DNA testing offers a far better way to detect Down syndrome

The results of a new study offer very good news for expectant parents: a remarkably accurate prenatal genetic test for Down syndrome that requires only a blood sample from the mother.

Expectant parents have a lot to deal with. What can a mom-to-be eat, drink, or do to ensure the health of the baby? There is plenty of advice out there, both good and bad, but one area that has advanced rapidly in recent years is prenatal genetic testing. We now have tests that can identify a small but growing number of genetic disorders early in pregnancy.

Until recently, the only way to test the genetic makeup of a fetus was through amniocentesis ("amnio"), an invasive procedure in which a doctor inserts a long needly directly into the womb and collects a sample of amniotic fluid. This test is not only uncomfortable, but it also carries a small risk of miscarriage. Fortunately, it may soon become completely unnecessary, thanks in part to advances in DNA sequencing.

It turns out that small amounts of fetal DNA are circulating in the mother's blood from a very early point in pregnancy, just 10 weeks along. New technology allows us to take a simple blood sample from the mother and use that to examine the baby's DNA to look for trisomy–the presence of 3 copies of a chromosome rather than two. The new study primarily looked at trisomy 21, which causes Down syndrome, but it also looked at the ability of DNA screening to detect other trisomies.*

The new study by Mary Norton and colleagues, by far the largest of its kind to date, involved almost 19,000 women at 35 locations in 6 different countries, all of whom were undergoing routine prenatal screening. The study compared the standard blood test, the "triple screen", to a new cell-free DNA sequencing (cfDNA) test from Ariosa Diagnostics (recently acquired by Roche). I wrote about a similar study involving 1914 patients last year, which used the same kind of DNA testing. The new study is ten times larger, and the results are even better. All of the participants had both standard screening and DNA testing.

What did they find? First, the number of mothers who had both standard and cfDNA testing was 15,841. (For technical reasons, some of the 19,000 participants didn't get both tests.) Out of all pregnancies, there were a total of 38 fetuses with trisomy 21. Standard screening detected 30 out of 38, while cfDNA detected all 38. Much more impressive–startling, really–was the difference in false positives. The standard screen had 854 false positives, while cfDNA testing only had 9. This is nearly 100 times better.

Or to put it another way: among all the mothers, standard screening reported 884 "positive" results, of which only 30 were correct. Thus if you were a mom who got the standard screen, and you had a positive result, there was only a 3.4% chance that your baby would have Down syndrome. Nearly all of these mothers would probably elect amniocentesis to confirm, and go through the anxiety as well as risk of miscarriage that amnio entails.

The cfDNA test reported only 47 positive results, of which 38 were correct. Thus there was an 81% chance (38/47) that the baby will a positive result would have Down syndrome.

Although the numbers were smaller, cfDNA testing was also far more accurate for detecting trisomy 13 and 18, rarer conditions that cause health problems so severe that most infants die before reaching their first birthday.

With a false positive rate nearly 100 times lower than the standard blood test (9 versus 854 false positives), the superiority of DNA testing for prenatal screening is clear. Let's hope that we can soon replace the older blood tests and spare parents-to-be the anxiety and unnecessary follow-up testing caused by thousands of false positive results under the current standard of care.

*For those who want to understand the technical details, this paper by Sparks et al. (2012) explains the sequencing and statistical methods.




How disruptive are MOOCs? Hopkins launches new genomic data science series.

If you could visit a college classroom a hundred years from now, would you expect it to look just like one of today's classrooms?

I suspect not.

Yet if you walked into almost any college classroom today, you’d see a scene right out of the 19th century: students sitting in a classroom listening to a professor talk. Perhaps the professor is using a laptop to project slides, rather than writing on a chalkboard, but other signs of the 21st century can be hard to find. 

Of course the content today is different, particularly in the sciences–no one even knew what DNA was 200 years ago–but the way we teach has barely changed in the past two centuries.

Well, it’s changing now. The advent of massive, online open courses (MOOCs) is making high-quality content available to millions of people for the first time, and much of it is free. These online courses, even in their infancy, have generated a huge response, revealing the hunger out there for learning. 

Professors too are discovering something: how rewarding it is to reach thousands of students rather than just a handful. And capturing lectures on video allows us to mix and match material, updating just the parts that need changing and re-using the good stuff with relatively little effort. 

Hopkins joined the MOOC revolution last year, with a Data Science series of courses led by Brian Caffo, Jeff Leek, and Roger Peng. This June we’re launching a new specialization on genomic data science, that five of my colleagues and I are teaching. Coursera announced our 6-course specialization just a few days ago; Jeff Leek and Roger Peng produced a teaser video that you can watch here:


The specialization will cover statistics, Python programming, the Bioconductor and Galaxy environments, and more, all designed as a gentle introduction to genomic data science for students, postdocs, or even established scientists who want to start learning the tools behind genomics and “precision medicine.” 

Most of the videos are “in the can” and we’re doing the post-production now, creating the quizzes and projects that go along with the classes. Students can sign up now, but they'll have many chances to take the series. Once it launches, the classes will run every month, forever. (Okay, maybe not forever, but for a while.)

One of the first questions people ask me about MOOCs is, how can you offer all this content for free? Isn’t this a threat to the traditional university model?

You bet it is. But it’s happening, whether universities like it or not. As with other disruptive technologies, MOOCs are likely to change dramatically over the next few years, but they aren’t going away.

We're just learning how to deliver knowledge through MOOCs and we’ll get better at it. Right now MOOCs exist in a separate world from universities, but that’s likely to change. Maybe the experiment will fail, and MOOCs will fade away. Maybe that classroom in the year 2115 will look just like my classroom today. I’m betting it won't.