EDIT: Yes well for those who haven't read the book the reference is to the sexual transmission of data. Implied (but not confirmed) by nano-tech, however no implementation details were provided (as I recall), so encoded DNA might have been a possibility as a medium. Just sayin'.. /shrugs/
My first thought was, how long did it take to write and read this data? I don't expect exceptional speed, but I do wonder about throughput and seeking.
FTA: While it took years for the original Human Genome Project to analyze a single human genome (some 3 billion DNA base pairs), modern lab equipment with microfluidic chips can do it in hours.
OK, so assuming 1 hour for 3 billion bases, it's 1000 hours for 3 trillions bases (3 Terabits) or 1 million hours for ~3 petabits (or 400TBs of data). Yeah, that's a long time, roughly 100 years :)
Reading DNA is advancing fairly quickly. Illumina's HiSeq X Ten system produces 6Tbases per day. (Capex and opex/year are both in the 8 digit range.)
Seeking can be performed through entirely different means, as DNA is content-addressable. You can put in a magnetic bead attached to a strand complementary to what you're seeking, and pull that out of the mix. This is still a physical process that can take quite a bit of time.
Read and write were a couple of days each. Would be faster now. That said, we did 650kB; the article is misleading as they are talking about copies of the same information.
My quibble is with the claim that 3TB (3.5") hard drives are the "densest storage medium in use today". Yes, the article is two and a half years old, but by then 32GB MicroSD cards had been around for a couple of years. 100 such cards would exceed the capacity of a 3TB hard drive and weigh a tenth as much.
Yeah, the article isn't great. We go through how we calculated densities; we didn't use hard drive enclosures, but did include thickness for HD platters in calculating density. In general, magnetic tape mostly wins as they are so thin.
Because this is a discovery, not an invention. If they could make high-performance biological storage devices the size of hard drives, you don't think they would be doing this already?
This probably required a machine significantly larger than most desktop computers, with a pricetag in the $100,000+ range. That and it has a R/W speed of ~30Kb/s...
It will be in our hard drives when the technology has been developed to make that happen.
I mean, graphene is already looking to be a far more efficient and powerful alternative to silicon. Why isn't that in our processors yet? It's because creating a logic gate and creating an x86 CPU are entirely different things.
The same way that encoding a bunch of data into DNA and replicating it a few billion times is entirely different from making an on-demand biological data storage device.
Same reason we don't heat our homes with nuclear weapons. More capacity than we need by orders of magnitude of orders of magnitude, but rather hard to extract & manage just what you want.
I said that all-you'll-ever-need capacity is meaningless if you can't get what little you need without extreme efforts. It's a "drinking from a firehose" problem.
A question that's somewhat akin to asking, 'How robust are humans to flying projectiles?' ;-)
Of course, it's all about the energy. I'm not a biologist, so I won't speak to the details of how resilient DNA is to specific types of exposure or why, but I am a physicist and I can tell you that if you blast DNA with high energy radiation, like gamma rays, it's gonna have a bad time. That said, if we want to talk about robustness in terms of what it's likely to be exposed to, then it's pretty darn robust. We are exposed to a wide band of EM radiation on a constant basis, and can even withstand exposures on the higher end without becoming 'corrupted'.
I'd also note that standard, non-specially designed electronics aren't very resilient to radiation either. Not to mention, if you could reliably read/write data using DNA, I'd expect that redundancy would become... easy. Security on the other hand...
It's weird when you read about tech advances that are so profound, you have to stop and consider the philosophical implications. I have a dime on my desk that probably weighs 2 grams... something that small could contain more raw data than a person could possibly read in their entire life... that could contain 1.4 million hours of cd quality audio. Insane!
Considering and quantifying the some total of your actions in life is largely philosophical in nature. Yes its not writing a paper on the applications of Friedrich Nietzsche in 2015, but everyone also isn't an academic.
Much like figuring out how much to tip your waiter is a mathematical exercise. Just one that might not shake the foundations of academic mathematics.
Just because territory is well travel, doesn't mean everyone has walked there.
"Wikipedia trivia: if you take any article, click on the first link in the article text not in parentheses or italics, and then repeat, you will eventually end up at 'Philosophy'."
Particularly impressive when you consider the human genome is only ~700 MB.[1] So this is like a million times as much data as human DNA stores, but in the same DNA format. Impressive.
They chose to encode one bit per base (rather than two) to give them flexibility in encoding the binary values to avoid particularly tricky sequences for DNA synthesis or sequencing (such as ones that have many repeating bases, e.g. AAAAA, which are known as homopolymers, or ones that have high GC content).
Not astronomical gains. Ignoring the technical challenges of a biological system like DNA, base 4 can encode the same data in exactly half as many characters, so capacity would double.
We are talking about data, which is the log of the number of combinations, like any measure of information. If this interests you, definitely look into basic information theory, and then move onto coding theory.
Take, for example, 32bit integers vs 64bit integers (unsigned for simplicity). Two 32 bit integers can represent exactly the same number of combinations that a single 64 bit number can. Sure, there's an exponential number more combinations in a 64bit integer than a 32 bit, but the number of combinations is not how the storage capacity is measured.
The difference is exponential. 1 quaternary digit can encode twice as much information as 1 bit, yes. But 2 digits can store 16 possible combinations, whereas 2 bits can store 4.
X digits of binary vs quaternary can represent 2^x vs 4^x possibilities.
Even at 10 digits, that is 2^10 = 1024 vs 4^10 = 1048576.
Doesn't DNA have a halflife of like 13 years or something? Wonder what the error rate is on 700TB of DNA just sitting there for a year. I guess you could engineer a system with multiple redundancies and checksums, with that kind of density raid2 or raid1000 doesn't make a difference.
Haha that paper was published today! Are you the author?
This is pretty interesting. I'd love to read the paper but it looks like I can't download the article yet since its not finished uploading to the site.
DNA storage is pretty silly. The claim is 5.5petabits/mm^3 with a 100x redundancy. The problem is that massive storage is only as good as your ability to encode/decode into it.
Let's be clear: this work encoded just 5.27 megabits. It's stored in what's basically a large molecular hash table where each piece of key-value data is replicated a million times for redundancy. Each piece then read 100x to correct for the -abundant- errors in each piece. So they encoded less than a megabyte.
The problem with encoding information into DNA is that writing serial polymers accurately is difficult and slow. In this paper they're using an inkjet printed DNA array. It takes a day to make them, resulting in a bandwidth of:
(5.27 Mbits) / (24x60x60 sec) = 66 bits/sec
Reading is a little faster. The fastest system, the HiSeq 2500 reads 120Gbits raw in 27 hours. Factoring in the necessary 100x redundancy, one has a -maximum- read rate of:
So for 5.5 petabits it would take 16,000 years to write the data into a cubic mm but only 16 years to read it at the current rate.
If we get a little scifi, and assume we build programmable
polymerases and get nanopore (direct read) sequencing. Even then, physics limits you to something like 1000 read/writes per sec per pore/polymerase. Instrumenting to these will probably limit per-feature size to being larger than 100micron on a fabricated chip, giving us an ultimate read/write limit around:
(2cm/100um)^2 x 1000 bit/sec = 40Mbit/sec
for a giant 2cmx2cm chip. With the necessary error-correction and redundancy, it's probably going to cap out around 1Mbit/sec at best.
Those 700TB take about 2months to read/write at these rates, and we'll have much-better solid state storage technologies by the time we figure out how to all that with DNA.
Largely agreed, though I think information storage in sequenced polymers in general is fairly interesting, but we are long ways off. Also, there are applications for DNA storage that are somewhat interesting when weight becomes very important (space travel) or biocompatibility (barcoding food ingredients).
This is pretty old, and not sure why it's here again today. That said, I was an author on the paper, and I'm happy to answer questions during a useless meeting I have to attend in 30 minutes.
What kind of IO bandwidth can you currently get from DNA? Does reading from it damage the DNA? What kind of hurdles need to be overcome to bring this to market and are they hurdles we can overcome in the near future?
edit; Hey looks like you answered most of these else where, so thanks.
75 comments
[ 2.9 ms ] story [ 155 ms ] threadEDIT: Yes well for those who haven't read the book the reference is to the sexual transmission of data. Implied (but not confirmed) by nano-tech, however no implementation details were provided (as I recall), so encoded DNA might have been a possibility as a medium. Just sayin'.. /shrugs/
OK, so assuming 1 hour for 3 billion bases, it's 1000 hours for 3 trillions bases (3 Terabits) or 1 million hours for ~3 petabits (or 400TBs of data). Yeah, that's a long time, roughly 100 years :)
http://xkcd.com/1217/
Seeking can be performed through entirely different means, as DNA is content-addressable. You can put in a magnetic bead attached to a strand complementary to what you're seeking, and pull that out of the mix. This is still a physical process that can take quite a bit of time.
This probably required a machine significantly larger than most desktop computers, with a pricetag in the $100,000+ range. That and it has a R/W speed of ~30Kb/s...
It will be in our hard drives when the technology has been developed to make that happen.
I mean, graphene is already looking to be a far more efficient and powerful alternative to silicon. Why isn't that in our processors yet? It's because creating a logic gate and creating an x86 CPU are entirely different things.
The same way that encoding a bunch of data into DNA and replicating it a few billion times is entirely different from making an on-demand biological data storage device.
Any idea on how long this will take to get to market, and whether it's being worked on?
Storage companies buy by the petabyte.
The argument that we don't need more storage doesn't work.
I said that all-you'll-ever-need capacity is meaningless if you can't get what little you need without extreme efforts. It's a "drinking from a firehose" problem.
Of course, it's all about the energy. I'm not a biologist, so I won't speak to the details of how resilient DNA is to specific types of exposure or why, but I am a physicist and I can tell you that if you blast DNA with high energy radiation, like gamma rays, it's gonna have a bad time. That said, if we want to talk about robustness in terms of what it's likely to be exposed to, then it's pretty darn robust. We are exposed to a wide band of EM radiation on a constant basis, and can even withstand exposures on the higher end without becoming 'corrupted'.
I'd also note that standard, non-specially designed electronics aren't very resilient to radiation either. Not to mention, if you could reliably read/write data using DNA, I'd expect that redundancy would become... easy. Security on the other hand...
Sneeze once and everyone has a copy? :)
Much like figuring out how much to tip your waiter is a mathematical exercise. Just one that might not shake the foundations of academic mathematics.
Just because territory is well travel, doesn't mean everyone has walked there.
I know a more then one prof who'd agree with that statement.
"Wikipedia trivia: if you take any article, click on the first link in the article text not in parentheses or italics, and then repeat, you will eventually end up at 'Philosophy'."
[1] http://stackoverflow.com/questions/8954571/how-much-memory-w...
with base2 we get 2^5 = 32 possible combinations
with base4 we get 4^5 = 1024 possible combinations
2*(2^x) != (4^x)
Take, for example, 32bit integers vs 64bit integers (unsigned for simplicity). Two 32 bit integers can represent exactly the same number of combinations that a single 64 bit number can. Sure, there's an exponential number more combinations in a 64bit integer than a 32 bit, but the number of combinations is not how the storage capacity is measured.
X digits of binary vs quaternary can represent 2^x vs 4^x possibilities.
Even at 10 digits, that is 2^10 = 1024 vs 4^10 = 1048576.
This is pretty interesting. I'd love to read the paper but it looks like I can't download the article yet since its not finished uploading to the site.
Let's be clear: this work encoded just 5.27 megabits. It's stored in what's basically a large molecular hash table where each piece of key-value data is replicated a million times for redundancy. Each piece then read 100x to correct for the -abundant- errors in each piece. So they encoded less than a megabyte.
The problem with encoding information into DNA is that writing serial polymers accurately is difficult and slow. In this paper they're using an inkjet printed DNA array. It takes a day to make them, resulting in a bandwidth of:
(5.27 Mbits) / (24x60x60 sec) = 66 bits/sec
Reading is a little faster. The fastest system, the HiSeq 2500 reads 120Gbits raw in 27 hours. Factoring in the necessary 100x redundancy, one has a -maximum- read rate of:
(120 Gbits / 100) / (27x60x60 sec) = 12 Kbits / sec
So for 5.5 petabits it would take 16,000 years to write the data into a cubic mm but only 16 years to read it at the current rate.
If we get a little scifi, and assume we build programmable polymerases and get nanopore (direct read) sequencing. Even then, physics limits you to something like 1000 read/writes per sec per pore/polymerase. Instrumenting to these will probably limit per-feature size to being larger than 100micron on a fabricated chip, giving us an ultimate read/write limit around:
(2cm/100um)^2 x 1000 bit/sec = 40Mbit/sec
for a giant 2cmx2cm chip. With the necessary error-correction and redundancy, it's probably going to cap out around 1Mbit/sec at best.
Those 700TB take about 2months to read/write at these rates, and we'll have much-better solid state storage technologies by the time we figure out how to all that with DNA.
What kind of IO bandwidth can you currently get from DNA? Does reading from it damage the DNA? What kind of hurdles need to be overcome to bring this to market and are they hurdles we can overcome in the near future?
edit; Hey looks like you answered most of these else where, so thanks.