You’ve heard the saying “flash is the new disk and disk is the new tape,” which traces its origins back to Jim Gray*. But what if DNA-based data storage could one day replace tape? And what if computing could be done “in chemistry” rather than “in silico,” utilizing the substrate of synthetic DNA?
This is the bold vision of Catalog. The Boston-based startup founded in 2016 by MIT scientists has a plan to make DNA data storage commercially viable — and they’ve got former IBM exec Dave Turek leading the way. This week, the company publicly announced it had secured $10 million in Series A funding and named Turek CTO.
“Bringing David Turek on-board, as our CTO, is game-changing for the company and is an important milestone for the field of DNA-based data storage and computation technology,” said Catalog CEO and Co-founder Hyunjun Park. “The new capital will be used to reach commercialization.”
On the occasion of Catalog clinching its latest funding round (Horizons Ventures led the oversubscribed round, joined by Airbus Ventures), HPCwire spoke with Turek at length about Catalog’s goals, the underlying technology and why he’s excited to be leading the technical strategy and business development.
Turek comes to Catalog after more than two decades at IBM. An accomplished HPC strategist, Turek spearheaded a number of supercomputing projects at IBM, notably Blue Gene, Roadrunner (the first petascale system) and the pre-exascale CORAL systems (Summit and Sierra).
Catalog worked with UK company Cambridge Consultants to develop what Turek referred to as factory in a box to tackle the future of data storage. The room-sized block of industrial equipment is in contrast with almost magical-sounding abilities promised by Turek, who after his storied career in HPC is now getting tutored in chemistry twice a week.
“Catalog invented a clever data encoding scheme,” Turek said, “but they also invented a machine using inkjet technology to print DNA to reflect that encoding scheme. It’s to the point now where it can be automated, if you will. And we can use conventional semi-off-the-shelf technology to print data in DNA, and by virtue of doing this, we can also compute on it as well.”
As for that inkjet-inspired DNA writer, Turek said, “we start out with a fairly conventional mechanical device passing a plastic medium through rollers like a regular printer at about 16 meters a minute. Instead of printing ordinary ink, the printheads print an ink composed of DNA that we’ve invented. [The machine] prints cells as dots on this plastic substrate. And then it gets an enzyme placed on top of it to cause a reaction. And the net-net is this sheet covered with dots goes through the machine and eventually all the DNA in it gets washed off and put into a solution where you can operate on that DNA. When I say operate on it — you can do things like search for particular bits, or particular data wrappers or strings of data and so on.”
“In a certain sense, it represents a confluence of hardware, software and chemistry in one box, where you take digital information, you pass it into the machine, and under software control, it gets rendered into biological construct DNA — by the way, it’s not active DNA, you’re not going to create life forms out of it, anything like that — and then you apply chemistry to it almost as you would think of applying computer operations to data, except now they’re chemical operations. And by virtue of that, you can do whatever you want to the data.”
The company says the DNA storage technology makes it possible to fit all the world’s data into the space of a coat closet.
As explained by Catalog, “Synthetic DNA is ideal for data storage purposes for reasons that go beyond its longevity. It can hold a million times the data in the same volume as what is offered by magnetic and solid-state media. It takes almost no energy to store, can be replicated easily and inexpensively once encoded, and is trivial to transport, as a thousand petabytes of data (one exabyte) in DNA form will be roughly the size of a sugar cube.”
The company adds that “current research efforts to develop DNA-based data storage struggle with the high cost and low synthesis speeds, limiting their near-term economic viability for storing meaningful quantities of digital information.”
As a proof of concept for its approach, Catalog recently encoded all of the English text of Wikipedia into DNA, storing 16 gigabytes of data — significantly more information than has been captured by other DNA storage projects, according to the company. Turek attests that Catalog’s technology is two orders of magnitude faster and cheaper than other DNA-based data encoding projects. “It has to do with the way we approach the application, the chemistry and data encoding and such,” he said. “We’ve done a lot to minimize the amount of chemical steps that are involved in doing this.”
If you have the contents of Wikipedia in your pocket or all the world’s data in your hall closet, how do you locate the specific data you want and wouldn’t it take a long time?
This gets to the most fascinating aspect of the chemistry approach to computing. Catalog says its proprietary encoding scheme enables direct computation by enzymes and other DNA molecules, facilitating massively parallel computation — and setting the stage for breakthroughs in search, inference and digital signal processing.
No matter the size of the data set, the data you want can be found in the same amount of time, said Turek. “The reason for that is the chemistry I run against a trillion molecules is the same chemistry I’m going to run against one molecule, and it all takes place in the same amount of time. I’m going to add more reagent and other stuff, but the time horizon to do chemistry against a trillion molecules is the same as it’s going to take against one. So the bigger the data set, the easier it is for me, to convince you that I’m finding things efficiently. Because the time horizon is the same.”
Explaining further, he said, “It’s sort of a variation on this idea of random access. So I can specify a pattern, if you will, that represents the part of Wikipedia that I’m looking for. And then I can automatically translate that into chemistry. That will excise those molecules that correspond to that part of Wikipedia and just sort those out for you and present them to you. It sounds somewhat like magic, I’m sure. But you look at the length of the molecule, you look at the composition of the base pairs as you compose it, and you look at the key, the building blocks we have that helps you decode that or encode it, and you can go through that process of decoding all this stuff and then select exactly what you want out of it. So no, I don’t have to look at the whole thing. I can simply say, here’s the section I want, here’s what it looks like, go fetch; and I go fetch with chemistry.”
The storage medium is also highly mobile. The entirety of Wikipedia fits in a thimble-sized container and can be easily cloned. “If you say, I want a million copies of Wikipedia, that’s actually a pretty trivial endeavor; I can produce that very cheaply and quickly,” said Turek.
The invention in DNA storage has been mainly on the write side because in the biological world, there is no need for writing, Turek explained. “You simply take DNA from living organisms and sequence them using a commercial sequencer. So the marketplace as it exists today has put invention and money and development into reading DNA. We did the reverse, we put money and invention into writing DNA. And we’re leveraging the investments the industry’s made on the read side. But we think there are many, many orders of magnitude improvement that we can get on reads as well. So right now there is technically a bit of an imbalance between read speed versus write speed, but that can be remedied.”
Currently, Catalog’s technology is capable of megabits-per-second write speeds. Turek indicated that it is on track to reach gigabyte-per-second speeds on the write side, at which point the company will switch focus to optimizing read speeds with the goal of achieving parity for read and write times.
An innovator and big thinker, Turek gets a charge from conceptualizing and delivering paradigm-changing technologies, and that’s what brought him to Catalog. He describes how IBM in the last couple years, prompted in part by interest from the Department of Energy, has been working toward a new era “of bits, qbits, and neurons, of artificial intelligence working with quantum, working with classical computing and all coming together.”
“As I was learning more about DNA computing, I thought there was a role for biologically-based computing here as well — not neuromorphic, which is you know, sort of a mimicking of brain-like function in silicon — but real biological computing and storage,” said Turek. “What attracted me to it was the potential to get aligned with something that has the potential to be disruptive in the marketplace, but also something working very much at the leading edge, where we could be transformative in terms of what we can invent — the volume of IP — and really set the industry in the right direction. So it was pursuit of an opportunity to really lead an emerging technology into the marketplace.”
Asked if Catalog is planning to offer a complete solution, Turek said the commercialization strategy is still coalescing. “Right now the answer would be yes,” he said. “And by complete solution, that would be read, write, robotically control, software infused. And by the way, there are applications for machine learning as well. But yeah, from start to finish, we’re going to take the data, we’re going to write it, we’re going to store it, we’re going to read it, and we’re going to do it efficiently in a way that’s suitable to your needs.”
Turek said Catalog is working on proof of concepts with government agencies and a number of Fortune 500 companies across sectors such as media and entertainment, banking and finance, and oil and gas. “Some of these are who’s-who HPC players, but some are non-HPC players — many names you would recognize — who are doing commercial stuff but are awash with data and they’re trying to figure out how to store and how to operate and how to minimize costs.
“We’re at what I would say is the beginning of the commercial beginning.”
* Tape is Dead Disk is Tape Flash is Disk RAM … – Jim Gray (PPT)