Monday, 14 December 2009
Monday, 19 October 2009
Volume 138, Issue 5, 4 September 2009, Pages 820-821
Available online 3 September 2009.
In the Preface of his book, Bray insists his work is not about single cells possessing consciousness. “I say repeatedly in the book as clearly as English words will allow that in my opinion single cells are not sentient or aware in the same way that we are,” he writes (p. ix). The key part of this claim, however, is the clause “in the same way that we are.” Bray is not equating cellular and human cognition. However, he is up to something profound and perhaps slightly sneaky. Rather than place consciousness on a pedestal that only humans can reach, Bray seeks to define it along a continuum of systems and states. In this view, consciousness is not an absolute, and by some definition cells can indeed be said to be aware of themselves and of their surroundings.
To explore these ideas, the book swiftly covers most of the important work to-date in the nascent field known as systems biology. It provides an excellent introduction to the field that is broadly accessible and would do well as required reading for graduate students in the biological sciences. Bray gives a colorful tour of diverse topics encompassing micro- and molecular biology, artificial intelligence, neuroscience, biochemistry, and robotics. In another context these subjects might have had less to do with one another—here, Bray convincingly shows how they are parallel fields in search of a central theory of cognition. His skill is in framing the basics of molecular systems biology by way of deep and thought-provoking questions: What is the basis for consciousness? To what extent are machines and cells really intelligent?
Single cells can exhibit extraordinarily complex behaviors. To respond to their environment, single-celled organisms such as the amoeba adopt a complex set of internal cell states that Bray likens to feelings of hunger, curiosity, and fatigue (Chapter 1). He points out that these very complicated behaviors make it difficult to argue that they are entirely hard wired—there is some dynamic learning even at the cellular level. To be capable of this, Bray proposes that the amoeba must possess an internal model of its environment and thus a primordial sense of self.
To understand these behaviors, Bray first turns to the field of artificial intelligence. To Bray, the key finding of artificial intelligence (with regard to biology) is that surprisingly simple man-made rules can give rise to extraordinarily complex behaviors. He touches repeatedly on this concept by way of numerous examples—citing artificial constructs such as the Game of Life, robotic vacuum cleaners, and even the 1980s video game Pac-Man. Bray also alludes to a potential link between consciousness and the ability to replicate. Although living systems are machines, they are the only machines that routinely undergo self-repair and self-replication. That replication is linked to awareness is an intriguing notion, and it echoes previous suggestions by John von Neumann, Daniel Dennett, and others. In fact, rudimentary self-replicating robots have now been constructed (Zykov et al., 2005, Nature 435, 163–164). It would be interesting to know how Bray incorporates such creatures into his worldview.
In the chapters on protein switches and signals, Bray seeks to understand and describe the molecular mechanisms that bestow cells with computational ability. To begin, he gives an excellent and eye-opening account of thermal diffusion as the basis for all of life. Although biologists often visualize cellular processes as wiring diagrams, linking together collections of genes, RNAs, proteins, or small molecules, Bray points out that these “wires” are really nothing of the kind. They are diffusing molecules, more akin to cell phone transmission towers than point-to-point wires. Later, Bray revisits thermal diffusion to point out not only that it establishes connections in the cell, but also that these connections can be noisy resulting in biological individualism (cellular free will, one wonders?).
A second key mechanism of cellular computation, according to Bray, is allostery. Allosteric proteins are switches, in which the protein's activity can be turned on or off through modification at an independent site. Such proteins, Bray argues, serve precisely the same function in cells as do neurons in brains or transistors in electronic circuits. He discusses many ways in which allosteric switch-like behavior can appear in biological systems, including enzymatic reactions and kinase-phosphatase signaling. Given many individual switches, Bray points out, it is very easy to store a near infinite number of cell states. He also makes the intriguing suggestion that protein complexes may serve as rudimentary memory devices whose states persist much longer than the lifetime of a single cell. Although the individual proteins within a complex must turn over, the complex as a whole can persist in the same active state for much longer periods.
Yet another principle of cellular computation might best be described as “co-activity between two connected switches.” According to the theory of learning known as synaptic plasticity, a synapse is more likely to be reinforced if both the axonal and dendritic sides of the synapse are excited synchronously or in rapid succession. Bray hypothesizes that this may be the crucial concept at work in ligand-receptor interactions—a provocative idea.
Chapter 6 is perhaps the most important of the book. In it, Bray moves from individual switches to consider how many switches together behave as a type of neural network. If the neural network analogy is accurate, it has some important predictions for biological systems. One of them is combinatorics—each gene will be controlled by a potentially very large and complex set of signals coming from other genes. Another one is redundancy: A neural network, and by analogy a gene network, should be robust enough to withstand removal of individual components. In support of the redundancy argument, Bray cites the finding that most genes do not produce any noticeable growth defect when removed from the genome. On the other hand, there is conflicting evidence in this regard. Guri Giaever and colleagues recently showed that, in fact, most genes are required for life. This requirement is not revealed by single gene knockouts in nutrient-rich conditions—it is revealed only in the particular stress conditions for which each gene is evolved to handle (Hillenmeyer et al., 2008, Science 320, 362–365).
Regardless, Bray's greater message is right on target. There is a type of neural network in every cell, involving connections at multiple levels including protein signaling, transcription, translation, and metabolism. Classical biology thinks in terms of sequential pathways of proteins and metabolites, but the reality is an interconnected “hairball.” To see evidence of this, one needs only to read about the networks of protein-protein, transcriptional, and genetic interactions that are now being systematically elucidated by numerous technologies. Proteins such as kinases and transcription factors form myriad connections with other molecules, over a wide spectrum of connection strengths (or in neural network terminology, “weights”). In any particular cell type or organism we observe one particular set of weights, but many different patterns of weightings are possible to produce the same function and no particular pattern is sacred.
Recent studies suggest that some metabolic networks and protein complexes may remain stable over evolutionary time, whereas informational networks involved in signaling and regulation may be dramatically reprogrammed, even over short evolutionary distances (Tuch et al., 2008, Science 319, 1797–1799). Such gross changes are consistent with the neural network model, in which the same function can be encoded by many alternative patterns of connections and weights “under the hood.” Given this property, it seems likely that cells can tolerate much more evolutionary drift in their gene and protein networks than has been previously appreciated. Such implications are profound and have not yet been adequately grasped by the biomedical community. We should not be so focused on mapping what proteins are wired to what other proteins. The reality may in fact be much more daunting: everything is connected to everything. It is the weights that matter.
One area of research that seems relevant to Bray's argument, but that was missing from it, is the field of biological computing. In work pioneered by Leonard Adleman, DNA and proteins are used explicitly as a highly parallel, man-made computer to solve computationally hard problems. Nowhere else has “the cell as computer” analogy been so direct and accurate. It also would have been nice to have had greater attention paid to the ongoing systematic efforts to decipher the cell's gene and protein networks. Investigators in the ‘omics sciences (genomics, proteomics, metabolomics, etc.) are working hard to decipher the patterns of molecular wiring and weights that Bray describes. It seems that Chapter 10 (Genetic Circuits) might be a natural place to include these topics. Finally, Bray's philosophical discussion of what it would take to construct a minimal cell would be bolstered by mention of recent work by several groups to engineer exactly such a cell based on genomes that are naturally small, such as that of Mycoplasma.
Ultimately, questions about cellular cognition are as much philosophy as biology. In this regard, Bray's view seems particularly well aligned with that of David Chalmers, who in his 1996 book The Conscious Mind: In Search of a Fundamental Theory credits all information-bearing systems with some level of consciousness but not necessarily with cognition or awareness. Reading Bray's musings at the intersection of philosophy and biology, one also wonders whether there is room for a field of “existential biology.” In existentialism, it is individuals who invent their own values and create the very terms under which they excel. What a perfect system of thought in which to encompass not only people as individuals but also artificially intelligent computer programs and nonhuman life forms including—if we can stomach it—single cells…
Friday, 16 October 2009
Minton, Treharne & Davies Ltd
Dear Mr Minton,
I wonder if you can help me please. I'm a science teacher based in Scotland trying to raise pupil interest in science through topical issues. We've been following the Trafigura slops-dumping incident and its consequences and would like to ask you a question, if we may.
On Tuesday 13 Oct 2009 solicitors Carter-Ruck issued a press release stating that a judge recently "fully endorsed Trafigura’s long-maintained position that the Probo Koala’s slops simply could not have caused deaths, miscarriages, stillbirths, birth defects or other serious or long-term injuries."
This followed thorough analysis, they continued, by "20 independent experts in what has been by far the most detailed consideration of these matters anywhere in the world".
This statement, as you know, flatly contradicts the report you wrote on 14 September 2006. In that report you concluded that the slops were "a mixture of spent NaOH, naphtha, free water and an organo-metallic catalyst used in the Merox reaction. This mixture was also likely to contain several reaction intermediates and by-products, including but not limited to Na2S, NaHS, NaSR, RSSR and H2S."
These compounds, you continue, "are capable of causing severe human health effects through inhalation and ingestion. These include headaches, breathing difficulties, nausea, eye irritation, skin ulceration, unconsciousness and death. All of these effects were as reported in the incident."
Can you help my students and me to understand this drastic difference of expert opinion? Is the chemistry you lucidly described in your report simply wrong?
Did you make mistakes in your analysis, and if so could you explain for the benefit of our science students exactly what those mistake were?
I look forward to hearing from you.