signatures in the floating DNA, then compared its characteristics against the genetic maps of the mother and father. That way, scientists were able to construct a genome-wide genetic map of the foetus, which they could then scan for variations and mutations."
Update: OK, I was a bit confused by the DNA extraction process. From what I can tell, they only took DNA from the parents once, at ~12 weeks gestation. They then GWASsed them and found sites where the fetus would be an obligate heterozygote because the mother was homozygous for one allele while the father was homozygous for another. (So if mom is AA and dad is GG, the fetus must be AG, assuming no point mutation occurs there.)
But if the blood is only drawn once, 10% of the maternal DNA will be fetal DNA, so how are they confidently calling sites as homozygous in the mother? I suppose this makes sense if Birdseed is robust to 5% error in the data, and if so this is quite clever. (5% because 10% of the DNA is fetal, and at the obligate het sites half of that 10% will be maternal so only 5% total will be of paternal origin.)
Now, their method is for high-depth whole-genome sequencing, which is still very expensive, but there's no reason they couldn't use targeted sequencing with this approach (they address this in the discussion). That would allow for a potentially inexpensive test for inherited genetic disorders without invading the fetal privileged space (which is what we currently do, and which is high-risk).
The hard part is figuring out the maternally inherited foetal DNA, because it's already present in the mother's bloodstream. For features which are present in the mother but not the father, it's fairly easy with the data they have. For vice versa, it's probably not going to be possible for some time.
Agreed, mostly, at least from first principles. The enrichment will be very modest. On average for any given chromosome, 45% of DNA in the mother's blood should be maternal DNA from one chromosomal copy or the other. This gets us to 90%.
The remaining 10% is split between one paternal copy and one maternal copy.
This means that--if we ignore other differences such as fragment length--you're potentially trying to label the chromosomal copy that comprises 50% of DNA as the one inherited by the fetus, versus the one comprising 45% of DNA as the one not inherited by the fetus. That's a very small enrichment.
Now, because of the length differences that they identified, they may be able to vastly enrich for fetal DNA (or precipitate out maternal DNA), obviating this problem.
> without invading the fetal privileged space (which is what we currently do, and which is high-risk). [emphasis mine]
The standard process is amniocentesis, and according to wikipedia, that has a miscarriage rate "as low as 1 in 1600", plus an unspecified risk of infection. While that's almost certainly higher than the risk due to drawing the mother's blood, I wouldn't call it a "high-risk" procedure.
Would you feel the same way if the risk of death from a diagnostic test for cancer was 1/1600? The answer probably hinges upon the type of cancer, the treatment available, etc, but clearly this is not something you'd use for a population screening tool.
Up until ~4 years ago, the risk of miscarriage from amniocentesis was thought to be 1/200, but if the 1/1600 number is true, that's good. Nevertheless, it's not something that you're going to want to be doing to every pregnant woman. On the other hand, we draw plenty of blood from pregnant women, so we can essentially get this without additional risk to the fetus, except the risk of elective termination.
> "The latest research isolated foetal genetic signatures in the floating DNA, then compared its characteristics against the genetic maps of the mother and father. That way, scientists were able to construct a genome-wide genetic map of the foetus, which they could then scan for variations and mutations."
Now I'm interested to read more. The article can be found behind the following paywall: http://stm.sciencemag.org/content/2/61/61ra91.abstract .
Update: OK, I was a bit confused by the DNA extraction process. From what I can tell, they only took DNA from the parents once, at ~12 weeks gestation. They then GWASsed them and found sites where the fetus would be an obligate heterozygote because the mother was homozygous for one allele while the father was homozygous for another. (So if mom is AA and dad is GG, the fetus must be AG, assuming no point mutation occurs there.)
But if the blood is only drawn once, 10% of the maternal DNA will be fetal DNA, so how are they confidently calling sites as homozygous in the mother? I suppose this makes sense if Birdseed is robust to 5% error in the data, and if so this is quite clever. (5% because 10% of the DNA is fetal, and at the obligate het sites half of that 10% will be maternal so only 5% total will be of paternal origin.)
Now, their method is for high-depth whole-genome sequencing, which is still very expensive, but there's no reason they couldn't use targeted sequencing with this approach (they address this in the discussion). That would allow for a potentially inexpensive test for inherited genetic disorders without invading the fetal privileged space (which is what we currently do, and which is high-risk).