Using analog computation circuits, engineers design cells that can compute logarithms, divide and take square roots

May 15, 2013 by Anne Trafton
MIT engineers have created synthetic biology circuits that can perform analog computations such as taking logarithms and square roots in living cells.

MIT engineers have transformed bacterial cells into living calculators that can compute logarithms, divide, and take square roots, using three or fewer genetic parts. Inspired by how analog electronic circuits function, the researchers created synthetic computation circuits by combining existing genetic "parts," or engineered genes, in novel ways.

The circuits perform those calculations in an analog fashion by exploiting natural that are already present in the cell rather than by reinventing them with , thus making them more efficient than the pursued by most synthetic biologists, according to Rahul Sarpeshkar and Timothy Lu, the two senior authors on the paper, describing the circuits in the May 15 online edition of Nature.

"In analog you compute on a continuous set of numbers, which means it's not just black and white, it's gray as well," says Sarpeshkar, an associate professor of and computer science and the head of the Analog Circuits and group at MIT

Analog computation would be particularly useful for designing cellular sensors for pathogens or other molecules, the researchers say. Analog sensing could also be combined with digital circuits to create cells that can take a specific action triggered by a threshold concentration of certain molecules.

"You could do a lot of upfront sensing with the analog circuits because they're very rich and a relatively small amount of parts can give you a lot of complexity, and have that output go into a circuit that makes a decision—is this true or not?" says Lu, an assistant professor of electrical engineering and computer science and biological engineering.

Lead author of the Nature paper is MIT postdoc Ramiz DanielJacob Rubens, a graduate student in microbiology, is also an author of the paper.

Analog advantages

Sarpeshkar has previously identified thermodynamic similarities between analog and the chemical circuits that take place inside cells. In 2011, he took advantage of those similarities to model biological interactions between DNA and proteins in an electronic circuit, using only eight transistors.

In the new Nature paper, Sarpeshkar, Lu and colleagues have done the reverse—mapping analog onto cells. Sarpeshkar has long advocated analog computing as a more efficient alternative to digital computation at the moderate precision of computation seen in biology. These analog circuits are efficient because they can take in a continuous range of inputs, and they exploit the natural continuous computing functions that are already present in cells. In the case of cells, that continuous input might be the amount of glucose present. In transistors, it's a range of continuous input currents or voltages.

Digital circuits, meanwhile, represent every value as zero or one, ignoring the range of possibilities in between. This can be useful for creating circuits that perform logic functions such as AND, NOT and OR inside cells, which many synthetic have done. These circuits can reveal whether or not a threshold level of a certain molecule is present, but not the exact amount of it.

Digital circuits also require many more parts, which can drain the energy of the cell hosting them. "If you build too many parts to make some function, the cell is not going to have the energy to keep making those proteins," Sarpeshkar says.

Doing the math

To create an analog adding or multiplying circuit that can calculate the total quantity of two or more compounds in a cell, the researchers combined two circuits, each of which responds to a different input. In one circuit, a sugar called arabinose turns on a transcription factor that activates the gene that codes for green fluorescent protein (GFP). In the second, a signaling molecule known as AHL also turns on a gene that produces GFP. By measuring the total amount of GFP, the total amount of both inputs can be calculated.

To subtract or divide, the researchers swapped one of the activator transcription factors with a repressor, which turns off production of GFP when the input molecule is present. The team also built an analog square root circuit that requires just two parts, while a recently reported digital synthetic circuit for performing square roots had more than 100.

"Analog computation is very efficient," Sarpeshkar says. "To create digital circuits at a comparable level of precision would take many more genetic parts."

Another of the team's circuits can perform division by calculating the ratio of two different molecules. Cells often perform this kind of computation on their own, which is critical for monitoring the relative concentrations of molecules such as NAD and NADH, which are frequently converted from one to the other as they help other cellular reactions take place.

"That ratio is important for controlling a lot of cellular processes, and the cell naturally has enzymes that can recognize those ratios," Lu says. "Cells can already do a lot of these things on their own, but for them to do it over a useful range requires extra engineering."

That extra engineering included modifying the circuits so that they can compute with inputs over a range of 1 to 10,000—much wider than the range of a naturally occurring cell circuit.

"It's nice to see that frameworks from electrical engineering can be concisely and elegantly mapped into synthetic biology," says Eric Klavins, an associate professor of electrical engineering and adjunct associate professor of at the University of Washington who was not part of the research team.

The researchers are now trying to create analog circuits in nonbacterial cells, including mammalian cells. They are also working on expanding the library of genetic parts that can be incorporated into the circuits. "Right now we're using three of the most commonly used transcription factors in biology, but we'd like to do this with additional parts and make this a generalizable platform so everyone else can use it," Lu says.

"We have just scratched the surface of what sophisticated analog feedback circuits can do in living cells," says Sarpeshkar, whose lab is working on building further new in . He believes the new approach of what he terms "analog synthetic biology" will create a new set of fundamental and applied that can dramatically improve the fine control of gene expression, molecular sensing, computation and actuation.

Explore further: How a molecular Superman protects the genome from damage

More information: "Synthetic analog computation in living cells", www.nature.com/nature/journal/… ull/nature12148.html

Related Stories

Researchers build a toolbox for synthetic biology

Aug 03, 2012

For about a dozen years, synthetic biologists have been working on ways to design genetic circuits to perform novel functions such as manufacturing new drugs, producing fuel or even programming the suicide ...

Mimicking cells with transistors

Sep 28, 2011

As the world has become less analog and more digital -- as tape decks and TV antennas have given way to MP3 players and streaming video -- electrical engineers’ habits of thought have changed, too. In ...

Cell-inspired electronics

Feb 25, 2010

(PhysOrg.com) -- A single cell in the human body is approximately 10,000 times more energy-efficient than any nanoscale digital transistor, the fundamental building block of electronic chips. In one second, ...

Recommended for you

Structure of an iron-transport protein revealed

3 minutes ago

For the first time, the three dimensional structure of the protein that is essential for iron import into cells, has been elucidated. Biochemists of the University of Zurich have paved the way towards a better ...

Over-organizing repair cells set the stage for fibrosis

53 minutes ago

The excessive activity of repair cells in the early stages of tissue recovery sets the stage for fibrosis by priming the activation of an important growth factor, according to a study in The Journal of Ce ...

Cell architecture: Finding common ground

Oct 16, 2014

When it comes to cellular architecture, function follows form. Plant cells contain a dynamic cytoskeleton which is responsible for directing cell growth, development, movement, and division. So over time, changes in the cytoskeleton ...

User comments : 9

Adjust slider to filter visible comments by rank

Display comments: newest first

210
1 / 5 (3) May 15, 2013
Indeed: Analog computation has much more inherent power and precision over our digital rendition of what natural or programmed data and events should look like. We needed digital for our instrumentalities thereby allowing us to manipulate and quantify our experiences/science/nature. Even our mathematics shows this: Most, or very, very many processes in nature are analog, so few truly digital. So we find our descriptors in our differential (calculus, for example)) bias used to described a primarily analog world. So then our digital machinations/machines revealed so much to us during the rudimentary and pubescent development of our understanding of science. Now, mother nature reveals to us that we may use her technology, the technology of life, the ULTIMATE technology to unlock supreme mysteries of the cosmos - both inner and outer space. EXCELLENT! And I submit, what better medium for bringing about our mastery of optical computing than these flexible tools?
word-to-ya-muthas
axemaster
5 / 5 (1) May 16, 2013
Whaaa? I hardly dare to ask, but... inner and outer space?

And I know it's pointless to seriously respond, but analog computation does not have greater precision than digital. In fact it's vastly inferior because it can only be resolved down to the noise floor. That's why we usually try to convert analog sensor signals to digital as fast as possible - each operation you add creates more and more noise until the output becomes completely useless.
210
1 / 5 (3) May 16, 2013
Whaaa? I hardly dare to ask, but... inner and outer space?

And I know it's pointless to seriously respond, but analog computation does not have greater precision than digital. In fact it's vastly inferior because it can only be resolved down to the noise floor. That's why we usually try to convert analog sensor signals to digital as fast as possible - each operation you add creates more and more noise until the output becomes completely useless.

No, analog IS the ultimate digital. WE have only mastered the technology for representing natural phenomenon and daily stats using our first breakthrough technology. BUT the faster digital gets and the more precision you give it, the more analog-like it becomes. Do the math yourself. The noise floor is a natural boundary into which a signal may descend. Detecting the signal is the issue, not resolving it. We use D/A converters to gain a digital representation of a signal because we have digital electronics to help us -
210
1 / 5 (3) May 16, 2013
When you integrate a displacement equation you divide it into discrete washers or rectangles, and you ALWAYS have error, inherent, no matter how small you make the limits of integration. BUT, but, as you do make them, the limits smaller, you 'smooth' the curve back into its analog best form. Likewise digital: The better you digitize, the closer and closer you get to analog. Our best computers will be analog units using either light/optics or natural radiation as their compute-transmission medium, even, as their fabric where now we use varying backplanes embodying fibre channel, etc, etc, etc. Analog IS digital, without the sampling error and analog is digital ad infinitum. Our limit has been the machines we had to use bounded by our growing inventiveness. Rudimentary analog was first, then digital, then super digital, then analog returns in the ultimate form/no error digital - which is analog.
word-
210
1 / 5 (3) May 16, 2013
The noise floor is THE NOISE FLOOR for all: in RF it is ambient, unfiltered manmade and natural electrical and RF signal, in digital it is THE SAME. If you have to CONVERT anything, you inject error as well as noise. Digital without need for conversion or conversion noise, is, ANALOG.

INNER SPACE is where our molecular tools lie. Our methods of lithography and the nature of physical matter may not allow us to actually 'write' circuits as we do now. So we may need to 'grow' or 'place' the components as nature does in living systems. Our knowledge of, say dark matter and the cosmological constant began with thought experiments relevant to Einstein's work at the atomic level and his understanding of relativity. E=MC2 was revealed at the atomic level and then was seen to apply to the stars. Thought proceeded from the inner cosmos, where our LHC is even to this day and extends outward to explain the OUTER cosmos at the macro-scale.
wod-
alfie_null
not rated yet May 16, 2013
210 - The point of this research is to create something that does a sufficient job of emulating some ideal process. Not that our bodies are full of analog processes. Get the distinction?

For most criteria, digital emulation is always going to be superior to analog emulation. You can easily control precision and accuracy. You can represent functions easily, some of which would be impossible to do in the analog domain.
210
1 / 5 (3) May 16, 2013
210 - The point of this research is to create something that does a sufficient job of emulating some ideal process. Not that our bodies are full of analog processes. Get the distinction?

My statements were not aimed at mere human processes, but at the use of the bio-molecular tools. THey do in fact offer us Unbounded precision whereas the digitizing process has to limit this process to bring a range of data into human view/consumption. Digital is more like a series of rapid snapshots whereas analog is a HD motion picture and that is the EXACT best analogy. Conversion losses bred error and in the realm of large data flow, you leave a lot of data on the cutting floor.

word-
210
1 / 5 (3) May 17, 2013

For most criteria, digital emulation is always going to be superior to analog emulation. You can easily control precision and accuracy. You can represent functions easily, some of which would be impossible to do in the analog domain.

No, "A square wave is a non-sinusoidal periodic waveform (which can be represented as an infinite summation of sinusoidal waves), in which the amplitude alternates at a steady frequency between fixed minimum and maximum values, with the same duration at minimum and maximum. The transition between minimum to maximum is instantaneous for an ideal square wave; this is not realisable in physical systems." Digital emulation just makes the data amendable to our DIGITAL computers! An analog system made of bio-molecular parts would need NO such conversion and would leave out no data and add no error. here: http://en.wikiped...are_wave
Hence digital signal processing is and has always been a stop-gap to going back to powerful analog.

word-
rsklyar
1 / 5 (2) May 20, 2013
Other swindling engineers at the Massachusetts Institute of Technology and Children's Hospital Boston are stealing unboundedly in British cheating journal Nature Materials when dependably covered by their Vice President for Research Maria T. Zuber as http://ru.scribd....g-timely