News New memory card uses DNA to store your data — Biomemory's card costs $1,100 to store one kilobyte of data

[Admin]

French startup Biomemory commercializes a storage technology that stores data within dried, manipulated DNA strands. Claimed to be first to market, the 1,000 Euros per kilobyte of storage is undeniably expensive.

New memory card uses DNA to store your data — Biomemory's card costs $1,100 to store one kilobyte of data : Read more

 

[usertests]

I want small glass/holographic discs that can store hundreds of terabytes or petabytes for millions of years.

 

[COLGeek]

I want small glass/holographic discs that can store hundreds of terabytes or petabytes for millions of years.

Just imagine the number of obsolete adapters to access those discs in the future...

 

[AgentBirdnest]

How does a user write data to the card? How is it read?
Or is it a completely nonfunctional conversation piece, and not able to interface with computers?

 

[USAFRet]

How does a user write data to the card? How is it read?
Or is it a completely nonfunctional conversation piece, and not able to interface with computers?

This isn't an actual product yet.
"The French biotech firm slates Biomemory Prime for a 2026 release"

 

[bit_user]

I think it's pretty amazing just that they got a test vehicle to work reliably.

While that is pretty nifty, a base-4 system should be able to easily surpass a base-2 or binary system in storage efficiency. Biomemory's DNA Encode does not do that— in fact, converting "hello" into regular old binary also spits out a 40-character result.

Could be lots of reasons for that. Perhaps it's meant as a form of parity or for bio-security (you wouldn't want a terrorist or rogue researcher downloading the DNA sequence of some hemorrhagic fever virus and using this to synthesize its DNA).

 

[stevefan1999]

This is already sufficient to encode an Ed25519 private key in. You can have practically 3 of them in 1K plus some redundant space for error correction (would be interesting to see if Reed-Solomon error correction works on DNA).

Useful as a heritage to the family's treasure, as it is not easily replicatable and is assumed to be truly unique over at least 100 years. Sure, someone can have stolen it, but nobody can clone your card, given you have not leaked the private key bits electronically. Easily the most expensive way to flex your e-Signature/Certificate.

How is it better than a Yubikey you might ask? I don't know. It's just cool to do it like that. Ain't nobody is insane enough to do $1000 on 1K DNA in the first place, right?.

But speaking of which, Yubikeys are easily under attack by world government and national level threat actors -- This shiny DNA card needs some specialized equipment to read and write. Again, its taking data safety to an extreme...

 

[AgentBirdnest]

The link in the article explains how writing/reading works.

Write: You give Biomemory (the company) the data that you want to have stored. They give that information to a lab in Germany, who produces the DNA and puts it into that little circular capsule embedded in the card. You get sent two cards.
Read: If you ever want to read the data, it has to be sent back to the lab. They'll pull the data (which is a destructive process) and they give you the DNA sequence, which you can decode yourself using Biomemory's DNA translator.

Totally impractical for the average user. :-D But really fascinating.

 

[thisisaname]

Write once read never?

Only 150 years! it would last longer with some good parchment and high quality ink.

 

[AgentBirdnest]

Rough estimates based on quick price checks; forgive me if I'm a bit off...

Tape - 0.2¢/GB
HDD - 1.5¢/GB
BDR - 2.5¢/GB
SSD - 5¢/GB
M-DISC - 10¢/GB
DNA - $1,100,000,000 / GB. :-D

 

[purposelycryptic]

How does a user write data to the card? How is it read?
Or is it a completely nonfunctional conversation piece, and not able to interface with computers?

You DON'T write to it. You give them your KB of data, and they assemble DNA into a pattern equivalent to it. It's strictly ROM. Except currently, they have the only devices to read it, too. So it's more like... OM?

 

[umeng2002_2]

I don't see the point if only the company can store and read data on it.

 

[slightnitpick]

The link in the article explains how writing/reading works.

Write: You give Biomemory (the company) the data that you want to have stored. They give that information to a lab in Germany, who produces the DNA and puts it into that little circular capsule embedded in the card. You get sent two cards.
Read: If you ever want to read the data, it has to be sent back to the lab. They'll pull the data (which is a destructive process) and they give you the DNA sequence, which you can decode yourself using Biomemory's DNA translator.

Totally impractical for the average user. :-D But really fascinating.

Yeah, I assumed it was something like this. Commercial DNA synthesis providers sell double-stranded DNA for less than 0.1 USD per base pair, so you can get 1000+ base pairs for on the order of $100 or less. This will be on the order of a couple of micrograms of DNA, some of the molecules will have errors, but on average they'll be sequence correct (think of it as a *huge* RAID with many trillions of parity disks).

Resuspend the tube that they ship you with water, transfer it to a filter paper ( http://wang.ucsd.edu/protocol/2. molecular cloning/2.3 Amplification/Shipping_and_Receiving_Plasmids_on_Filter_Paper.pdf ), let it dry, and store the filter paper in an air tight bottle with a desiccant, and you'll probably have something equivalent to what this company is selling.

When you want to read it, pull the filter paper out and rehydrate the DNA, then send it to a Sanger sequencing facility to get the read out for $5 or something around that.

To make things easier, the ends of the DNA can encode primer recognition sites using commonly available primers. The most common primers are generally provided free by the sequencing facilities, so if possible, use them to save some money.

There are current limits both for synthesis, and some for sequencing. For instance DNA sections which are rich in G/C content or A/T content do not synthesize well, and can skip some bases or prematurely terminate when sequencing. This is also true for DNA that creates secondary structure and long stretches of the same base, or has repeats. There are ways around this, but it sometimes can be more expensive or have length limitations. This will limit the maximum encoding to somewhat less than 4 bases per position.

I would guess that much of the 40-mer for the above "hello" encoding is for defined 'stuffer' sequences on the ends either for primer recognition sites, or just to prevent any degradation of the ends from impacting the text coding portion if these are synthesized as linear molecules and not cloned as circular molecules.This is not the case. It looks like they're using sequences of 8 to encode each digit. Maybe they're covering the entire Unicode specification? The first 3 of 8 seem to indicate letter, number, etcetera, while the next 5 presumably encoded the actual digit.

Ultimately the readability issue will be which encoding is being used.

 

[slightnitpick]

Thinking about this a bit more, and given the caveats I mentioned above.

They're charging $1,100 for up to 1024*8 DNA bases. I would guess they are likely cloning this into a plasmid vector to make it closed circular. They may or may not be making the vector clonally pure (which would be transferring the mixed population of the DNA assembly into bacteria, replicating the plasmids in that bacteria, and then sequencing individual bacterial colonies to isolate sequence-perfect plasmids). Regardless, this isn't a horrible price for up to 8192 base pairs of synthesized DNA. Given the length of synthesized DNA they're probably using something like Oxford nanopore technology to sequence the entire plasmid as one read. It's about $15-$20 to do that commercially today, and can be done relatively rapidly. Making the plasmid clonally pure would take at least 2-3 days, but a non-pure plasmid could be synthesized in under a day.

The 8 base pair "byte" that encodes each character consists of a leader sequence of 3 base pairs, which under ideal circumstances can code a maximum of 64 'pages', and a 5 base pair follower sequence which ideally can encode a maximum of 1024 individual characters per page (but realistically a bit less than this). Together this allows encoding a single Unicode plane of 65,536 characters, theoretically, and has the possibility of allowing more characters depending on whether the leader sequence allows dynamic 'byte' sizing along the lines of UTF-16.

 

[TerryLaze]

I want small glass/holographic discs that can store hundreds of terabytes or petabytes for millions of years.

But only if you have to spin them on a tabletop for them to play back the info.

 

[Krotow]

Soon will be available in Wish and AliExpress under 2TB name.

 

[Conor Stewart]

The 150 year lifespan isn't really a flex, yes it will beat SSDs and HDDs but there are other, cheaper long term storage solution if you don't want much storage.

FRAM is a good alternative, 512 kByte for just under £20 and that is on a breakout module. It can be used like SRAM chips but it is non volatile. The chip can be read or written with standard hardware (SPI) and it can apparently handle 10 trillion writes per byte and the SPI bus can run at up to 40 MHz with SRAM like data rates. Data retention is only 10 years at 85 Celsius, 95 years at 55 Celsius and 200 years at 35 Celsius.

Bare in mind the FRAM chip I am taking about is meant for data storage for microcontrollers or other small applications, larger and cheaper FRAM chips should be possible.