TED2008

Paul Rothemund: DNA folding, in detail

Filmed:

In 2007, Paul Rothemund gave TED a short summary of his specialty, DNA folding. Now he lays out in clear, abundant detail the immense promise of this field -- to create tiny machines that assemble themselves.

- DNA origamist
Paul Rothemund folds DNA into shapes and patterns. Which is a simple enough thing to say, but the process he has developed has vast implications for computing and manufacturing -- allowing us to create things we can now only dream of. Full bio

So, people argue vigorously about the definition of life.
00:12
They ask if it should have reproduction in it, or metabolism, or evolution.
00:15
And I don't know the answer to that, so I'm not going to tell you.
00:20
I will say that life involves computation.
00:22
So this is a computer program.
00:25
Booted up in a cell, the program would execute,
00:27
and it could result in this person;
00:30
or with a small change, it could result in this person;
00:33
or another small change, this person;
00:36
or with a larger change, this dog,
00:38
or this tree, or this whale.
00:41
So now, if you take this metaphor
00:43
[of] genome as program seriously,
00:45
you have to consider that Chris Anderson
00:47
is a computer-fabricated artifact, as is Jim Watson,
00:49
Craig Venter, as are all of us.
00:52
And in convincing yourself that this metaphor is true,
00:55
there are lots of similarities between genetic programs
00:57
and computer programs that could help to convince you.
00:59
But one, to me, that's most compelling
01:02
is the peculiar sensitivity to small changes
01:04
that can make large changes in biological development -- the output.
01:07
A small mutation can take a two-wing fly
01:10
and make it a four-wing fly.
01:12
Or it could take a fly and put legs where its antennae should be.
01:13
Or if you're familiar with "The Princess Bride,"
01:17
it could create a six-fingered man.
01:19
Now, a hallmark of computer programs
01:21
is just this kind of sensitivity to small changes.
01:23
If your bank account's one dollar, and you flip a single bit,
01:26
you could end up with a thousand dollars.
01:28
So these small changes are things that I think
01:30
that -- they indicate to us that a complicated computation
01:33
in development is underlying these amplified, large changes.
01:35
So now, all of this indicates that there are molecular programs underlying biology,
01:39
and it shows the power of molecular programs -- biology does.
01:45
And what I want to do is write molecular programs,
01:49
potentially to build technology.
01:51
And there are a lot of people doing this,
01:53
a lot of synthetic biologists doing this, like Craig Venter.
01:54
And they concentrate on using cells.
01:57
They're cell-oriented.
01:59
So my friends, molecular programmers, and I
02:01
have a sort of biomolecule-centric approach.
02:03
We're interested in using DNA, RNA and protein,
02:05
and building new languages for building things from the bottom up,
02:08
using biomolecules,
02:11
potentially having nothing to do with biology.
02:12
So, these are all the machines in a cell.
02:15
There's a camera.
02:19
There's the solar panels of the cell,
02:21
some switches that turn your genes on and off,
02:22
the girders of the cell, motors that move your muscles.
02:24
My little group of molecular programmers
02:27
are trying to refashion all of these parts from DNA.
02:29
We're not DNA zealots, but DNA is the cheapest,
02:33
easiest to understand and easy to program material to do this.
02:35
And as other things become easier to use --
02:38
maybe protein -- we'll work with those.
02:40
If we succeed, what will molecular programming look like?
02:43
You're going to sit in front of your computer.
02:45
You're going to design something like a cell phone,
02:47
and in a high-level language, you'll describe that cell phone.
02:49
Then you're going to have a compiler
02:51
that's going to take that description
02:53
and it's going to turn it into actual molecules
02:54
that can be sent to a synthesizer
02:56
and that synthesizer will pack those molecules into a seed.
02:58
And what happens if you water and feed that seed appropriately,
03:01
is it will do a developmental computation,
03:04
a molecular computation, and it'll build an electronic computer.
03:06
And if I haven't revealed my prejudices already,
03:09
I think that life has been about molecular computers
03:12
building electrochemical computers,
03:14
building electronic computers,
03:16
which together with electrochemical computers
03:18
will build new molecular computers,
03:20
which will build new electronic computers, and so forth.
03:22
And if you buy all of this,
03:25
and you think life is about computation, as I do,
03:26
then you look at big questions through the eyes of a computer scientist.
03:28
So one big question is, how does a baby know when to stop growing?
03:31
And for molecular programming,
03:35
the question is how does your cell phone know when to stop growing?
03:37
(Laughter)
03:39
Or how does a computer program know when to stop running?
03:40
Or more to the point, how do you know if a program will ever stop?
03:43
There are other questions like this, too.
03:46
One of them is Craig Venter's question.
03:48
Turns out I think he's actually a computer scientist.
03:50
He asked, how big is the minimal genome
03:52
that will give me a functioning microorganism?
03:55
How few genes can I use?
03:57
This is exactly analogous to the question,
03:59
what's the smallest program I can write
04:01
that will act exactly like Microsoft Word?
04:02
(Laughter)
04:04
And just as he's writing, you know, bacteria that will be smaller,
04:05
he's writing genomes that will work,
04:09
we could write smaller programs
04:10
that would do what Microsoft Word does.
04:12
But for molecular programming, our question is,
04:14
how many molecules do we need to put in that seed to get a cell phone?
04:16
What's the smallest number we can get away with?
04:20
Now, these are big questions in computer science.
04:22
These are all complexity questions,
04:24
and computer science tells us that these are very hard questions.
04:26
Almost -- many of them are impossible.
04:28
But for some tasks, we can start to answer them.
04:30
So, I'm going to start asking those questions
04:33
for the DNA structures I'm going to talk about next.
04:34
So, this is normal DNA, what you think of as normal DNA.
04:37
It's double-stranded, it's a double helix,
04:40
has the As, Ts, Cs and Gs that pair to hold the strands together.
04:42
And I'm going to draw it like this sometimes,
04:45
just so I don't scare you.
04:47
We want to look at individual strands and not think about the double helix.
04:49
When we synthesize it, it comes single-stranded,
04:52
so we can take the blue strand in one tube
04:55
and make an orange strand in the other tube,
04:58
and they're floppy when they're single-stranded.
05:00
You mix them together and they make a rigid double helix.
05:02
Now for the last 25 years,
05:05
Ned Seeman and a bunch of his descendants
05:07
have worked very hard and made beautiful three-dimensional structures
05:09
using this kind of reaction of DNA strands coming together.
05:12
But a lot of their approaches, though elegant, take a long time.
05:15
They can take a couple of years, or it can be difficult to design.
05:18
So I came up with a new method a couple of years ago
05:21
I call DNA origami
05:24
that's so easy you could do it at home in your kitchen
05:25
and design the stuff on a laptop.
05:27
But to do it, you need a long, single strand of DNA,
05:29
which is technically very difficult to get.
05:32
So, you can go to a natural source.
05:34
You can look in this computer-fabricated artifact,
05:36
and he's got a double-stranded genome -- that's no good.
05:38
You look in his intestines. There are billions of bacteria.
05:40
They're no good either.
05:43
Double strand again, but inside them, they're infected with a virus
05:45
that has a nice, long, single-stranded genome
05:47
that we can fold like a piece of paper.
05:50
And here's how we do it.
05:52
This is part of that genome.
05:53
We add a bunch of short, synthetic DNAs that I call staples.
05:54
Each one has a left half that binds the long strand in one place,
05:57
and a right half that binds it in a different place,
06:01
and brings the long strand together like this.
06:04
The net action of many of these on that long strand
06:07
is to fold it into something like a rectangle.
06:09
Now, we can't actually take a movie of this process,
06:11
but Shawn Douglas at Harvard
06:13
has made a nice visualization for us
06:15
that begins with a long strand and has some short strands in it.
06:17
And what happens is that we mix these strands together.
06:21
We heat them up, we add a little bit of salt,
06:25
we heat them up to almost boiling and cool them down,
06:27
and as we cool them down,
06:29
the short strands bind the long strands
06:30
and start to form structure.
06:32
And you can see a little bit of double helix forming there.
06:34
When you look at DNA origami,
06:38
you can see that what it really is,
06:40
even though you think it's complicated,
06:43
is a bunch of double helices that are parallel to each other,
06:44
and they're held together
06:47
by places where short strands go along one helix
06:49
and then jump to another one.
06:51
So there's a strand that goes like this, goes along one helix and binds --
06:53
it jumps to another helix and comes back.
06:56
That holds the long strand like this.
06:58
Now, to show that we could make any shape or pattern
07:00
that we wanted, I tried to make this shape.
07:03
I wanted to fold DNA into something that goes up over the eye,
07:06
down the nose, up the nose, around the forehead,
07:08
back down and end in a little loop like this.
07:11
And so, I thought, if this could work, anything could work.
07:14
So I had the computer program design the short staples to do this.
07:17
I ordered them; they came by FedEx.
07:20
I mixed them up, heated them, cooled them down,
07:22
and I got 50 billion little smiley faces
07:24
floating around in a single drop of water.
07:28
And each one of these is just
07:30
one-thousandth the width of a human hair, OK?
07:32
So, they're all floating around in solution, and to look at them,
07:36
you have to get them on a surface where they stick.
07:39
So, you pour them out onto a surface
07:41
and they start to stick to that surface,
07:43
and we take a picture using an atomic-force microscope.
07:45
It's got a needle, like a record needle,
07:47
that goes back and forth over the surface,
07:49
bumps up and down, and feels the height of the first surface.
07:51
It feels the DNA origami.
07:54
There's the atomic-force microscope working
07:56
and you can see that the landing's a little rough.
07:59
When you zoom in, they've got, you know,
08:00
weak jaws that flip over their heads
08:02
and some of their noses get punched out, but it's pretty good.
08:03
You can zoom in and even see the extra little loop,
08:06
this little nano-goatee.
08:08
Now, what's great about this is anybody can do this.
08:10
And so, I got this in the mail about a year after I did this, unsolicited.
08:13
Anyone know what this is? What is it?
08:17
It's China, right?
08:20
So, what happened is, a graduate student in China,
08:22
Lulu Qian, did a great job.
08:24
She wrote all her own software
08:26
to design and built this DNA origami,
08:28
a beautiful rendition of China, which even has Taiwan,
08:30
and you can see it's sort of on the world's shortest leash, right?
08:33
(Laughter)
08:36
So, this works really well
08:39
and you can make patterns as well as shapes, OK?
08:41
And you can make a map of the Americas and spell DNA with DNA.
08:44
And what's really neat about it --
08:47
well, actually, this all looks like nano-artwork,
08:50
but it turns out that nano-artwork
08:52
is just what you need to make nano-circuits.
08:53
So, you can put circuit components on the staples,
08:55
like a light bulb and a light switch.
08:57
Let the thing assemble, and you'll get some kind of a circuit.
08:59
And then you can maybe wash the DNA away and have the circuit left over.
09:02
So, this is what some colleagues of mine at Caltech did.
09:05
They took a DNA origami, organized some carbon nano-tubes,
09:07
made a little switch, you see here, wired it up,
09:10
tested it and showed that it is indeed a switch.
09:12
Now, this is just a single switch
09:15
and you need half a billion for a computer, so we have a long way to go.
09:17
But this is very promising
09:21
because the origami can organize parts just one-tenth the size
09:23
of those in a normal computer.
09:28
So it's very promising for making small computers.
09:29
Now, I want to get back to that compiler.
09:32
The DNA origami is a proof that that compiler actually works.
09:35
So, you start with something in the computer.
09:39
You get a high-level description of the computer program,
09:41
a high-level description of the origami.
09:44
You can compile it to molecules, send it to a synthesizer,
09:46
and it actually works.
09:49
And it turns out that a company has made a nice program
09:50
that's much better than my code, which was kind of ugly,
09:54
and will allow us to do this in a nice,
09:56
visual, computer-aided design way.
09:57
So, now you can say, all right,
10:00
why isn't DNA origami the end of the story?
10:01
You have your molecular compiler, you can do whatever you want.
10:03
The fact is that it does not scale.
10:05
So if you want to build a human from DNA origami,
10:08
the problem is, you need a long strand
10:11
that's 10 trillion trillion bases long.
10:13
That's three light years' worth of DNA,
10:16
so we're not going to do this.
10:18
We're going to turn to another technology,
10:20
called algorithmic self-assembly of tiles.
10:22
It was started by Erik Winfree,
10:24
and what it does,
10:26
it has tiles that are a hundredth the size of a DNA origami.
10:27
You zoom in, there are just four DNA strands
10:31
and they have little single-stranded bits on them
10:34
that can bind to other tiles, if they match.
10:36
And we like to draw these tiles as little squares.
10:38
And if you look at their sticky ends, these little DNA bits,
10:42
you can see that they actually form a checkerboard pattern.
10:44
So, these tiles would make a complicated, self-assembling checkerboard.
10:47
And the point of this, if you didn't catch that,
10:50
is that tiles are a kind of molecular program
10:52
and they can output patterns.
10:55
And a really amazing part of this is
10:58
that any computer program can be translated
11:00
into one of these tile programs -- specifically, counting.
11:02
So, you can come up with a set of tiles
11:05
that when they come together, form a little binary counter
11:08
rather than a checkerboard.
11:11
So you can read off binary numbers five, six and seven.
11:13
And in order to get these kinds of computations started right,
11:16
you need some kind of input, a kind of seed.
11:19
You can use DNA origami for that.
11:21
You can encode the number 32
11:23
in the right-hand side of a DNA origami,
11:25
and when you add those tiles that count,
11:27
they will start to count -- they will read that 32
11:29
and they'll stop at 32.
11:32
So, what we've done is we've figured out a way
11:34
to have a molecular program know when to stop going.
11:37
It knows when to stop growing because it can count.
11:40
It knows how big it is.
11:42
So, that answers that sort of first question I was talking about.
11:44
It doesn't tell us how babies do it, however.
11:47
So now, we can use this counting to try and get at much bigger things
11:50
than DNA origami could otherwise.
11:54
Here's the DNA origami, and what we can do
11:55
is we can write 32 on both edges of the DNA origami,
11:58
and we can now use our watering can
12:01
and water with tiles, and we can start growing tiles off of that
12:03
and create a square.
12:07
The counter serves as a template
12:09
to fill in a square in the middle of this thing.
12:12
So, what we've done is we've succeeded
12:14
in making something much bigger than a DNA origami
12:15
by combining DNA origami with tiles.
12:18
And the neat thing about it is, is that it's also reprogrammable.
12:21
You can just change a couple of the DNA strands in this binary representation
12:24
and you'll get 96 rather than 32.
12:28
And if you do that, the origami's the same size,
12:31
but the resulting square that you get is three times bigger.
12:34
So, this sort of recapitulates
12:39
what I was telling you about development.
12:40
You have a very sensitive computer program
12:42
where small changes -- single, tiny, little mutations --
12:45
can take something that made one size square
12:48
and make something very much bigger.
12:50
Now, this -- using counting to compute
12:54
and build these kinds of things
12:57
by this kind of developmental process
12:59
is something that also has bearing on Craig Venter's question.
13:01
So, you can ask, how many DNA strands are required
13:05
to build a square of a given size?
13:07
If we wanted to make a square of size 10, 100 or 1,000,
13:09
if we used DNA origami alone,
13:14
we would require a number of DNA strands that's the square
13:16
of the size of that square;
13:19
so we'd need 100, 10,000 or a million DNA strands.
13:21
That's really not affordable.
13:23
But if we use a little computation --
13:25
we use origami, plus some tiles that count --
13:27
then we can get away with using 100, 200 or 300 DNA strands.
13:31
And so we can exponentially reduce the number of DNA strands we use,
13:34
if we use counting, if we use a little bit of computation.
13:39
And so computation is some very powerful way
13:42
to reduce the number of molecules you need to build something,
13:45
to reduce the size of the genome that you're building.
13:48
And finally, I'm going to get back to that sort of crazy idea
13:51
about computers building computers.
13:54
If you look at the square that you build with the origami
13:56
and some counters growing off it,
13:59
the pattern that it has is exactly the pattern that you need
14:01
to make a memory.
14:04
So if you affix some wires and switches to those tiles --
14:05
rather than to the staple strands, you affix them to the tiles --
14:08
then they'll self-assemble the somewhat complicated circuits,
14:11
the demultiplexer circuits, that you need to address this memory.
14:14
So you can actually make a complicated circuit
14:17
using a little bit of computation.
14:19
It's a molecular computer building an electronic computer.
14:21
Now, you ask me, how far have we gotten down this path?
14:24
Experimentally, this is what we've done in the last year.
14:27
Here is a DNA origami rectangle,
14:30
and here are some tiles growing from it.
14:33
And you can see how they count.
14:35
One, two, three, four, five, six, nine, 10, 11, 12, 17.
14:37
So it's got some errors, but at least it counts up.
14:49
(Laughter)
14:53
So, it turns out we actually had this idea nine years ago,
14:54
and that's about the time constant for how long it takes
14:57
to do these kinds of things, so I think we made a lot of progress.
15:00
We've got ideas about how to fix these errors.
15:02
And I think in the next five or 10 years,
15:04
we'll make the kind of squares that I described
15:06
and maybe even get to some of those self-assembled circuits.
15:08
So now, what do I want you to take away from this talk?
15:11
I want you to remember that
15:15
to create life's very diverse and complex forms,
15:17
life uses computation to do that.
15:21
And the computations that it uses, they're molecular computations,
15:23
and in order to understand this and get a better handle on it,
15:27
as Feynman said, you know,
15:29
we need to build something to understand it.
15:31
And so we are going to use molecules and refashion this thing,
15:33
rebuild everything from the bottom up,
15:37
using DNA in ways that nature never intended,
15:39
using DNA origami,
15:42
and DNA origami to seed this algorithmic self-assembly.
15:44
You know, so this is all very cool,
15:47
but what I'd like you to take from the talk,
15:50
hopefully from some of those big questions,
15:51
is that this molecular programming isn't just about making gadgets.
15:53
It's not just making about --
15:56
it's making self-assembled cell phones and circuits.
15:58
What it's really about is taking computer science
16:00
and looking at big questions in a new light,
16:02
asking new versions of those big questions
16:05
and trying to understand how biology
16:07
can make such amazing things. Thank you.
16:09
(Applause)
16:12

▲Back to top

About the Speaker:

Paul Rothemund - DNA origamist
Paul Rothemund folds DNA into shapes and patterns. Which is a simple enough thing to say, but the process he has developed has vast implications for computing and manufacturing -- allowing us to create things we can now only dream of.

Why you should listen

Paul Rothemund won a MacArthur grant this year for a fairly mystifying study area: "folding DNA." It brings up the question: Why fold DNA? The answer is -- because the power to manipulate DNA in this way could change the way we make things at a very basic level.

Rothemund's work combines the study of self-assembly (watch the TEDTalks from Neil Gershenfeld and Saul Griffith for more on this) with the research being done in DNA nanotechnology -- and points the way toward self-assembling devices at microscale, making computer memory, for instance, smaller, faster and maybe even cheaper.

More profile about the speaker
Paul Rothemund | Speaker | TED.com