Scientists have long known that living cells use a complex system of signals to sense their environment and to transmit this information internally and to their neighbors. Specific signaling molecules, their concentration, and the way this changes over time are some of the factors that go into this system.
Although simple in principle, the system turns out to be extraordinarily powerful and complex. Which is why decoding it is hard. One problem is the difficulty of spotting the signaling molecules and measuring the way their concentration changes.
For example, when gamma radiation damages mammalian cells, it triggers the release of a protein called nuclear p53. This is released in many rapid pulses, a signal that causes the cell to pause operation to check for damage (a process called cell cycle arrest).
However, UV radiation causes a longer single pulse, which immediately triggers cell death. But the total amount of p53 released can be the same in both cases.
Modern molecular sensors cannot spot this difference. It’s rather like listening to a radio show with a Morse code receiver—you can tell if the transmitter is working, but not what it’s broadcasting.
So biologists desperately need a better way to measure these molecular signals.
Enter Jackson O’Brien and Arvind Murugan at the University of Chicago. These guys have developed a way to measure changes in molecular signals using a powerful form of molecular computation. They say their approach creates the building blocks for a new way to study and exploit cell signaling: “Our work lays the foundation for temporal pattern recognition through analog molecular computation.”
The emerging technology behind O’Brien and Murugan’s work is a form of DNA computing that synthetic biologists have great hopes for. The process is based on the way that one piece of single-strand DNA can displace another in a double-strand DNA, a technique that can be precisely controlled using well developed tools
These tools can precisely control the rate and reversibility of these “displacement strand reactions” over many orders of magnitude. So this creates switch-like behavior—the reaction is either on and off. And combining several different switches makes logic operations possible.
That, in turn, paves the way for all kinds of computational tasks. Researchers have shown how displacement strand reactions can perform complex calculations and even mimic the behavior of deep-learning networks.
O’Brien and Murugan’s contribution is to outline the DNA circuitry that can sense the presence of specific signals and the way they change over time.
Pulsatile signals vary in several ways. The period of pulses—the interval between them—can change. The length of the each pulse can vary—pulses with the same period can be short or long, for example. This is known as the duty fraction—the fraction of time the pulse is “on.” And of course, the number of pulses can change.
Importantly, the total amount of signal can be the same even when the period, duty fraction, and number of pulses varies wildly.
The new work is to design the molecular machinery that can measure each of these features individually and independently. “We demonstrate the decoder for each of these temporal features, one at a time,” say O’Brien and Murugan.
And the results look promising. The researchers have simulated the behavior of their circuits and say they work well: “We demonstrate our design principles using abstract chemical reaction networks and with explicit simulations of DNA strand displacement reactions.”
There are challenges ahead, of course. The circuits can look for predetermined changes in molecular signals, but more flexibility would be useful. “It would be interesting to develop molecular circuits that can learn relevant temporal features dynamically as in machine learning approaches,” suggest the researchers.
And the circuitry does not yet measure changes in the amplitude of the signal, which can be another important feature.
Beyond that, the next stage is to build this circuitry and put it into practice. That, of course, is an ongoing challenge for synthetic biologists in general. Past successes seem to show that researchers with skills that span the wet/dry lab divide have the greatest successes, because they can move quickly to test new ideas.
The rewards should be huge. O’Brien and Murugan speculate that their molecular computer could have dramatic applications. They imagine a DNA origami pill that delivers drugs only when it receives a specific pattern of signals.
For example, a cell’s inflammatory response and its adaptive immune response trigger different signal patterns of the transcription factor NFkB. A pill could be programmed to recognize just one of these and release its payload accordingly.
That may be some way off. Nevertheless, DNA displacement strand logic is an exciting technology with huge potential.
Ref: arxiv.org/abs/1810.02883 : Temporal Pattern Recognition Through Analog Molecular Computation
AI-powered 6G networks will reshape digital interactions
The convergence of AI and communication technologies will create 6G networks that make hyperconnectivity and immersive experiences an everyday reality for consumers.
The power of green computing
Sustainable computing practices have the power to both infuse operational efficiencies and greatly reduce energy consumption, says Jen Huffstetler, chief product sustainability officer at Intel.
Using data, AI, and cloud to transform real estate
AI can enable business transformation to deliver positive outcomes for clients and propel sustainability goals, according to Sandeep Davé, chief digital and technology officer at CBRE.
How this Turing Award–winning researcher became a legendary academic advisor
Theoretical computer scientist Manuel Blum has guided generations of graduate students into fruitful careers in the field.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.