A Thumb Drive Made of Genes?
What do Shakespearian sonnets, a music video by the band OK Go (shown below), and an Amazon gift card all have in common?
All of these are pieces of data that have been successfully encoded in strands of DNA. The process takes the 0s and 1s of binary code and turns them into the As, Ts, Cs, and Gs of DNA nucleotides that can then be synthesized, and sequenced later to retrieve the information.
Why DNA? In theory, it could solve many of the problems of modern data storage, and hold greater densities of information for longer periods of time without the typical worries over obsolescence or corruption.
Since pioneering work from Harvard in 2012, research has advanced in terms of how much information can be encoded in DNA, and how accurately.
But what will it take to make DNA a mainstream data storage option?
Columbia University computational biologist Yaniv Erlich’s team reported in Science this week that they successfully stored an operating system, an 1895 French film, and several other files at densities previously unachieved. He joins electrical and computer engineer Olgica Milenkovic, of the University of Illinois at Urbana-Champaign, and biochemist Sriram Kosuri, of the University of California-Los Angeles, in a discussion of the hurdles ahead for biological data storage.
Yaniv Erlich is an assistant professor of Computer Science and Computational Biology at Columbia University as well as a core member at the New York Genome Center. in New York, New York.
Olgica Milenkovic is a professor of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign in Urbana, Illinois.
Sriram Kosuri is an assistant professor of Biochemistry at the University of California-Los Angeles in Los Angeles, California.
IRA FLATOW: This is Science Friday, I’m Ira Flatow. DNA– it’s the blueprint for life, but it’s also a string of letters– the As, the Cs, the Gs, and the Ts, not unlike the zeros and ones of the binary code we use to store data. And this similarity has led scientists to speculate that perhaps someday we could use DNA itself to store data more efficiently than silicon. Even better, it would last longer, avoid corruption problems, and never go the way of the VHS tape or the floppy disk.
And this is not a new concept. Researchers have been working on it for years and have successfully stored and retrieved everything from Shakespearean sonnets to photos to a music video from the band OK Go. And one of my guests has even stored an entire computer operating system in DNA. You can play Minesweeper off the code his team retrieved when they sequenced it again.
In 2013, we talked to a researcher who said all the knowledge of humankind written in DNA could fit in the back of a station wagon. So how soon are we actually going to archive all of that? When can you play Minesweeper off a DNA operating system? That’s what we’re going to be talking about today. If you want to join us, 844-724-8255. You can also tweet us @scifri.
Let me introduce my guest. Yaniv Erlich, Assistant Professor of Computer Science and Computation of Biology at Columbia University in New York. He’s a core member at the New York Genome Center. Welcome to Science Friday.
YANIV ERLICH: Hello.
IRA FLATOW: Olgica Milenkovic is professor of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign. Welcome to Science Friday.
OLGICA MILENKOVIC: Hello.
IRA FLATOW: Thank you for joining us. Sri Kosuri is an assistant professor of Biochemistry at UCLA in Los Angeles. Welcome to Science Friday.
SRIRAM KOSURI: Thank you for having us on.
IRA FLATOW: And you have a vial of DNA sitting in front of you, and what is stored on this tiny little bit of it?
YANIV ERLICH: So in this tiny little bit of DNA, we have here a French movie called The Arrival of a Train. It’s a French movie that was filmed in around 1885, so over 100 years ago. And now it’s on DNA.
We have a computer operating system. We have an Amazon gift card of $50. We have a manuscript, one of the most influential manuscripts in information theory. And we also put the Pioneer plaque that we sent to space, so we put the figure over there, and a computer virus on DNA.
IRA FLATOW: All in that tiny little, it looks like the head of a pin.
YANIV ERLICH: Exactly.
IRA FLATOW: Yeah, why is the gift card in there?
YANIV ERLICH: So the gift card is a funny thing. We wanted to encourage someone to reproduce our study. So in fact, I contacted one of my Twitter followers that was really interested in what we are doing. And I told him that you can download the sequencing data from this European archive that we store the data.
Here is the code. And if you can recover the files, you can have the $50 and just purchase something nice for you. So he did this operation, was able to get the file, and got a book about machine learning. We are all geeks.
IRA FLATOW: [LAUGHS] You’re in good company. Sri, we were talking about DNA as if putting data on it were an everyday thing, right? But this was sort of a gee whiz story even a year ago. How do you get the encoding on DNA?
SRIRAM KOSURI: Well, the encoding, when we did it many years ago in a very dumb way was essentially just ones or zeros replace Ts or Cs and Gs or As. So it was a very redundant code that wasn’t super dense when we did it just as a kind of proof of concept. What Yaniv done in this new work is really quite nice, which DNA has a lot of negative sides to it. So when we synthesize our sequenced DNA, we have all sorts of dropouts or certain sequences are hard to read or write.
And so Yaniv’s code, which he borrowed from computer science and these fountain codes, are basically taken from spotty transmission algorithms that are meant to convey information over noisy communication lines and used it to encode DNA, encode this information in DNA. And what that’s allowed him to do is really bring to scale the types of ideas that have been around for even before we were working on it, decades now, into something that seems pretty reliable in his newest work.
IRA FLATOW: Olgica, you let me ask you, how reliable is DNA? Why put all your photos on it? What are the benefits?
OLGICA MILENKOVIC: So if you think of DNA as the media, and my background is coding theory, and basically coding theorists have been [INAUDIBLE] magnetic optical racetrack memories for a very long time. And the interest is really to allow reliable writing, reliable reading of the information. And storing in DNA has many benefits, because DNA offers exceptional densities as we all know. But it’s also a naturally robust media that can survive for long periods of time.
The issues and the bottlenecks in terms of reliability come from the process of writing, which is basically synthesizing the DNA strings, and reading, which is some form of sequencing. And that’s where the interesting action, at least in coding theory happens, because the readout systems used for DNA storage systems are very different from what we have seen in classical recorders. The types of errors, the types of patterns that are error-prone when recorded are nothing like we’ve seen before, and they require very new approaches in coding theory, which is why I’m very excited about this field.
IRA FLATOW: Yaniv, is it error-free recording or how?
YANIV ERLICH: So it’s not error-free. We have some errors in the synthesis and the sequencing, as Olgica said. But our method allows us to go and to correct for these errors, and it’s highly robust. For instance, we showed that every time we read the DNA, we consume some amount of material.
Now for instance, my daughter, she loves to hear Frozen. So every day we hear maybe five times, the song of Frozen. So after a week, if Frozen was encoded in DNA, we would run out of material at all. So using our technique, we show that we can copy the DNA, copy the copy, copy the copy of the copy, and so one, nine times.
And we can retrieve– although we introduce more errors, we can retrieve and correct these errors and get a file that is fully accurate. We can watch this movie, we can play Minesweeper on the operating system. So it’s fully robust.
IRA FLATOW: But it’s not as fast yet as a hard drive, as writing and reading that kind of stuff.
YANIV ERLICH: Much slower than the hard drive, but copying exact copies of the data is extremely fast because it’s an enzymatic reaction.
IRA FLATOW: Hmm. And Olgica, I understand that you’re working on making DNA not just readable, but random access and rewritable like a hard drive, right?
OLGICA MILENKOVIC: So this is a big, interesting project that I’ve been dreaming of the last few years about. And as I mentioned, given the background of my group, which has been working in data storage for over 20 years now, we realized that there is so much that can be done with this media. And as you mentioned, random access is one thing. Another thing we did as well is make it portable, because as people that are really related to the storage industry, we know that once you have the data stored, you also need to read it in a very delay-tolerant manner, which we are not very close to doing right now.
But you also want to have your readout system handy. You don’t want to carry big sequencing devices that cost anywhere from half a million dollars to even more with you. So our recent work is focused on making systems that are portable. And we switched to using nanopore sequencing devices for that purpose.
IRA FLATOW: Let’s go into the phones. A lot of geeky folks like us are calling in. Let’s go to Eric in Fayetteville, Arkansas. Hi, Eric.
ERIC: Hi, Ira. I’m curious, how vulnerable is this data to corruption? Could a DNA drive mutate, and what would to the data if it did?
IRA FLATOW: Could it mutate? Wow, Yaniv, is it going to mutate? Thank you for the question.
YANIV ERLICH: That’s a great question, thank you so much. So we showed, in fact this copying process mutates the DNA quite a lot. It introduces many, many errors. But we have this type of error correcting codes of redundancy that allows us to complete that information. I will give you an analogy. The way that we encode the data is not kind of like we just transmit the file.
We first organize the file as if it was Sudoku puzzle. And then what we show– but it’s a very simple Sudoku puzzle, like a kids’ version. And we just send many, many hints. Every DNA molecule is a hint about the file.
Now, think that that I will give you a DNA Sudoku puzzle for kids, and I’m going to be mean. I’m going to erase some of the cells. Probably you can still go and complete the Sudoku puzzle, although I was mean and didn’t give you the entire puzzle. This is the same way that it works here. Although not all the DNA molecules are going to make it, some of them are erroneous, we can still solve the puzzle and get back the file.
IRA FLATOW: Mm-hm. Sri, you’re a biochemist. Do you worry about things mutating, like DNA?
SRIRAM KOSURI: Well, I think Yaniv is talking about the physical nature of DNA. I think sometimes people think about oh well, this is DNA and it could turn into a real life virus. I think we’ve talked about things like that in the past, but all the storage mechanisms we’ve been talking about thus far are outside of a living organism. So getting DNA back into an organism is extremely difficult.
And so the safety profile of such a thing is A, your likeliness that you’ll actually encode something living is little to none. But even then, getting it inside of the entire machinery of the cell is also extremely unlikely. So I think there’s very little thought that there would be danger there, especially since the size of these DNA molecules are quite short.
IRA FLATOW: Let’s go to Washington, thanks for that answer. Let’s go to Washington. Benjamin, welcome to Science Friday.
BENJAMIN: Thank you. I was curious as to whether you could use this technique with any other chemical structure other than biologically living DNA. Could you design something that’s sort of silicon-based, that would still use this same molecular encoding strategy?
YANIV ERLICH: I think that Olgica did some work?
OLGICA MILENKOVIC: Yes.
IRA FLATOW: Sri? Yeah.
OLGICA MILENKOVIC: Oh, please go ahead, Sri, if you want to say–
SRIRAM KOSURI: Oh, I think one point to make is that this is a biological polymer. And this is just a very specific version of information storage inside of a polymer suit. So one point could be, you could use any other polymer, one that might be easier to read and write, for instance.
But I think one advantage– there’s a couple of advantages of DNA, one being that there exists a wealth of enzymes and natural things that have evolved that make our lives a lot easier for example, than the ability to make copies from individual molecules, incredibly easy through the use of polymerases that have evolved for billions of years. Though we use a different polymer, it would be nice to have the tools to be able to deal with things like that.
OLGICA MILENKOVIC: If I may just add–
IRA FLATOW: Olgica, let me just interrupt because I have to be rude here. This is Science Friday from PRI, Public Radio International. OK, Olgica, go ahead.
OLGICA MILENKOVIC: Oh, sorry, yes. So what I wanted to say to the caller is that there are a lot of endeavors right now that try to use different polymers and synthetic [AUDIO OUT].
IRA FLATOW: Oh, her line has dropped now.
OLGICA MILENKOVIC: Oh, sorry. Can you hear me?
IRA FLATOW: Yes, go ahead.
OLGICA MILENKOVIC: Yeah, a group in France is using synthetic polymers, which have the big advantage of being very easy and cheap to synthesize, which is not the case with DNA. But as Sri pointed out, the fact that it’s a synthetic polymer really makes copying the part of the system that Yaniv is so efficiently using with DNA, copying is not very efficient with synthetic poly– [AUDIO OUT] Oh, go ahead.
IRA FLATOW: So we’re sort of using a CRISPR then, to edit the DNA? Is that one of the tools we might use?
OLGICA MILENKOVIC: Potentially. But it’s still pretty expensive and not very precise.
IRA FLATOW: How do all of you see using DNA in the long run? Is it going to be on a thumb drive, on a server farm? Yaniv, what do you think so far?
YANIV ERLICH: So I think the way that you use DNA is that most users will not even know that they are using DNA as their storage media. We think about something like a cloud service where you want to store some files for a very long amount of time, very cheaply. And there will be a service [INAUDIBLE] for a great price for you. But you will not know that it uses DNA. But this service can now take all the benefits of DNA, the compactness of this molecule, and do all the operation of sequencing, synthesizing, and taking care, rather than us actually carrying DNA with us around the city.
IRA FLATOW: And Sri, how are we going to make this cheaper and make it practical to use?
SRIRAM KOSURI: Yeah, I think that’s probably the million dollar question. Right now we’re on order, a million fold too expensive on the synthesis side. And we have seen drops that are very large in biology. And there are ways to think about getting to such drops. But that’s a long way from here to there, getting to about a million fold drop in cost.
IRA FLATOW: Olgica, what does your vision for the future look like?
OLGICA MILENKOVIC: I believe that it’s going to be a long road ahead. I agree with Sri. Ideally, I would like to see a DNA flash memory, because I think that is the DNA analog of a flash memory, because that would be the first thing we could hope for and the technology is very close to allowing us to do something like that, except for the cost. And I agree with Sri again that we need to drive the cost of synthesis and the delay of sequencing down in order to make this plausible.
IRA FLATOW: Let me get one quick call in from Amy in Manhattan. Hi Amy.
AMY: Hi. I was wondering, do we need to worry about coding DNA being hacked. And would the self-correcting mechanism make that less likely?
IRA FLATOW: Yaniv? Can you hack it?
YANIV ERLICH: So you cannot really hack it. It’s not a living organism. It’s like thinking about your coffee in the morning, that you drink the milk and there is DNA of cow, and somehow this DNA is taking over you. It’s just a molecule that you cannot really hack it, or it’s not a living organism. So I don’t think there is any risk over here.
IRA FLATOW: OK, this is quite fascinating. We’ll have to revisit this. Yaniv Erlich is assistant professor of Computer Science and Computational Biology at Columbia University in New York. Olgica Milenkovic is professor of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign. And Sri Kosuri to is assistant professor of Biochemistry at UCLA in Los Angeles. Thank you all for joining us today.