Microsoft Wants To Use DNA For Cloud Data Storage (technologyreview.com)

Last July, researchers from Microsoft and the University of Washington said that they had successfully encoded about 200 megabytes of data onto synthetic DNA molecules. The company is now planning to take the technology commercial. "Computer architects at Microsoft Research say the company has formalized a goal of having an operational storage system based on DNA working inside a data center toward the end of this decade," reports MIT Technology Review. "The aim is a 'proto-commercial system in three years storing some amount of data on DNA in one of our four centers for at least a boutique application,' says Doug Carmean, a partner architect at Microsoft Research." From the report: Internally, Microsoft harbors the even more ambitious goal of replacing tape drives, a common format used for archiving information. Major obstacles to a practical storage system remain. Converting digital bits into DNA code (made up of chains of nucleotides labeled A, G, C, and T) remains laborious and expensive because of the chemical process used to manufacture DNA strands. In its demonstration project, Microsoft used 13,448,372 unique pieces of DNA. Experts say buying that much material on the open market would cost $800,000. According to Microsoft, the cost of DNA storage needs to fall by a factor of 10,000 before it becomes widely adopted. While many experts say that's unlikely, Microsoft believes such advances could occur if the computer industry demands them.

Comments

Better than SongSmith

[omnichad]

At least Microsoft Research is doing something more useful than this.

Re:Better than SongSmith

[repka]

I watched the whole thing.

Re:Better than SongSmith

[omnichad]

Hope you don't suffer the same fate again.

So that's how we create the Andromeda Strain

[BarbaraHudson]

Combining DNA in a near infinitude of combinations is going to require the whole thing be handled like a level 5 biohazard, because you'll be producing prions and sequences of DNA that aren't found in nature, and to which we have zero resistance. Makes a hard disk or an ssd crash look positively benign, since all you'll lose is your data.

It would also be great for making known bio-weapons - just record multiple instances of sequences for, say, smallpox, then break the seal in a populated space.

Re:So that's how we create the Andromeda Strain

[omnichad]

I don't know their methods, but with a little overhead in your data size, you could probably render the resulting sequence biologically inert.

Re:So that's how we create the Andromeda Strain

[The Real Dr John]

No, you need a vector to do something with the DNA or RNA, such as a viral vector. These are all very specific in their targets, and they are used every day in thousands of labs around the world to shuttle DNA and RNA into cells. But it is not a good digital storage mechanism for big data.

This will never be used to store digital information. You could do it much more easily with a much simpler, man made system of chemical 1 and 0 s in any type of medium, silicon or biological. The triplet code in DNA/'RNA requires a lot of complicated enzymatic machinery to copy, store, replicate and read. It will not make a simple, reliable and robust system for humans trying to store and retrieve lots of information.

Re:So that's how we create the Andromeda Strain

[BarbaraHudson]

Not really. Just look at how much junk DNA we carry around, and yet our DNA still works.

Re:So that's how we create the Andromeda Strain

[BarbaraHudson]

DNA is used to make all sorts of fun things, like proteins, prions, and viruses. To be able to store arbitrary information, the dna encoding will have to allow for the creation of sequences not found in nature, or it won't achieve the desired density.since the data would have to be encoded in longer sequences of naturally-occurring dna. Either way, you've got the dna equivalent of a 3d printer.

Re:So that's how we create the Andromeda Strain

[Rick Schumann]

Which is why I said this. I said it not just to be funny.
For even greater potential hazard, be sure to thoroughly encrypt the data before committing it to strands of DNA.

Worst idea ever. Nice job, Microsoft.

Re:So that's how we create the Andromeda Strain

[interkin3tic]

There's no biosafety level 5, it only goes up to four.

More importantly, no to the rest of it.These are not going to be living things, they're going to be dried nucleotides on paper most likely. There is going to be no transcription or translation and creation of proteins. First of all that's much more difficult and doesn't happen on its own. Second that would defeat the point of data storage. Having the DNA doing stuff would cause its degradation and loss.

It's like saying "don't download that encyclopedia on that external hard drive! It might achieve sentience!" Nothing is happening to the data either way, and in both cases, making "life" would be impossible.

Life requires a lot more than DNA. There are some plant viruses IIRC that can reproduce simply by injecting their DNA or RNA sequence into plant cells. But I didn't hear about any such human viruses. Viruses require protein machinery to take over the cell in addition to their DNA. You synthesize the smallpox genome and inject it into your veins, you're not going to develop smallpox.

... I mean, I wouldn't try that myself, but my fears over doing that are purely illogical.

The smallpox genome is also a 186 kilobase sequence. It's not something that's sure to show up with much frequency even if all the DNA in MS's storage were to get into your cells. If anyone knows a way of calculating how much DNA you'd need to synthesize at random before you came up with those specific 186000 nucleotides, I'd be very interested, but I'm guessing it's a lot.

Finally, synthesizing nucleotides is old hat. The scale and cost is the new thing here. You want to synthesize a smallpox genome? You can do that already. There aren't even any laws against it yet! It's going to cost you a lot and again, DNA itself wouldn't do shit besides freak people out, but you can. It'd be much easier just to find smallpox itself. But either way, there's nothing completely new here besides it's now cheap and fast enough to consider doing for data storage.

Quit getting spooked by biology.

Re:So that's how we create the Andromeda Strain

[Rick Schumann]

All it would take is someone having the bright idea of creating a bacteria to serve as a 'backup system', to replicate the DNA strands.

Re:So that's how we create the Andromeda Strain

[omnichad]

But the junk DNA doesn't seem to have any ill effects against the core programming. If the entire dataset was encoded to be "junk" then it won't do anything.

Re:So that's how we create the Andromeda Strain

[sjames]

Yeah, but would anyone be that surprised if the latest MS install disk encoded to a hyper virulent super Ebola?

Re:So that's how we create the Andromeda Strain

[omnichad]

Right now they're on the Extend phase of adopting DNA. Extinguish isn't for another cycle or two.

Re:So that's how we create the Andromeda Strain

[Anonymous Coward]

It's not all junk because it doesn't appear to do anything. At the minimum it's spacers to separate functional areas, but it's often involved in moderating gene expression rates, either by obstructing, exposing, or acting as an on-ramp for enzymes.

Without regulation you'd never get tissue differentiation.

Re:So that's how we create the Andromeda Strain

[BarbaraHudson]

Given that this could in theory produce DNA more dangerous than anything found in nature, you'd damn well better have a level 5.

There are some plant viruses IIRC that can reproduce simply by injecting their DNA or RNA sequence into plant cells. But I didn't hear about any such human viruses

The flu virus hijacks your body's cells to reproduce pretty much the same way. Surely you've heard of the flu virus.

Prions also aren't living, just chunks of protein, but they cause mad cow disease, and they can be distributed by eating infected meat, by blood transfusion, and experiments have show that it can be distributed and successfully infect mice in aerosol form.

Re:So that's how we create the Andromeda Strain

[BarbaraHudson]

Try enough different combinations and you're going to get some that are not junk. It's like the million monkeys typing on a million typewriters.

Re:So that's how we create the Andromeda Strain

[interkin3tic]

The flu virus hijacks your body's cells to reproduce pretty much the same way. Surely you've heard of the flu virus.

Citation needed. Yes influenza has DNA, but it also has an envelope in order to get into the cells. DNA just floating around in the air isn't going to get into a cell. DNA floating around in your bloodstream even is going to get shredded by your immune system. So please, prove to me that naked DNA outside of cells can cause viruses inside the cells.

The suggestion that there should be a biosafety level 5, higher than an intact ebola virus, for DNA by itself is patently absurd.

I'm not sure what prions have to do with anything either. This is not going to be biochemical reactions with transcription, translation, chaperones, and proteins. Proteins are not going to be generated.

The company working with MS, Twist biosciences, make DNA on silicone dots.. Not using living cells with translation machinery. It sounds like there's no proteins involved in the synthesis or the "writing" of data.

Reading the data will most likely be through nanopore technology. Again, no translation, generation of proteins, or any other biological process going on there.

So tell me, where the hell is a prion or any other protein going to be generated in that process?

Finally, I'm not convinced it's a mathematical certainty that the sequences for any pathogen WOULD be generated. They're extremely long sequences. If there were legitimate concerns that viruses could spontaneously generate from protein-free, sterile DNA, and if MS were unable to keep themselves from injecting their data into their veins and if there are viruses whose DNA can magically worm their way into human cells, it's entirely possible for MS to automatically scan for DNA being written that happens to be similar to a pathogen's genome and disrupt it. Say "only write that half in this block, put a spacer with a bunch of stop codons, then put the other half somewhere else.

Again, you're describing something that is less likely than a regular hard drive achieving sentience and going skynet.

Re: So that's how we create the Andromeda Strain

[Anonymous Coward]

Valid question: forgive my terminology. virureses are free form floating around. Everywhere. They mutate also. The flu, many times a year.

Is it possible that some floating DNA can randomly come in contact with some viruses or whatever else nessicary , over time to create something new? We do live in a giant Petri dish and its how specimen evolve?

However unlikely, is it possible? And if it is possible, can it be a netter of time to combine all this stuff? It maybe inert, it maybe something bad (infectious or deadly to bio)

Re: So that's how we create the Andromeda Strain

[interkin3tic]

Viruses aren't actually "everywhere." Sterility is actually pretty easy if human hands and bodily fluids aren't involved, as they wouldn't be here. So there would be no viruses or bacteria.

Likewise, the DNA isn't going to be spraying everywhere any more than computing using a hard drive involves platters and magnets flying all around the air.

Viruses and bacteria evolve because they're much more than nucleotides, and because they duplicate their own nucleotides with some degree of error. DNA by itself is inert and does nothing aside from fall apart at extremely long timescales. It's not going to be replicating itself and introducing errors.

"However unlikely, is it possible" is in fact not a valid question in this case since likelihood does in fact matter. "However unlikely, is it possible that a hard drive with an encyclopedia on it could achieve sentience?" If you can say "no" to the question, then you can say "no" to your question. If you have to say yes to the hard drive question, then realize there's little use to pondering extremely unlikely hypothetical situations. Finally, if you're worried about viruses, be worried about non-fictional ones. Ebola is real and is out there already. You have billions of cells WITH proteins and DNA. The possibility that a world-ending virus is going to form inside you is far more likely than one being kicked out without any proteins by MS.

Re: So that's how we create the Andromeda Strain

[BarbaraHudson]

Hard drives have never been hermetically sealed. Even the ones that use helium today leak.

Re:So that's how we create the Andromeda Strain

[BarbaraHudson]

First, viruses are not alive, they're just DNA or RNA with a protein coating. Neither the DNA, RNA, or the coating are produced by the virus - they're produced by the host cell. Viruses are like bricks - they're inert objects made in a factory.

Viroses neither "eat", nor "excrete". No metabolism. They cannot move on their own. They literally go wherever the environment carries them.

We know what ebola is, and how to treat it - and people can survive without treatment, same as some (a very few) survive for decades with HIV because they have a natural immunity . We don't know what you get when you just randomly generate dna sequences. Most will be harmless, but there is absolutely no guarantee that some won't, just as it's guaranteed that some will find their way into the environment (there is no hard drive, not even the new helium ones, that is perfectly sealed).

If you don't see the relevance of prions, you should look it up.

So yes, viral DNA does indeed seemingly "magically worm its' way into human cells", and your statement "I'm not convinced it's a mathematical certainty that the sequences for any pathogen WOULD be generated. They're extremely long sequences." by demanding mathematical certainty, shows you are foolish, same as the statement that pathogens have extremely long sequences is full of shit.

DNA? OJ?

[Anonymous Coward]

PAROLE?

What the fuck!

They should create a whole new company for this

[Rick Schumann]

Microsoft should just spin off a whole new company specifically for this project.
They can name it The Umbrella Corporation.

Re:They should create a whole new company for this

[cyberchondriac]

Or Weyland-Yutani. This is how you get weaponized xenomorphs.

No jokes about wanking?

[edittard]

No jokes about wanking yet? This place isn't what it used to be.

Re: No jokes about wanking?

[Anonymous Coward]

We're all busy running a train on your mom, sorry.

Re:No jokes about wanking?

[michelcolman]

Only 200 megabytes? I once stored more than 8 petabytes in DNA in a cloud in the bathtub! Should have filed a patent, I'd be rich now...

Re:No jokes about wanking?

[Wolfrider]

That was YOU?! I couldn't use that bathroom for 45 minutes after you were done! Next time try a courtesy spray, you insensitive clod!
:B

/ damn green clouds hanging around...

Re: No jokes about wanking?

[Anonymous Coward]

It was a spray. On your mom.

talk about virus vulnerabilities...

[dAzED1]

I know MS has long had a history of being especially prone to viruses based on basic design decisions, but this is almost retro!

Re:talk about virus vulnerabilities...

[rogoshen1]

retroviral indeed.

Re:talk about virus vulnerabilities...

[Wizardess]

Um, yes, it gives the concept of viral data a whole new meaning. Be careful with your data patterns so that none of them generate a killer disease in the Microsoft cloud in 2072.

{^_-}

Viruses?

[saccade.com]

This will give a whole new meaning to the phrase "Computer Virus".

Success depends on simplicity

[Gravis Zero]

Using DNA for data storage is a real possibility but what they need to do more than anything is to simplify the encoding and decoding so that it is both speedy and more importantly, costs next to nothing. What this really means is building complex molecular machines which is something we have yet to manage. It might take 50 years before we manage to figure out how make complex molecular machines but the result will be amazing in the same way that graphics rendering thought up 50 years ago is amazing on modern GPUs.

Re:Success depends on simplicity

[fustakrakich]

I'm wondering, what do you feed it (Miracle-Gro?), and how often you have to water it.

This reminds me of IBM doing something like that

[Anonymous Coward]

Posmanipulating atoms for the atomic hard drive:
https://youtu.be/f2OKVQmODC8

They even made a movie:
https://youtu.be/oSCX78-8-q0

Unlikely??

[wbr1]

I doubt it is unlikely. I am no expert, but the cost of genetic sequencing has dropped tremendously in the PAST 20 years? Why should we expect the costs of creating our own sequences to not become easier with time and new methods?

I am not here to debate the ethics of such a thing, but to hand wave it away as if it will not happen is likely to be hubris.

Re:Unlikely??

[Anonymous Coward]

Cost isn't the problem - accuracy is. We've got cheaper at sequencing by massively increasing parallelism, but you then need to non-deterministically reassemble the information back into order. That's costly too, and doesn't give 100% accuracy. We're making progress, but I don't think read errors on a par with hard disks are likely for a long time.

I think creating sequences to order is much more reliable than reading them back again.

For the last time, stop ripping off Star Trek.

[Anonymous Coward]

...and come up with your own ideas. Or at least rip off some of the GOOD ideas instead of all the stupid ones.

Less touchscreens and voice controls and more holodecks and subspace communication and transporters and warp drives.

Microsoft Wants To Use DNA For Cloud Data Storage

[Anonymous Coward]

I want a pink pony.

Nonsense, DNA is tasty

[goombah99]

DNA will be eaten in short order. other than RNA or writing it on sand on a tidal flat, they could not have picked a less viable long term storage media.

Microscopic elephant in the room: nucleases

[omaha393]

The potential is exciting, but the big issues are cost of synthesis, which is fixable, and stability, which is unfortunately more pervasive. DNA is super stable, but only when its uncontaminated. Nucleases (enzymes specifically designed to cleave DNA) are everywhere and certain chemical or UV exposure would ruin the base sequences irreparably. Basically if the hard drive isn't in sterile conditions it risks being easily destroyed, so a misplaced sneeze could delete everything. Also cytosine (C) has a tendency to convert to Uracil (U, basically a substitute for T in RNA). I'm sure engineers will probably fix all these issues, but there's always nucleases that evolve to counter anti-nuclease activity.

Re:Microscopic elephant in the room: nucleases

[interkin3tic]

Engineers have fixed those issues by such technology as "not sneezing into your test tubes" and "freezers" and "not storing DNA in sunlight."

If your DNA is out in the open collecting dust and microbes, you're asking for it to be contaminated. DNA is more than stable enough to work on without using so much as gloves. Consider that Lincoln's DNA from when he was assassinated is still readable despite having been stored at room temperature for over a century in non-sterile conditions. RNA? Sure, that would get shredded pretty rapidly. DNA though is extremely robust. The people at microsoft and twist biosciences aren't stupid.

Nucleases are everywhere, but you wouldn't be rubbing data DNA on your shoe prior to working with it. They're not going to "evolve" to counter anti-nuclease activity because the DNA is going to be kept away from the bacteria with the nucleases.

DNA as data storage but still no safe OS

[TheOuterLinux]

Proprietary software and for profit enterprising at work. Besides, agents Scully and Molder (X-Files) can tell you all about "DNA storage." Surely our government in real life wouldn't encourage people to store their DNA? Ancestry.com, 23 And Me, CDC scaring baby boomers into blood testing for hep C, and now M$ DNA storage, to which I'm sure there's a patent to profit billions from if it works. It's probably just some fany micro-electrophoresis like what they use in forensics and with a way to interpret the passing through the gel as data. You won't be ripping that "CD" anytime soon. DRM at an insanely high level. But, if you can't copy and paste in a reasonable amount of time or accuracy, what's the point? File management becomes impossible. They already want to kill desktop software, do they have to go after the filesystem too?

Right

[Anonymous Coward]

Storage Media, huh?
Cost a nickel a byte. (That's 50 billion dollars per Terabyte)
Write speed: oh, around 1-10 Megabyte per day
Read speed: around 4 GB/day
MTBF: probably around 10 minutes per DNA linkage break.
Definitely, sign me up.

This sounds familiar

[newbie_fantod]

Microsoft has finally caught up to Johnny Mnemonic.

But does it compile?

[sabbede]

If yes, what does it do?

What's the data life measured in Teloyears?

[Anonymous Coward]

Will enzymes will be repairing the data or will there be a Raid of these to remove the bugs.

God

[Anonymous Coward]

Turns out God was the greatest programmer of all