cellular supremacy
Will the capabilities of cellular computing ever surpass those of classical computers? Lewis Grozinger and colleagues argue in favor, developing a notion they call "cellular supremacy."
“The promise of synthetic biology lies in its potential to provide new substrates for computation, production, pollution control and medical diagnosis (among many areas), and to harness the “wetware” inside the living cell for human-defined purposes” (Grozinger et al.) (emphasis added).
SynBio enthusiasts have been touting the transformative potential of engineered biological systems since the field was conceived about 25 years ago. Firms like McKinsey, CapGemini, and Boston Consulting Group predict that synthetic biology will add billions of dollars added to the economy over the next decade, and yet the field seems to have been stuck in an eternal stage of prototyping.
Why should researchers, the government, and private companies continue to invest in synthetic biology? Grozinger et al. argue that there exists a set of problems for which cellular computing is significantly better suited than traditional computing—an idea that they call “cellular supremacy” (an intentional parallel to quantum supremacy).
This proposition goes beyond “the implementation of conventional computations with unconventional computing substrates”—a common goal of first-wave SynBio, in which much research centered around building modules like digital logic gates in biological systems. Rather, the authors write, “supremacy must be derived from the fact that the type of computation performed by a conventional computer is qualitatively different to that executed by living systems” (Grozinger et al.) (emphasis added). In other words, cellular computing will find its edge in problems that cannot be solved with classical computing methods.
As an aside: thinking about computation, we often consider the most commonly-used (and best fleshed-out) model of computation, the Turing Machine (which classical computers are based on). Turing defined a particular class of problems that were solvable with his model, but cells, the authors argue, have been “optimised over billions of years of evolution” to perform very different computational tasks than the deterministic, digital ones Turing describes.
“[L]ogic circuits offer a relatively bland computational palette compared to the richness of biology. For this reason, future developments in cellular computing should focus on models of computation that both accommodate and exploit the natural abilities of the cell, and avoid forcing biological systems into artificial (and often unsuitable) architectures” (Grozinger et al.) (emphasis added).
Computer science includes many alternative models of computation that share more in common with cells than the Turing Machine does, but “[t]he fact remains that we still currently lack any formal framework within which to argue that a cell computes, according to any understood model of computation” (Grozinger et al.). That being said, the existing parallels between biological systems and alternative models of computation provide strong evidence that they are suited to solve a class of problems than classical computers cannot.
What will give SynBio-based solutions an edge over classical computing?
In short: their architectural richness and ability to accommodate and exploit noise. The authors present five specific advantages of cellular computing (but note that this is a non-exhaustive list). Here are a couple:
Noise
The cellular environment is messy. Biochemical signals are not nearly as clear as electrical signals; yet, cells exhibit reliable responses to these noisy signals. Unlike classical computers, cells not only tolerate, but harness, noise to compute. This capability could make them well-suited for implementing probabilistic algorithms, helping overcome the challenge of implementing non-deterministic computation in silicon substrates.
Concurrency
Some algorithms are not completely reliant on steps executing in a particular order—a property known as concurrency. In these cases, work can be distributed and executed across multiple processors simultaneously, significantly speeding up computation.
Cells do this naturally, at both the single-cell and population levels. In biochemical reaction networks within cells, for example, individual reactions need not take place in a specific order for a cell to behave in a particular way. Individual cells or populations may be used to compute solutions to problems described by models of distributed and concurrent computation from computer science (Petri nets, process algebras, population protocols, and more).
Others
The other special qualities of cells that the authors list are:
Reconfigurability: cells’ ability to change their internal structure in response to internal or external signals
Representation: cells use non-binary representations of signals
Evolution: populations of cells evolve over time, seeking out novel solutions to problems
“[T]he cellular environment is a radically different computing substrate than silicon. Although this difference might make cells unsuitable for computational tasks traditionally dominated by conventional computers, it could also offer opportunities to explore more unconventional models of computation” (Grozinger et al.).
thoughts
The paper makes a convincing case for why cellular systems are good (and unique) information processors, but I think it could go a step further in its descriptions of the five advantages of cellular computing to show precisely how cells are superior to classical computers in these respects. In most cases, it seems like an analog has already been developed in silicon (i.e. I don’t think the paper fully demonstrates a need for cellular computing).
That being said, experiments have yet to show the full range of cellular advantages described in the paper, which might be what keeps the authors from making even bolder claims about cellular supremacy. DNA cloning (building synthetic DNA constructs and putting them inside cells) is not automated, making the experimental process slow and tedious; I would venture to say that this is one of the big reasons that synthetic biology as a field has seen slower progress than initially forecasted. It will be interesting to see what direction cellular computing takes over the next decade or two and whether or not we achieve ‘cellular supremacy’ in one domain or another.
(I’ve only covered the five advantages of cellular computing that the authors propose, but they also include extensive sections on implementations of cellular computers and their applications.) Find the paper here:
Grozinger, L., Amos, M., Gorochowski, T. E., Carbonell, P., Oyarzún, D. A., Stoof, R., Fellermann, H., Zuliani, P., Tas, H., & Goñi-Moreno, A. (2019). Pathways to cellular supremacy in biocomputing. Nature Communications, 10(1), 5250. https://doi.org/10.1038/s41467-019-13232-z