Fyodor URNOV: I got an email from a parent of a child who is going progressively blind. This parent writes to me and says, “Look, my child is an amazing student and she’s doing so well, but she’s going to lose vision. It’s going to happen, we don’t know when. Can you stop it?”
Every day, Dr. Fyodor Urnov’s inbox is flooded with emails like that one. And they put him in a difficult position. Because he has the technology to stop it. Fyodor is a professor at U.C. Berkeley and a leading researcher in the field of genomic therapies. He develops medications that change people’s D.N.A. to cure genetic diseases like the one described in that email.
URNOV: We are not just sitting here handwringing at the fault that’s in our stars; we can actually fly to the stars and touch them and manipulate them.
But having the ability to do something and actually doing it are two very different things. Medicine has always suffered from a problem called the “know-do” gap. It’s the difference between what we actually do for patients and what we could do, given all that we know. Breakthroughs in biomedicine are allowing doctors to do things they could never do before. But sometimes these advances don’t fit into our financial or regulatory systems. That means it can take a long time for patients to actually benefit — time that many of them don’t have to spare. The National Institutes of Health invests more than $40 billion in biomedical research each year, and the private sector in the U.S. spends more than twice that. Clearly we value these discoveries. Why is it so hard to use them?
From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show: we’ll talk about the promise of lifesaving genetic treatments. But first, how can we find the people who might benefit from them?
Gaurav SINGAL: What if artificial intelligence could examine the fingerprints, the breadcrumbs that patients leave throughout the healthcare system.
And Fyodor Urnov will tell us how editing the human genome can cure disease — and why his answers to those desperate emails aren’t so straightforward.
URNOV: Our ability to engineer these CRISPR medicines has far outpaced how these medicines are actually built, tested, and put into human beings.
* * *
SINGAL: I’m Gaurav Singal. I’m a physician and computer scientist, studied artificial intelligence and robotics, and ultimately became a doctor to see patients.
JENA: When you say you have a background in robotics, does that just mean you used to play with legos?
SINGAL: Yeah, no, actually I helped build a Lego-based team of autonomous, soccer-playing robots as an undergraduate.
JENA: And they were in the World Cup recently?
SINGAL: They were in the Robo Cup, actually. We played at Carnegie Mellon and got destroyed by robots that had seven wheels. We had two wheels and, you know, lost.
JENA: Are you joking or is this—.
SINGAL: I’m a hundred percent serious.
When he’s not playing with Legos, my friend Gaurav spends his time using computer science to solve healthcare problems. Most recently, he was the Chief Data Officer of a company called Foundation Medicine, which develops tests that diagnose cancer patients with specific genetic mutations. Now he sees patients at Brigham and Women’s Hospital in Boston and advises other companies that are using big data and artificial intelligence to solve problems in medicine. Artificial intelligence in a doctor’s office may sound as science-fiction as, say, soccer-playing robots … but the fact is that artificial intelligence already permeates our lives.
SINGAL: Credit card companies have been using artificial intelligence to help map risk scores as part of your credit evaluation. Spotify uses artificial intelligence to make personalized recommendations. Things like Google Photos have the ability to match photos of my children all the way from when they were born till now when they’re 9 and 7. That’s incredible. The metric for a long time has been: can computers do things as well as humans? But you see places like this task of matching infant pictures to childhood pictures where computers outperform humans. And once you cross that threshold, you get to a real opportunity where computers could complement humans. So when we get to medicine, I think this becomes particularly relevant. What are computers and artificial intelligence good at? Two things, at least. No. 1, pattern matching. No. 2, doing things very quickly. So where those two things are important, there may be a real role for computers and artificial intelligence.
One example where that’s the case is diagnosing strokes.
SINGAL: When a patient has a stroke, part of the blood flow to their brain has been blocked, and every minute that goes by, more and more neurons die. If you wait too long, that brain tissue has already died, and in fact, opening up the blood vessel no longer has any benefit.
When a patient shows up at the emergency room with a suspected stroke, they need to get treatment fast. But first, to confirm what kind of stroke they had, those patients usually get a C.T. scan, which has to be read by a radiologist.
SINGAL: That’s a very busy environment. It may be the case that that C.T. scan of the head — very important, very time sensitive — is in a queue of equally important and equally urgent scans that that radiologist has to read. So, it may take five minutes, 10 minutes, 20 minutes for that radiologist to review it to see if it has a stroke, only after which can that patient be evaluated and hopefully treated if it’s in the time window for treatment. One place that artificial intelligence has already had an impact is analyzing scans of these head C.T.s in the emergency room faster than the radiologist ever could. One example is a company called Viz.ai that has an F.D.A.-approved algorithm for detecting stroke in the emergency room. It lives on the scanner of hospitals all over the country.
I should note that Viz.ai is one of the companies that Gaurav has consulted with.
SINGAL: Every time a scan is taken of someone’s head, that algorithm runs on that scan and determines if it believes there’s a stroke there. If a stroke is detected by the algorithm, a radiologist immediately reviews it, determines if that patient has indeed had a stroke, and rushes the next steps for intervention. The result of this has been faster detection of strokes, often by a dozen minutes or more.
JENA: Stroke is really, really common. Are there examples where this technology is being deployed in areas where the diseases are much less common, what we might call rare diseases?
SINGAL: I think rare diseases are the next frontier for computational diagnostics. They’re not diseases that most providers see every day, by definition, and, as a result, believed to be highly underdiagnosed. Meaning, for some of these conditions, the subset of people who know that they have the condition is a small minority of the people who actually have the condition. A common expression that we probably both heard in training is: “when you hear hoof beats, think horses, not zebras” — that the rare diseases one thinks about later. On the flip side of that, if you’re a patient with a rare disease, that can be a very frustrating experience. It can mean going from doctor to doctor, from proposed diagnosis to proposed diagnosis, a litany of tests and evaluations and treatments, all without any benefit, while you’re on what is often termed a “diagnostic odyssey.” That can take months, it can take years, it can take lots of expense and heartache and frustration. If there were a way to make that diagnosis earlier, that could be tremendously beneficial to the patient, to the health system, avoiding all this unnecessary work, and getting on the right treatment sooner.
The U.S. government defines “rare diseases” as those that affect fewer than 200,000 people in the country. Some affect only a handful of people. But the word “rare” can be misleading when talking about rare diseases because there are more than 7,000 of them. Taken all together, more than 30 million people in the United States have been diagnosed with a rare disease — that’s around 10 percent of the population. So improving how we find and care for those patients could have a really big impact.
SINGAL: What if artificial intelligence could examine the fingerprints, the breadcrumbs that patients leave throughout the healthcare system as part of their routine care — through the radiology imaging they’ve gotten, the E.K.G.s that have been done, the lab work they’ve had done — What if it were possible for A.I. to interpret that in the background, passively, without anybody even needing to think about the rare disease and have proactively brought it up, and alert the patient or the physician that, “Hey, this is something that you might wanna keep an eye out for. I noticed this pattern.” That feels like an incredibly powerful area, and I believe there’s real examples where that’s now possible.
SINGAL: One example might be looking at an E.K.G. There’s a lot more signal in an E.K.G. than just, “Have you had a heart attack or not?” It’s an incredibly rich analysis of the electrical conduction of the heart. And in fact, there are now a number of different research scientists across the country that have demonstrated that if you give a computer sufficient training in analyzing an E.K.G., using artificial intelligence, a computer can predict which subset of patients is more likely to have one of these rare conditions.
JENA: Don’t you worry about false positives in this sort of scenario?
SINGAL: That’s a great point. Artificial intelligence isn’t perfect. Like any screening test, it flags people who may have the condition and then they can go get the confirmatory testing. The problem is today, we’re not flagging enough people who are at risk, and as a result, the majority of people with some of these conditions don’t know they have it.
Fyodor Urnov, the geneticist we heard from earlier, is all too familiar with the diagnostic odyssey that Gaurav described.
URNOV: It’s heartbreaking. But the good news is a key technology advance has happened literally in the past five years that I think is an enormous call-to-action for pretty much the entirety of the biomedical community.
Fyodor isn’t talking about artificial intelligence. He’s talking about an advance in the next step on the path to diagnosis, the one that would come after a patient is flagged by A.I. as being at risk for a rare disease: genetic testing.
URNOV: The first complete sequence of the human genome was obtained two decades ago. It took about a decade and 3 billion dollars. Today, there is a room on the U.C. Berkeley campus and on the U.C.S.F. campus across the bay, or at Stanford, or a large number of places where you can walk into the door and a technician will take one of those little swabs, you know, the ones you use for covid tests, and swirl them in your mouth or your nose. And 24 hours later you can get a link to your complete genetic sequence. If you ask me when I was a graduate student in the mid-nineties up at Brown, “Do you think we’ll ever get to a time where we could do that in a day?” I’d go, “Oh, come on. We’d more likely move faster than the speed of light.” And yet here we are. This isn’t some hypothetical, something that will exist in 2033. This exists in January, 2023. So, the technologies to read D.N.A. at an incredible rate are here. And they got so much cheaper. And so, these folks who described to me the harrowing odyssey of, like, “What’s wrong with my child?” should not suffer. We as a society, we as a species, owe it to folks in such predicaments to develop and deploy a scalable solution of rapid genetic diagnosis.
SINGAL: A.I. could serve a lot of other functions in healthcare, including helping design new treatments, helping predict which treatment is better for which patient. But the idea of screening for rare diseases or diagnosing undiagnosed conditions, these feel absolutely here and now.
And as Gaurav said earlier, artificial intelligence is already being used in emergency rooms to instantaneously screen for a more common disease, stroke.
SINGAL: You’re improving patients’ lives by decreasing morbidity from stroke in a way that saves payers money and increases the number of procedures that are done in a fee-for-service environment. And so you have incentive alignment between payers and healthcare systems to do what’s in the best interest of patients. Implementation is not trivial. You have to deploy software, algorithms, workflow tools behind the firewall of hundreds and thousands of health systems. You have to get providers to use the tools, you have to get them to make decisions based on them. There are a lot of hurdles that have to be overcome, and yet, when incentives are aligned, that can happen.
The question is what will be required for the example of detecting rare cardiovascular condition from E.K.G.s to become real? The payer is caught in a bind here because if we screen more for rare conditions, we identify more patients who will need to have expensive treatments. On the other hand, you have an entire industry that’s developing novel medicines for these patients who can’t get the medicine today because they don’t even know they have the condition that’s being treated.
Fyodor Urnov is among the scientists developing those novel medicines. But his work comes with its own set of economic challenges.
URNOV: We owe it to the patients and the families to aggressively build a new framework.
That’s after the break. I’m Bapu Jena, and this is Freakonomics, M.D.
* * *
JENA: What is DNA? Explain it in a way that someone who doesn’t have a medical background would understand.
URNOV: Asking a geneticist, “what’s D.N.A.?” is like asking an astronomer, “what’s a star?” You know, it’s a ball of light.
Before the break, I mentioned that Dr. Fyodor Urnov is developing treatments for genetic diseases. But first, let’s take a step back. What is a genetic disease? And, seriously, what’s D.N.A.? Like what is it, really?
URNOV: Um, it’s a molecule, so it consists of atoms just like everything else. And it has two remarkable properties that pretty much no other molecule has. It can carry in it genetic information. Just like a piece of paper can carry a sentence, D.N.A. can carry genetic sentences. In you and I, it carries 20,000 such genetic instructions, which we call genes. But the other property of D.N.A., which inspired me to devote my life to it, is that it’s conceptually and molecularly straightforward to make a copy of it.
But that copying process isn’t foolproof. As D.N.A. reproduces itself again and again, sometimes there are little typos, or mutations, in the genetic instructions it holds.
URNOV: That’s the basis of a process, which we call evolution, that gave us this wonderful constellation of bacteria, animals, plants, you and I. If D.N.A. never changed, you and I would still be little microbes floating around in the primordial soup. So this intrinsic ability of D.N.A. to change is the basis of life, paradoxically.
Some of these changes have beneficial outcomes — like a mutation that occurred around 5,000 years ago that allowed humans to digest lactose for the first time. I’m happy I inherited that one. But sometimes, the outcomes of a typo in genetic instructions can change our lives for the worse. Those are what we call genetic diseases. One of the more common ones you may have heard of is sickle cell disease. Around 100,000 Americans suffer from it, so we classify it as a “rare” disease, even though it’s the most common inherited blood disorder in the country, and affects millions of people worldwide. People who inherit sickle cell disease can’t form normal red blood cells that carry oxygen; instead, they produce red blood cells that are rigid and sickle-shaped, like a crescent. That deformity causes extreme pain episodes that put sickle cell patients in the hospital on a regular basis. It also delays normal development in children; damages joints, nerves, and organs; and often causes strokes. All of these bad outcomes are the result of just one letter out of 6 billion in the genome being flipped.
URNOV: One tiny typo spreads its devastating effects through the entire body. And even with the best healthcare, our fellow Americans with sickle cell disease, their lifespan is around the mid-forties. So it takes away decades from your life.
Until now, the only way to cure sickle cell disease was with a bone marrow transplant, but the procedure is not for everyone. It can be difficult to find a well-matched donor, and bone marrow transplants are really hard on the body, especially as patients get older. What if instead of replacing the patient’s faulty bone marrow, doctors could actually fix the typo in the patient’s own bone marrow? That would be a much safer procedure — and eliminate the need for a well-matched donor, meaning anyone with the disease could potentially be cured. Thanks to a revolutionary gene-editing technology called CRISPR, scientists like Fyodor are now doing exactly that.
URNOV: The first thing to note about CRISPR is it’s one of those acronyms where what the acronym stands for is not useful to know because it doesn’t tell you anything about what it does. And there are great examples to the contrary, let’s say SCUBA, right, or “self-contained underwater breathing apparatus.” If you know what the acronym stands for, you’re like, “Aha!” But CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeats. And your audience is welcome to forget that immediately.
JENA: Or say it a hundred times, it’ll help you go to sleep.
URNOV: Or become the least popular person at a social gathering. “Hey, I know what CRISPR stands for!” So, I’m sitting in the recording studio at the School of Journalism of the University of California Berkeley, where I’m a professor. And probably the single biggest discovery in biomedicine of the past quarter century was made here on the Berkeley campus.
That discovery was made by the biochemist Jennifer Doudna — who, together with Emmanuelle Charpentier, won the 2020 Nobel Prize in Chemistry. To be clear, Jennifer and Emmanuelle didn’t create CRISPR. CRISPR is not a fancy new lab machine. It’s a microbial defense mechanism. It consists of just two molecules: an enzyme that acts as a pair of DNA scissors, and a special piece of genetic material that tells the enzyme where in the DNA to cut. And: it’s billions of years old. Early in the history of life, bacteria evolved CRISPR to fight off parasites that could attack and kill them.
URNOV: It’s basically a little molecular machine that carries in it a memory of a previous attack by a genetic invader, a snippet of the offender’s genetic material, like a law enforcement officer with a most wanted poster with a picture of somebody suspected of a crime. And it literally matches every piece of D.N.A. it sees. “Do you have a match to this 20 letter word that I’m carrying inside me? If yes, I’ll cut you and destroy you. If not, have a nice day.”
Jennifer and Emmanuelle’s big discovery was not that CRISPR exists. What they discovered was something that CRISPR can do.
URNOV: So it turns out that you can put CRISPR inside human cells, which seems insane. This thing comes from bacteria, which are billions of years apart from us evolutionarily. You can take CRISPR, you can give it a 20-letter match to a human gene that’s broken and it’ll fix it. We don’t understand why it’s been so successful in this incredible, new environment. But we’re grateful to Mother Nature, and of course to Jennifer and Emmanuelle for having the insight that you can program — this is the key word — CRISPR can be programmed.
Not only have scientists wielded CRISPR’s innate destructive function to eliminate toxic genes, but they’ve also come up with ways to make CRISPR serve a constructive function — that is, to precisely alter just one letter in the D.N.A., to repair a gene rather than getting rid of it altogether. Geneticists made use of that function to develop a cure for sickle cell disease, which is currently the first CRISPR-based therapeutic up for approval by the F.D.A.
URNOV: This is a great example of the ways in which we humans have borrowed from Mother Nature and then elaborated on her inventions. And we wouldn’t be talking about this if this hasn’t been used on people.
CRISPR isn’t the first approach to gene therapy; there are several approved medications that use modified viruses to deliver disease-treating genetic material into a patient’s cells. But CRISPR cures are the first to edit the genome itself. So far, CRISPR has been used to treat genetic diseases of the bloodstream, liver, eye, and immune system. For others, like those affecting the lungs, brain, and kidneys, scientists haven’t yet figured out how to get CRISPR into enough of the organ to actually heal it — so, to be clear, CRISPR is still far from a “cure all.” But as new techniques and technologies to deliver CRISPR are developed, more and more organs will come online. Fyodor expects the lungs to be next.
Having the power to cure genetic diseases by editing the human genome is a dream come true for geneticists. But when it comes to using that power to help people, the story gets more complicated.
URNOV: Many a time, when parents of children with severe genetic disease send me a note saying, “Dr. Urnov, is there anything you can do?” If they are willing to share what the mutation is, I can load it into some software in my computer — which is available to all, I wanna be clear, not some proprietary U.C. Berkeley software — and you can basically engineer, if you know what you’re doing, a CRISPR to fix that mutation. For many diseases, that engineered CRISPR on my computer screen can become a vial with that CRISPR that we can pretty quickly test for whether it can repair the defect safely and effectively. Start to finish, if you know what you’re doing, it’ll take well under a year. So do I write back to the parents and say, “Hi, guess what?” No. And here’s why. Engineering the medicine is the first step of probably a four-year process to protect patients from faulty medicines. I wanna really emphasize, I’m not sitting here and saying, “Get rid of the laws to protect patients from faulty medicines.” But going through that four-year process just to get to the clinic takes anywhere between eight to $10 million for one disease. If the disease is relatively prevalent, like sickle cell disease, And then if they charge what is currently being charged for these types of medicines, which is 1 to 2 to $3 million a patient, I can see why a company would invest years and millions to build a medicine. Okay, now I get this email from somebody and they have two children, and both children have this change and it’s unclear that anybody else on planet Earth has that genetic change. So who exactly is going to spend four years and $10 million building a medicine that’s gonna be used to treat two kids?
Many of these diseases are individually so rare that they do not form, a viable commercial proposition under the current system. We need to face the remarkable reality that our ability to engineer these CRISPR medicines has far outpaced how these medicines are actually built, tested, and put into human beings. We have never had a technology like CRISPR. We owe it to the patients and the families to aggressively build a new framework to provide these medicines to these individuals.
JENA: It scares me because it’s one thing to say to somebody that we don’t have a treatment because the biology doesn’t exist to provide that treatment. It’s another thing to say that we don’t have a treatment because there’s not sufficient commercial incentive to develop that treatment, to evaluate it, to test it, to market it, or that we have a set of regulatory policies that aren’t adept enough to recognize that there are some patients with some diseases who, you know, literally months matter in terms of getting access to care. We wanna get medicines to people faster, but we wanna make sure that we do so in a way that’s safe. And the F.D.A. is really tasked with managing that speed/safety tradeoff. But, of course, that trade-off should change when the parameters change, right? So if you have a new technology that will allow for personalized intervention in people with life-threatening diseases for which early treatment really does matter, we should be able to create a regulatory pathway that would allow for that. And then there’s the other bucket of alright, well, how do we pay for that? That commercialization issue is equally important.
URNOV: I cannot improve on what you just said. I’ll just add one point. Many genetic diseases are diagnosed in human beings at a stage where current technology — and I emphasize “current” because our field is moving very fast — current technology is essentially powerless. By the time we are looking at that human being in a clinic, it’s too late. A really profound and poignant example is the disease that killed Woody Guthrie, which is Huntington’s disease. It’s a broken gene, it’s actually a toxic gene, which is basically killing the brain. And by the time people develop symptoms, parts of their brain are just gone. And we don’t have a technology that can bring it back.
Remember the email from the very beginning of the episode that Fyodor received from the parent of a girl going progressively blind?
URNOV: We don’t know if we’ll be able to turn back time and bring back vision to 100 percent, but at the very least, if we can diagnose early enough, if we can intervene at the genetic level before it gets worse, I can tell you that the patient community, the vast, worldwide community of folks with genetic disease will applaud.
This brings us back to artificial intelligence, and its role in catching these rare genetic diseases as early as possible. Here’s Gaurav Singal again.
SINGAL: Now that there are effective treatments, it feels more important than ever that we use techniques like this to make sure that people who have this condition know they have it so they can get treated.
And in Fyodor’s eyes, this is a two-way street.
URNOV: I think we as a community owe it to the folks out there whose genetic changes we’re identifying as potentially dangerous or disease-driving to make sure that our ability to address those changes actionably in the clinic catches up to how quickly we can identify them.
JENA: Suppose that we can solve the economic puzzle. What’s your big picture, ideal vision?
URNOV: A world where genetic disease is diagnosed early in a way that’s so affordable that health insurance just covers it. And then in cases where that’s appropriate, the CRISPR medicine is manufactured and administered to that individual in a way that is scalable, affordable, and does not involve years and millions. Really having — I don’t know what we’re gonna call them— CRISPR clinics? Today, you know, I’m wearing glasses. It’s a way to correct my myopia. Do I see a future where CRISPR is deployed to repair genetic defects in a way that’s relatively commonplace? I do.
That’s a world I want to live in. So, how can we solve the economic puzzle? How can we make developing cures for rare genetic diseases profitable — and accessing those cures affordable?
LO: financing ends up being a tremendous roadblock. But with the right kind of financing, it actually ends up accelerating our ability to treat these patients.
And: what’s it like to receive a CRISPR cure?
OLAGHERE: It all amounted to a small syringe of D.N.A. that took about 30 seconds to infuse. And my life changed completely.
That’s coming up next week on Freakonomics, M.D. In the meantime, let us know what you thought about the show. I’m at email@example.com. That’s B-A-P-U at freakonomics.com. That’s it for today. I’d like to thank my guests this week, Dr. Gaurav Singal and Dr. Fyodor Urnov. And thanks to you, of course, for listening.
* * *
Freakonomics, M.D. is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and People I (Mostly) Admire. All our shows are produced by Stitcher and Renbud Radio. You can find us on Twitter at @drbapupod. And now you can find our episodes on YouTube too, by going to youtube dot com, slash @freakonomics — that’s the “at” sign, followed by freakonomics. If you know someone who doesn’t listen to podcasts but spends a lot of time on YouTube, let them know! This episode was produced by Julie Kanfer and Lyric Bowditch. It was mixed by Eleanor Osborne. Our executive team is Neal Carruth, Gabriel Roth, and Stephen Dubner. Original music composed by Luis Guerra. As always, thanks for listening.
JENA: For the last 22 years, every time I talk to my biology colleagues, I’m always happy that I did a PhD in economics. But when I’m talking to you, I’m thinking to myself, “daggonit, why didn’t I do something different 22 years ago?”
URNOV: But wait, you are an MD, right?
JENA: I’m an MD and a PhD in economics, yeah. Both.
URNOV: Wow. That’s two lives in one.
JENA: Exactly, yeah. I don’t know if it’s a good thing or a bad thing.
- Gaurav Singal, attending physician at Brigham and Women’s Hospital; former chief data officer at Foundation Medicine.
- Fyodor Urnov, professor of molecular and cell biology at the University of California Berkeley; scientific director at the Innovative Genomics Institute.
- “We Can Cure Disease by Editing a Person’s DNA. Why Aren’t We?” by Fyodor Urnov (The New York Times, 2022).
- “Artificial Intelligence–Parallel Stroke Workflow Tool Improves Reperfusion Rates and Door‐In to Puncture Interval,” by Ameer E. Hassan, Victor M. Ringheanu, Laurie Preston, and Wondwossen G. Tekle (Stroke: Vascular and Interventional Neurology, 2022).
- “Artificial Intelligence-Enhanced Electrocardiography in Cardiovascular Disease Management,” by Konstantinos C. Siontis, Peter A. Noseworthy, Zachi I. Attia, and Paul A. Friedman (Nature Reviews Cardiology, 2021).
- “CRISPR-Cas9 In Vivo Gene Editing for Transthyretin Amyloidosis,” by Julian D. Gillmore, Ed Gane, Jorg Taubel, Justin Kao, Marianna Fontana, et al. (The New England Journal of Medicine, 2021).
- “The Trade-off Between Speed and Safety in Drug Approvals,” by Anupam B. Jena, Jie Zhang, and Darius N. Lakdawalla (JAMA Network, 2017).
- “A Programmable Dual-RNA–Guided DNA Endonuclease in Adaptive Bacterial Immunity,” by Martin Jinek, Krzysztof Chylinski, Ines Fonfara, Michael Hauer, Jennifer A. Doudna, and Emmanuelle Charpentier (Science, 2012).
- “We Can Play God Now,” by People I (Mostly) Admire (2022).
- “Evolution, Accelerated,” by Freakonomics Radio (2017).