The Genetic Algorithm That Revealed My Possible Babies

Seeing a thousand of my potential progeny gave me a glimpse of new things that future parents will stress over.

“It is likely that within a few decades, people will look back on our current circumstance with a sense of disbelief that we screened for so few conditions. They will also be puzzled and dismayed, as I am now, that our healthcare system put so many couples in an unnecessarily difficult position, by not identifying their carrier status until a pregnancy was already underway.” — Francis Collins, Director of the National Institutes of Health

“Striving to better, oft we mar what’s well.” — William Shakespeare, King Lear


Lee Silver, the co-founder and chief science officer of GenePeeks, is tough to pin down. He admits he has severe ADD that makes it hard for him to focus for hours or even minutes. He certainly doesn’t like to stick with a job for more than a few years. “I get bored with things,” he says. GenePeeks arose during one of Silver’s flights of boredom. But the idea at its core is so big that he’s still at it six years later.

GenePeeks is attempting to prevent inherited diseases in future babies. By using an algorithm that mashes up the gene sequences of any two biological parents, the company will conceive your “virtual progeny.” Your VP don’t wear diapers or cry. They are just a data simulation of 1,000 potential genetic combinations.

Silver says this analysis can predict whether two people will produce a healthy child. The company calls the test, which has a price tag of nearly $2,000, the next step in genetic carrier testing, a way to give parents peace of mind before they conceive. The company’s marketing slogan is “Protecting our children is in our DNA.”

If “virtual progeny” sounds like something straight out of a dystopian imagination, it is. One of Silver’s many distractions, which have included jumping from teaching microbiology to international affairs at Princeton, collaborating on an off-off-Broadway romantic comedy about a graduate student who impregnates herself with chimp sperm (the New York Times called it “smugly silly”), was his 1997 book Remaking Eden: How Genetic Engineering and Cloning Will Transform the American Family. In Silver’s projected future, parents begin to choose embryos that are genetically healthier. This eventually leads to a new kind of class divide between what he calls the “gen-rich,” those who have the money to engineer their children, and the “naturals,” who don’t.

Soon after his theatrical flop, a mutual friend introduced Silver to Anne Morriss, a Harvard MBA who was fishing around for her next gig. Morriss had recently gone through a frightening experience with the birth of her son, conceived with donor sperm from a sperm bank. The boy was born with a single-gene recessive disease called MCAD deficiency, which meant that he didn’t make enough of an enzyme that is responsible for efficiently converting fat into sugar. If the disease wasn’t properly managed, he might have violent seizures if his blood sugar got too low, difficulty breathing and liver problems — and he might even die suddenly.

Most sperm banks, which are known for meeting only the bare minimum of regulations, don’t run carrier testing on their donors, let alone sequence their donors’ genomes. Most rely on donors truthfully disclosing their personal and family medical histories in interviews.

Lee Silver (Courtesy of GenePeeks)

So Morriss had no idea she and her donor carried the recessive gene for MCAD. Even if she had screened herself and the donor, MCAD is not included in most recessive disease tests. Luckily, her doctors caught the disease during newborn screening in the hospital, so from the start she and her partner could carefully manage it. Eventually they would learn that their son would most likely grow out of his particular case of MCAD.

The experience made her realize that many of these cases aren’t caught. Genetic disease is the leading cause of infant death in the U.S., accounting for 20 percent of annual infant mortality. While carrier testing has been happening since the 1970s among certain ethnic groups with high risk, universal testing has been cost prohibitive because it hasn’t been covered by health insurance.

It was only this year that the American Congress of Obstetricians and Gynecologists started recommending screening of all pregnant women, regardless of their ethnicity. One of the reasons for this new recommendation is that genome sequencing has shown some recessive diseases typically associated with certain ethnic groups are actually more common than we previously thought. For example, the mutation that causes Tay-Sachs disease, which has been associated with Ashkenazi Jews, also has been found in people of Irish descent.

Such discoveries have pushed the medical genetics community to suggest that genetic screening move beyond pregnant women. “The big question is: Why are we screening people when they’re pregnant?” asks Dr. Ronald Wapner, director of reproductive genetics at Columbia University Medical Center. “This is something the medical community should be doing before someone is pregnant. We should all be screened early in life.”

It’s this idea of pushing screening earlier that got Silver and Morriss talking. Today most genetic screening happens after conception, through an invasive test such as amniocentesis, if at all. And the carrier screening that does happen pre-conception usually doesn’t look for many of the less-common recessive diseases. So Silver and Morriss decided to work together to set a new standard for preventive medicine, by analyzing the combination of any two genomes to test for a wide range of diseases. “We want to bring all this science to the pre-conception moment and interrupt the transmission of risk,” Morriss told me.

Our desire to protect our children and our hope that they will be an improvement on us are innate. In today’s helicopter parenting culture, these desires can sometimes get amped up into a maddening drive toward perfectionism. It may be part of what many scientists and thinkers are defining as a new stage of human-driven evolution. Our scientific inventions, as author Yuval Noah Harari writes in his book Homo Deus, are creating the potential to “upgrade us into Gods.”

Nowhere is this more evident than in reproductive science. We can separate sex and procreation. We can choose sperm donors, egg donors and the technologies to help us conceive. This unprecedented control, in everything from in-vitro fertilization to preimplantation genetic testing, and pretty soon, gene editing, means that instead of spinning the reproductive roulette wheel, we are forced to ask ourselves, “What am I choosing for my child?” (Neuroscientist David Eagleman expands on this idea in this article for Neo.life.)

GenePeeks’ technology will help people with either a known or unknown history of recessive disease gain more control over their risk — and potentially escape their evolutionary fates. One question, however, is whether this opportunity is also sending us straight toward Silver’s dystopian vision, in which only those who can afford these new choices get to benefit. And at what point does our drive for perfection become unhealthy? I decided to go through the GenePeeks test to see for myself.

The website of GenePeeks plays perfectly to the anxious every-parent. The word “peeks” in the logo is a rainbow of colors, saying, in effect, “We’re open to any kind of family.” The company calls biological parents “participant parents” so as not to alienate single parents or families using donor eggs or sperm. A woman of ambiguous ethnicity smiles and holds a baby. An infographic demonstrates the steps a customer will go through. It all leads to a drawing of a computer screen displaying lines of diapered baby symbols.

Rebecca Silver, Lee’s daughter, is the company’s director of client experience. Over the phone, she walks me through the online registration process. I explain to her that five years ago, when I was conceiving my son with a sperm donor as a single mom by choice, I learned from a blood test that I carried the recessive mutation for Canavan disease, most frequently found in Ashkenazi Jews like me. The disease is a neurological birth disorder caused by a gene mutation that affects the nerve fibers of the brain. When both parents carry the mutation, there is a 1 in 4 chance that their child will get two copies of the mutation and thus inherit the disease. Generally the child doesn’t live past the age of four.

I’m a cautious optimist, anxious by nature. I feel more comfortable worrying about the worst-case scenario and then being pleasantly surprised when it doesn’t come to pass. Because I knew about the Canavan mutation, I chose not to pair myself with a Jewish donor when I was trying to get pregnant. But that was the extent of my decision. I knew a lot about my donor from his profile: he was 5-foot-10 with fair, rosy skin, light brown eyes, and blond wavy hair. He aced college tests while still in high school. His grandfather skied until the day he died at the age of 97. My donor could tie multiple kinds of neckties, and had a penchant for aphorism. But I didn’t know whether he carried any recessive diseases, including Canavan, because the sperm bank I used didn’t do those tests.

Without a definitive answer that he wasn’t also a carrier, the “what ifs” about Canavan lingered in the early weeks of my pregnancy. Because I was 40, and my chances for several genetic aberrations were higher, I chose at 11 weeks to undergo a test known as CVS (an acronym for chorionic villus sampling), in which cells are taken from the placenta and analyzed. That ruled out Canavan disease, Down syndrome, cystic fibrosis and Fragile X syndrome.

That was a relief because I knew that if he had any of those diseases, I would have terminated the pregnancy. I know this is not everyone’s choice, and that’s why some people don’t do any tests at all. They leave it up to fate or their God, and say they will accept what child they get on the day he or she is born.

Alexander is now almost five and I did get lucky — he’s healthy. His mix of genes gave him blond hair and fair skin like his donor, and my mouth, both in shape and tendency to talk a lot. I’m not planning to get pregnant again, but I was curious how I would feel and what more I might learn going through the GenePeeks process. Would I discover something lurking in my genes?

The test, which requires a doctor’s prescription and consultation with a genetic counselor, is not covered by insurance and therefore most likely out of reach for many families. For the sake of my experiment, GenePeeks covered the cost.

Silver tells me that only about 5 percent of the company’s customers learn that their virtual progeny carry a genetic risk. With such low odds, I wondered if it was worth the money. While the test is marketed to ease anxiety for new parents, especially those with a known history of genetic disease, it also could have the opposite effect by raising more questions and new choices about a child’s future.

“We look for diseases that patients would reasonably want to know about before pregnancy with the potential to intervene if risk is found,” says Regine Lim, a genetic counselor who works for GenePeeks. “On the milder end of the list, we have genes that cause hearing loss or vision loss. On the more severe end of the list there are genes that can impact fetal development and lead to pregnancy loss.”

It would be challenging to put my son’s actual donor through the test since he’s retired from donating. So Rebecca Silver matches me with a donor from one of the sperm banks that GenePeeks works with. He’s of Canadian and Irish decent with brown hair and brown eyes. “He’s a motivated go-getter and really confident in who he is, and very talented when it comes to making things happen,” she writes me in an email.

My fertility doctor approves an online prescription, and then Rebecca Silver sends me a spit test. The box arrives by FedEx the next day. It’s elegantly packaged with the same rainbow branding of the GenePeeks website. Simple instructions walk me through a process that takes all of two minutes. Spit in test tube, swirl test tube with a stabilizing liquid that keeps bacteria out, seal the tube, slip it in a box with a self-addressed envelope and stick it in the nearest mailbox.

When the box arrives at the lab, technicians will put my saliva sample through a machine and analysis platform made by the company Illumina that looks at my exome, the parts of the genome that encode proteins. Recessive genes like the one involved in Canavan would turn up here as variants of what is found in a typical exome.

(Baby photo by Lana K/Shutterstock; illustrations by Nick Vokey)

GenePeeks takes carrier testing to a new level. Today it looks for over 900 recessive genetic variations that correspond to over 1,000 diseases. For now, most prospective parents don’t get screened at all, but even those who do usually aren’t having that many genes analyzed. The American Congress of Obstetricians and Gynecologists says that screening up to several hundred genes, including the ones implicated in cystic fibrosis and spinal muscular atrophy, is “an acceptable strategy.” And that recommendation doesn’t include screening dads or sperm. That’s why GenePeeks’ analysis pushes genetic testing, and possibly parental anxiety, into a different realm.

Everyone with whom I spoke at the company referred to the “mathematical algorithm” that simulates genomes of my virtual kids. To understand what the algorithm does, I asked Lee Silver: why would it be any better than simply looking at both parents’ genes for any recessive mutations that are implicated in diseases? After all, basic genetic science tells us that if I conceived a child with a partner who also carried one copy of the Canavan gene, our child would have a 25 percent chance of having the disease.

Silver’s ADD takes us on a wild tangent that meanders across the history of our understanding of DNA, how he never believed that cloning was possible, but then when Dolly the sheep was born, he realized biology had fewer hard-and-fast rules than he thought. And one of the assumptions he now thinks should be overturned is the idea that variations in genes are entirely binary: either benign or the cause of disease. With many variations, the truth can be somewhere in the middle. Sometimes a mutation causes the body to produce too little of a key protein. But whether that protein shortfall causes a disease can depend on multiple factors, including the other mutations in the genome.

A 2012 study led by scientists at Wellcome Trust Sanger Institute found that every person has an average of 400 genetic defects — many of which are not causing any problems. (On top of that, people also have “epigenetic” variations, which are changes in how and when genes are activated, or “expressed.”) Because some inherited mutations cause disease in children only in combination with certain errors in a second copy of the gene, Silver thought the carrier screening paradigm was often too simplistic. It wasn’t enough to just look at one parent’s gene sequence and then the other, and compare whether they had the same commonly studied variants.

Instead GenePeeks uses the Monte Carlo method, a probability model, to look at 1,000 potential combinations of the two parents’ exomes. The subtleties of these combinations, Silver says, will be instructive.

He offers an example. Typical carrier screening showed that two parents who each carried a mutation in the BTD gene were likely to have a baby with a metabolic disease called biotinidase deficiency. People with this disease are unable to recycle biotin, one of the B vitamins. The GenePeeks algorithm showed, however, that because of the specific BTD variations that each parent had, their virtual progeny actually would have made more than 50 percent of the normal level of the protein—enough to avoid the disease. “Variants in a gene are not black and white,” Silver says. “The real world is a continuum.”

For single-gene recessive diseases like Canavan, GenePeeks could sort things out without creating 1,000 virtual progeny. But the company is preparing for a future in which it can sniff out more complex disorders, such as autism, that result from mutations in more than one gene. GenePeeks will need more combinations in order to predict the likelihood of those problems.

For now, it is focusing on proving that its simulations yield more precise predictions in single-gene recessive diseases. Last year, GenePeeks conducted a study with Reproductive Medicine Associates, a fertility clinic in New York. They compared conventional carrier screening of 308 pairings of egg donors and male participant parents to a GenePeeks analysis of the same people. The study, which has not yet been published, reported that traditional testing found two sets of donors and recipients at risk and GenePeeks’ algorithm found 11. For example, in one of the 308 pairings, conventional carrier screening found one parent to be a carrier for Smith-Lemli-Opitz syndrome, a severe developmental disorder. That parent carried the most common damaging variant, known as an allele, of the gene implicated in the disorder. But since nothing problematic was found in the other parent, the match was not flagged as being at risk of producing a kid with Smith-Lemli-Opitz.

However, in that other parent, the GenePeeks algorithm flagged a variant that had never been characterized in the public literature, what’s known as a “variant of unknown significance.” Another way of saying this is that the algorithm errs on the side of “better safe than sorry.” The variant “may not be a smoking gun that it’s going to be damaging,” Lim acknowledges.

The fact that the GenePeeks’ algorithm is detecting new and potentially borderline variants is probably good for the future of scientific understanding. But it may not be good for anxious, perfectionist parents. I now wondered even more if the GenePeeks test would pick up something else in me, something of “unknown significance” related to Canavan disease.

Wapner of Columbia’s Reproductive Genetics department told me that my increased worry is exactly what makes the test nowhere near ready for widespread clinical application. (GenePeeks won’t disclose exactly how many clinics are currently using the algorithm or how many patients have taken the test.) The reason, he says, is that because most of these new variants are very rare and no one really knows what they mean. “I think as a scientific endeavor, it’s fantastic,” he says. “There’s a large argument, however, that variants of unknown significance shouldn’t be reported to a family or used in counseling because there’s no definitive answer.”

Silver counters that even variants deemed to be of unknown significance are not necessarily complete unknowns. GenePeeks’ algorithm relies on data from “decades of research by molecular biologists who have developed tools to look at new variants and predict their effects,” he says. “But the clinical community has not yet accepted those.”

A few weeks after my test, an email arrives, saying my analysis is complete. I click through to my report, and read the comforting words “Your future child does not have an increased risk of inheriting disease.” Then the fine print: “As with any genetic screening method, these results do not guarantee the birth of a healthy child.”

Clicking through to the analysis, I half expect to see a digital rendition of my virtual progeny, but instead I read an arduously long spreadsheet of genes. Across the top row are columns with names such as Number of Bases Covered, Global Clinical Sensitivity, Global Clinical Specificity and Negative Predictive Value. It makes sense that this requires interpretation by a genetic counselor.

I call Regine Lim to walk me through my results. She explains that the spreadsheet reveals such things as how large each gene is, how common each of my variants is, and the likelihood that there are false positives in the results. It all starts to feel like a lot of information that just prompts more questions. I wonder whether this might produce the same anxiety, or even more, than just rolling the dice as I did with my son, with the minimally recommended testing.

If this test had been for real and it had uncovered Canavan disease — or a variant of unknown significance — in my donor, that information would change things. I would simply pick another donor.

A married couple can’t exactly switch it up like that. Lim emails me an example report of an anonymous couple whose future child was at risk because of a mutation on a gene known as SMPD1. If matched with another mutated copy of the gene from the other parent, it can result in Niemann Pick Disease. NPD is characterized by a buildup of fat and cholesterol. It can develop in infancy or adulthood, and the person usually dies within 10 years. The challenge with this disease, Lim explains, is that can be very severe or very mild. I ask her how she handled the diagnosis with the parents. She explains that there’s quite a lot of evidence to suggest that the mutations identified in their virtual progeny were very likely to cause disease. “But it’s difficult to identify if it’s actually going to be the more severe or the milder of the two,” she told them.

Knowing this information before conception, however, will give this couple more choices. They can spin the roulette wheel and prepare for a child with the disease. Or there is the more expensive choice with something close to a guaranteed healthy outcome: they could conceive using in-vitro fertilization, genetically test the embryos, and then choose to implant one that doesn’t carry the mutation.

This spectrum of choices will widen with other advanced reproductive technologies, and that’s where GenePeeks could start to go beyond preventative medicine and begin to look like Silver’s vision of genetic classes, with a divide between those who can afford to have healthier and otherwise enhanced children and those who can’t. At some point a GenePeeks virtual progeny might show less harmful mutations — say, for something like depression or ADD. There could be a scenario in which a genetic counselor offers gene editing in the mix of options, giving those who can afford it a way to select traits they believe will make their children “better.”

“What will happen is that we’ll start with eliminating cystic fibrosis and Tay-Sachs disease, and when the technology becomes available we’ll start choosing traits that having nothing to do with medicine or disease,” says Marcy Darnovsky, the executive director of the Center for Genetics and Society.

Maybe for those with a history in their family, the choice will be clear: to change their genetic fate. Maybe for the ever-worried, it will create a calming sense of control. I did feel better knowing for sure my theoretical donor didn’t carry Canavan disease, and that we would produce a child with no recessive disease. Paradoxically, it might also have the opposite effect, especially in the case of variants of unknown significance.

Some families may choose to pay for a genetic nip and tuck in order to conceive their idea of the more perfect child. In the end, these enhancements might bleach away the natural inconsistencies and genetic nuances that define humanity. One choice could erase a gene that might have produced an ADD-afflicted but wildly brilliant polymath professor and entrepreneur.

Then again, in some key ways we might end up no different than we have always been. Everyone will make decisions about genetic enhancement according to their own values, and some people, rich or poor, enhanced or disabled, will become extraordinary and others will not. We all carry mutations and there’s no such thing as the perfect baby.