Storing data in DNA could be world changing,  really.

This pile of mismatched hard drives that   I keep in this basket of shame is a nightmare.

It is only about 20 terabytes of video files,   but it is bulky and annoying and just full of  cables and a mess, but scale my hard drives up   to even more data, and you're going to need  a much bigger basket.

Google search alone,   for example, processes over 200 petabytes  of data every single day.

That is 200,000   of these single terabyte drives or  200 million single gigabyte drives.

I don't know anyone who uses a single gigabyte  drive, but you get the idea.

No one has that   kind of storage space.

Well, maybe somebody does.

Now imagine storing all the data in the world,   all of it, every book, every movie, every  document, everything.

That is about 10   trillion gigabytes of digital data.

So how  much space do you think that would take?

Well,   you would need 1,000,000,010 terabyte hard drives.

If each one of those is about five inches long,   end to end, they would reach nearly 79,000 miles,  which is about three times around the equator.

But using DNA to store data could keep all  of that information in a space as small as   a coffee cup.

So what is it about DNA that's  making scientists look into using it as data   storage?

And what has to happen to get all of  our zeros and ones into tiny little DNA strands?

Gosh, this is great footage.

Scientists began playing around with  storing data in DNA back in the nineties,   and the idea has been around since the  1960s.

But why?

It seems like kind of a   weird leap to go from tapes and hard drives to  a biological molecule as a storage method.

But   DNA is the original data storage format.

I'm  a geneticist, so I'm biased.

But every living   thing on earth stores the information it  needs to build its cells and tissues and   respond to its environment in its DNA.

And DNA  has some properties that would lead itself well   to being a digital data storage solution.

One is that it is super stable over time.

Keep DNA in a dark cool space, not even a freezer,  but just something like a wine cellar, and it'll   last for hundreds of years.

DNA in permafrost has  been decoded after as much as 30,000 years.

It's   a long time.

Two, DNA has a really dense storage  capacity.

All of the information needed to make   you is stored in almost every single one of your  cells.

And, tiny spoiler, a single gram of DNA,   about half the weight of a sugar cube can store  at least 215 million gigabytes of data.

So DNA   naturally has the capability to store a ton of  information for a long, long time.

And trust me   when I say that I will absolutely need this video  footage to be on a drive somewhere in 500 years.

Your great, great, great, great grandchildren  need to know ACS Reactions information.

Okay, so while theoretically all the data  in the world could fit in a coffee cup,   it would probably be more easily accessible and a  bit more useful if we stored it in a space about   the size of a household refrigerator,  but that's still pretty impressive.

The third thing that DNA has that makes it  perfect for storing data is a code.

So just   like the zeros and ones of binary computer code,  DNA stores information in the very specific order   of four nucleotides, A, T, C and G. And just like  rearranging the zeros and ones in a binary code   differently can give you everything from Jurassic  Park to Chicken Soup for the Chemist's Soul,   rearranging DNA letters can give you everything  from dinosaurs to chickens, and the two strands   of DNA are complementary.

And this means that  an A on one strand of DNA will always bind to   a T on the other, and a C will always bind  to a G. So if you have one strand of DNA,   you can always rebuild the other.

And this  makes copying information really easy,   and it is absolutely critical to how everything  else in this episode is going to work.

It's also phenomenally important  to all of life here.

But, you know,   priorities.

Another plus of storing data in  DNA is that it's not going to become obsolete.

We're going to be sequencing DNA a lot longer  than we ever used floppy drives.

And if it's   implemented carefully and thoughtfully, DNA  data storage has the potential to be way more   environmentally friendly than continuing to  produce hard drives out of metal and plastic.

The funny thing is that in 2018 when I  graduated from grad school, our gel imager   was still using Windows 98 and it only had a  floppy drive.

So in fact, I used floppy drives   until 2018.

There are going to be five big  steps for storing and accessing data in DNA.

First, you're going to need to turn the zeros  and ones of computer file language into the As,   Ts, Cs, and Gs of DNA language.

That will happen  in your computer.

Second, you need to physically   print out a strand of DNA that encodes those  specific As, Ts, Cs and Gs.

This is going to   happen on a DNA printer.

Then you can go store  your DNA in a tube somewhere.

Third, when you're   ready to retrieve your data, you need to access  the file or set of files of interest from your DNA   storage tube.

Fourth, when you want to read the  information you've accessed, you have to take that   DNA molecule and then read out the sequence of As,  Ts, Cs and Gs that are on it.

That happens in a   machine called a sequencer.

Fifth and finally,  you can then turn that text file of As, Ts, Cs   and Gs into zeros and ones again, and that happens  once again in your computer.

Okay, so now let's go   from theory to reality.

I'm going to convert  our very first ACS Reactions video into DNA.

So to do that, I first have to scroll way back  to the bottom of the ACS Reactions page to here,   in 2014.

It was a long time ago.

So I took that  video file and I wanted to turn it into binary   code.

So remember that at their core, computer  files are strings of zeros and ones.

And when I   did this for just the very first frame of ... Oh,  my computer is so dirty.

So I converted just the   first frame of that video into zeros and ones,  and it was thousands of pages of zeros and ones,   thousands of pages, truly.

I thought I was  going to print it all out, but then I didn't   want to kill all the trees.

So it's a lot  of zeros and ones is what I'm telling you.

So for fear of crashing my computer, I didn't  do the whole video because it's not necessary.

But now we need a method to turn all of  these zeros and ones into As, Ts, Cs,   Gs, and one of the easiest ways to do that is  to take pairs and say maybe 00 would equal A.

01 would equal T. 10 would equal C, and  11 would equal G. So you could take the   information you have and it would actually  be compressed by half because you'd go for   two characters down to one character, but that's  actually not very efficient or reliable.

Instead,   a team in 2017 used information theory principles  to compress and error correct their data.

Now look, it's complicated, but it's kind of  like zipping files on your computer to make   them smaller.

It's better to zip your files before  you store them as DNA.

This isn't going to happen,   right?

So much DNA.

The team was able to  store a full computer operating system,   an 1895 French film, Arrival of a Train  at La Ciotat, a $50 Amazon gift card,   a computer virus, a pioneer plaque,   and a 1948 study by information theorist Claude  Shannon in a little over 14 million DNA bases.

To put that into context, your human genome is  3.2 billion bases long.

14 million bases could   fit onto a spec way smaller than the period  at the end of a sentence.

With this method,   the team suggested that they could pack 215  petabytes of data into a single gram of DNA,   which was way more dense than any other  effort at the time.

And to give that a   little bit of context, 250 petabytes is like  1.5 billion Facebook photos.

It's a lot of   data.

That actually doesn't sound like enough.

The theoretical upper limit of DNA data storage   is a zettabyte in a gram of DNA.

That is 1  trillion gigabytes.

Even Elon Musk doesn't   have that much storage space, probably.

So  let's say we use this method to turn our   zeros and ones of our video file into As, Ts, Cs  and Gs.

Now we have a file of As, Ts, Cs and Gs.

So how do we turn that text file into a strand of  DNA?

Well, I'll tell you that this is where our   production budget runs out.

I can code stuff.

I  can't afford to print the DNA.

So you have to take   the nucleotides that are in a text file on your  computer and print them out as strands of DNA.

You can think of it like an inkjet printer but for  DNA.

Now, you can do this a couple of different   ways, but the most common and standard for a long  time has been phosphoramidite chemistry.

To get   how this works, we have to take a little bit of  a closer look at the As, Ts, Cs and GS that make   up the DNA code.

They're each different types  of nucleosides strung together on a backbone of   sugars and phosphates.

A nucleoside plus one  of these phosphates is called a nucleotide.

And these phosphates are key.

To build a strand  of DNA, cells connect the phosphate group of   one nucleotide to a free hydroxyl group on the  carbon of the next nucleotides sugar ring.

So   you can attach phosphate to sugar, phosphate to  sugar, phosphate to sugar over and over to build   a big long chain.

But unlike a normal nucleotide,  a phosphoramidite molecule has a phosphite in the   place of the phosphate.

That phosphite contains  one fewer oxygen atom than the normal phosphate,   and it can't connect to the next nucleotide.

This means it's considered protected.

If it's   at the end of a chain of nucleotides, you can't  add another, you can't keep growing that chain.

However, those phosphites can be easily  removed when necessary.

So to print out DNA,   you can start with your first letter  as a phosphoramidite.

Remember,   in our DNA data storage scenario,  this will stand for zeros and ones.

But in the actual thing that you're printing out,  it is a molecule called adenine or A for short.

Now, imagine that you have a phosphoramidite A and  you want to attach a G as the next space.

Well,   that phosphoramidite A is protected so you can't  add a G. Nothing will stick to it.

So in order to   add that G, you have to de-protect the phosphite  group on the A and then wash the Gs over it.

But   this means that only one G will bind because  the Gs themselves are protected.

So this way   you don't accidentally get a string of like G, G,  G, G, G, you just get an A and then your one G.   So now you have a two letter long strand of DNA, A  and G. To add the next letter, maybe a T, you can   de-protect the G and then wash a bunch of Ts over  it so that you get A, G, T and so on and so on.

So slowly, base by base, you can  turn your text file of As, Ts,   Cs and Gs into a real strand of DNA.

So now your files are strands of DNA.

You can place them into a tube just like  this and store them for hundreds of years.

But what about when you need to read that data  back out?

So first you need to find the piece   of DNA that you're looking for.

The whole point of  this storage method is that it's really dense.

So   you're not going to give each file its own tube.

You're going to put a bunch of files together   all in one tube.

But finding what you're looking  for in that tube of DNA can be really difficult.

Imagine that you've taken every book in the  world, torn out all of the individual pages,   and then tossed them together into an empty  swimming pool.

Finding the book you want to read   is going to be really difficult and you don't want  to have to look at every single page to do so.

One way to find just what you're looking  for is using a process called PCR,   or polymerase chain reaction.

Basically  it's a method that allows you to make lots   and lots of copies of a very specific piece  of DNA.

So you take your mish-mashed soup   of DNA files and you add two small pieces  of DNA to it that's going to specify the   one piece that you actually want to read.

Then you copy that piece billions of times,   so it overwhelms everything else in your soup.

It's like taking a bowl of chicken noodle soup   and duplicating the carrots so many times that  you have a five ton pile of carrots with a tiny   bowl of soup underneath.

If you just want to eat  carrots, it's a lot easier to do now.

So same   with your DNA.

If you pick a piece of DNA from  the tube to sequence, you're more likely going   to be looking at the DNA you want than the DNA you  don't want just because there's more of it around.

But this process can be kind of annoying since  you could use up your soup just to retrieve one   file.

The chicken and noodles and broth under  the carrot pile are basically useless now.

So   one group is embedding the DNA in little spheres  of silica to make the retrieval easier.

As a test,   they encoded 20 different images into DNA.

They  then put little DNA barcodes on the outside of all   of those spheres.

These were little tags hanging  off that would allow you to pick out the images   that you wanted.

So these barcodes corresponded to  keywords about the image, things like orange and   cat for an image of a tiger.

Then they could use  fluorescent tagging to sort out just the spheres   that they wanted to sequence.

And I know this  sounds a little bit like magic, so let me explain.

Because the tags they put on the outside of the  spheres are DNA, they can use the complementary   nature of DNA to make tiny matching tags that will  bind to only the tags that they're looking for.

Things like CAT.

You then stick a fluorescent  molecule onto those matching tags and then mix   them in with the sample of spheres.

They'll bind  to the tags on the images they're looking for.

So   now the images you want will have a fluorescent  tag on them and the images that you don't will   not have a fluorescent tag on them.

You can use  a super special sorting machine to separate the   fluorescing spheres out from the ones that aren't  fluorescing.

So now you have a group of just the   files you were looking for.

The team was able  to accurately pick out the image they wanted   from 20 images stored in DNA, and they say that  this could work for up to 10 to the 20th files.

That's a lot of files.

So now you found your file.

How do you read it?

Well, you can sequence the   DNA, which just means reading the order of the As,  Ts, Cs and Gs in that strain of DNA.

And there are   two main methods of doing this.

The first is by  building a copy of the DNA and the second is by   pulling that DNA through a tiny pore.

So the first  method takes advantage of the complementary nature   of DNA that we've talked about before in this  episode where an A always binds to a T and a C   always binds to a G. The sequencing machine looks  at a strand of DNA of unknown sequence.

In order   to determine its sequence, it slowly builds a copy  of it base by base.

So each new base you add gives   off a different color of fluorescent light, say  red for C or yellow for A.

A very sensitive camera   can pick up this light and turn it into a DNA  sequence that you can then read on a text file.

The other common method of DNA sequencing uses a  pore embedded in a membrane.

The sequencer slowly   pulls the DNA through that pore, and as it does  so, the DNA strand disrupts an electrical signal   running through that membrane.

Each nucleotide is  shaped a little bit differently, so it disrupts   the signal a little bit differently.

And you  can read those disruptions back out as A, T,   C or G. Now this kind of sequencing, because it  gives you your information back in real time,   can also be used for the data accession  problem that we talked about earlier.

If you start reading a piece of DNA and  it doesn't look like the file you want,   you can have the sequencer kick it out of the  pore and keep looking for the file that you do   want.

But no matter which way you sequence it, you  can then use whatever computer algorithm you use   to turn zeros and ones into A, T, C and G back  into the zeros and ones of your computer file.

So digital data in DNA sounds awesome,  but we're all still lugging around hard   drives.

So what needs to happen for us to  store all of our information in just this?

I mean, at least theoretical DNA tastes good.

Well, as you might have noticed already, all of   these methods are a little bit bulky and time  consuming.

At the moment it takes about a week   to order printed DNA from most of the companies  offering that service.

Then it takes about a half   hour of prep and a couple hours of sequence to get  that info back out when you want it.

And there's   also some computational time in there too.

So this  is all good for files that you don't need often,   but it's not great if you just want to show  someone the cute picture you took of your   dog yesterday.

Look at how cute she is.

DNA data  storage would be great for things you don't have   to access often replacing stuff like tape backups.

It's already primed to be a great archival method.

But the other thing we need to overcome  is cost.

Nobody's got enough money to   print all of the digital data that  we have.

Well, maybe someone does.

Currently it would cost $1 trillion to write one  petabyte of DNA.

Again, that's like 6 million   Facebook photos.

That's way too expensive  to make it accessible.

One way to make it   affordable might be to try new methods of writing  out that DNA.

EDS or enzymatic DNA synthesis is   one promising new method.

Like it sounds, this  uses an enzyme rather than a typical chemical   reaction to add a base to a growing nucleotide  chain.

Like the phosphoramidite synthesis,   you start out with your first nucleotide and then  slowly build your strand one by one.

But EDS uses   an enzyme that's carrying the next nucleotide  that you want to add to the growing chain.

Once it adds it, it then protects the  end so more nucleotides can't bind,   then it does this with the next nucleotide and  the next nucleotide and the next nucleotide.

So   this could be useful for making longer and  longer strands of DNA since the previous   phosphoramidite method starts to peter  out after about a few hundred bases.

The   combination of new DNA printing methods, the  ability to write longer strands of code and   decreasing sequencing costs means that by  2030, DNA synthesis could reach affordable   rates for long term storage, something around a  dollar a terabyte each for reading and writing.

Okay, so real talk in my professional geneticist  opinion, I really do believe that this is going   to be a reality for archival data storage in about  the next decade.

Sequencing costs are coming way   down.

DNA printing costs are coming down, and it's  a stable, compact, efficient storage solution for   the massive amount of data that we are all  creating every single day.

And the potential   for doing extra cool things like storing data in  DNA inside living bacteria or sending stable DNA   messages out into the solar system just makes the  sci-fi loving part of my brain sing.

So bring it   on, Elon.

I am ready to put your computational  data storage capacity to shame with DNA.

It's snack time.