Just before writing this month’s issue, I happened to tune into a BBC Radio 4 programme about a man who trains birds to act. He subverts the imprinting instinct that makes them treat the first moving object seen after hatching as a parent, and even gets them to fly in formation beind an ultralight aircraft. I missed the name, but a Web search found quite a few candidates, such as the Patuxent Wildlife Center Crane Migrators. At about the same time, I was reading a new book entitled The Gecko’s Foot by Peter Forbes. It’s all about understanding and imitating biological structures for use in engineering. Structures like the exquisitely multi-levelled compliance of the gecko’s foot, which is inspiring a sticky tape - not Spiderman but Geckoman - and synthetic gecko hairs for ultra-mobile robots.
Engineers show as much ingenuity in exploiting evolution’s products as evolution did in inventing them. That, on a microscopic scale, is the subject of this month’s issue: bacterial and genetic computation, and the systematic engineering of microbiological systems. I look at some of the projects built for the intercollegiate Genetically Engineered Machine competition or iGEM, including such explicitly computational systems as a binary counter developed by the Eidgenössische Technische Hochschule Zürich which runs on gene expression regulated by zinc finger proteins.
If this seems a long way from AI, think how far we’ve come since the first transistor amplifier or J-K flip-flop. In fact, the Zürich team’s counter uses a genetic J-K flip-flop. Other teams are reverse-engineering E. Coli bacteria, and some are connecting E. Coli control systems to bits taken from other organisms. There’s research on programming noisy and unreliable biological systems, and on new concepts of biological modularity. It’s probably not long before roboticists test-run their favourite control paradigms in an E. Coli chassis. I shan’t explain how to build a bacterial Prolog interpreter - come back in ten years' time for that. But I shall exhibit some of iGEM’s computational achievements, explain a bit of the background, and show how biology is becoming computing. Or, as two of the subject’s researchers write: we don’t fix radios by shooting randomly at their components; and we don’t build microprocessors from chunks of metal and silicon found lying about the countryside.
Biological Computing and the Systematic Engineering of Biological Systems
From Etch a Sketch to Bact a Sketch at iGEM
Etch a Sketch® is a toy many will recall from childhood Christmases. It’s a small red plastic case with a grey screen. Two knobs on the case drag a stylus vertically or horizontally along the screen’s backside, scraping off the fine aluminium powder with which it’s coated. Shaking the case recoats the screen and erases your picture; and when you’re 5 and find one in your stocking, it’s magic. Bact a Sketch is also an erasable drawing screen; but not one to give the average child. You draw using a UV pen, erase with heat, and the drawing surface is bacteria.
Bact a Sketch - also known as BioSketch - was Harvard’s entry for iGEM, the intercollegiate Genetically Engineered Machine competition. iGEM is an annual event: it began in 2003 with an MIT course in synthetic genetic systems, and preparations for 2006 are already underway. The idea behind it is to take advantage of the increasing capacity for bulk DNA synthesis, which is doubling every 18 months in a biochemical version of Moore’s Law. Competitors have to design, build, and characterise genetically encoded machines, going from idea to design to DNA to cell in 3 months. Its organisers hope that we’ll also learn a lot about how to engineer biological systems. As the University of Texas at Austin team say in their iGEM wiki page:
This year we have several project plans which, if they don’t self-assemble into green goo and eat the molecular biology building, should be fairly amusing. Being at heart a bunch of hackers, we believe that the greatest contribution to the field will come from actual experiments and thus we are plowing ahead with our experiments while modeling our system.
E. Coli edge detectors
For this feature, I’ve done a bit of reading on iGEM and the biological background of its and other projects in computational biology, in the hope that I can explain something of their workings to non-biologists. I’m not a biologist myself, so can’t claim too much professional knowledge, but I have tried to check the material carefully. Links are ones that seemed useful, interesting and authoritative: I won’t guarantee that they are always the best for their topic.
With that caveat, let’s continue with another iGEM entry: a bacterial edge detector from the Universities of Texas at Austin and California in San Francisco. In image processing, edge detection locates boundaries between areas of different light intensity in an image, possibly indicating the edges of objects. It’s a well-known task; but most edge detection algorithms do not run on massively parallel E. Coli.
On the way to their edge-detector, Austin/UCSF built the world’s first bacterial camera, using a genetically engineered light-to-pigment-converter which lead one expert to say “Die Amerikaner haben einen wunderschönen, lebenden Lichtsensor entwickelt” - “the Americans have developed a wonderful living light sensor”. Pictures taken with the camera, including a portrait of co-adviser Dr. Andy “Escherichia” Ellington, can be seen on an Austin photos page.
The bacterial camera is way too slow to replace conventional film, but that’s not its purpose. If we can program bacteria to lay down opaque pigment when stimulated by light, we should also be able to make them deposit polymers or metals: bacterial microlithography, and note that the bacterial camera has a resolution of about 100 megapixels per square inch. Also, the stimulus need not be light: we might program bacteria to detect TNT from buried landmines, indicating its presence by synthesising a fluorescent protein. Or, following the seven-segment display designed by Davidson College Synth-Aces, which takes one of its inputs from a caffeine detector, how about a bacterial probe you could stick into your morning coffee to see whether the cappuccino machine has dispensed a sufficient dose to wake you up for the day - a caffeine thermometer?
“They don’t really conjugate one at a time. They can go, but they can’t stop.”
Surveying the Web, I found a wealth of material about iGEM. I’ve already mentioned the iGEM wiki. Its contents include a list of awards given at the iGEM Jamboree. We can all have nightmares about what the Best Show Must Go Moment might have been like, and sympathise with whoever had to deliver the Best Honest Answer. The Best Analogy might well have helped me write this feature; and I would like to know how ETH - the Zürich team who I mentioned in the Introduction - won the George W. Bush Geography award, MIT the Least Transportable Visual Aid, and Oklahoma the Best “Hail Mary” Cloning. Pennsylvania State University won Best New Sport. Their wiki page says that this arose from the idea of a “bacterial maze,” in which bacteria would use logic to find their way through a microfabricated labyrinth. That seemed too difficult, so Penn State linearized the the concept and added transfer of a signal, naming the result a bacterial relay race.
Some iGEM projects are written up in Nature’s 24th November feature Synthetic biology: Designs on life. The feature describes Berkeley’s entry: a new way for cells to communicate, important because it would enable us to build complex control circuits that span groups of cells, instead of relying on the simple intracellular gene circuits I explain later on. Bacteria exchange genetic information by temporarily linking cells and transferring small rings of DNA called plasmids, a process known as conjugation. To quote Nature:
The group managed to trigger the conjugation response with synthetic circuits. But the bacteria turned out to be so eager to join up that they did so in huge bunches - and once they did, it was hard to separate them. “They don’t really conjugate one at a time,” said team spokeswoman Melissa Li. “They can go, but they can’t stop.”
Silicomimetic Noughts and Crosses in DNA
Where will synthetic biology lead? Milan Stojanovic, one of the researchers who devised MAYA, a Noughts-and-Crosses-playing automaton based on DNA logic gates, is quoted as saying that we might add logic to engineered cells, programming them with rules such as
Release a tumor-toxic molecule IF you detect tumor indicator molecule A OR tumor indicator molecule B.
Even if we discover that biology is too noisy and unreliable to carry out long chains of logical calculations, these simple rules could be used in biological devices which seek and destroy medical disorders such as tumours. Incidentally, MAYA’s logic gates work differently from the genetic control used by most of the iGEM-ers. MAYA’s gates are single molecules which take short strands of DNA as input. These strands bind to matching regions on the gates, causing part of the gate molecule known as a “stem-loop structure” to change its conformation from a hairpin shape to something more open. The gates also contain a catalytic region. In YES gates, for example, this is blocked by its neighbouring closed stem-loop. When a DNA strand binds to the loop and opens it, the loop moves away from the catalytic region, which can than perform the function for which evolution designed it. It cuts strands of RNA to which are attached both a fluorescent region and a quencher. These separate, causing the fluorescent region to emit light, hence to be a visible signal of the gate’s output. The entire system is explained in the Nature papers cited in my links, which also tell how the Noughts and Crosses strategy was encoded in logic and then implemented in DNA. As Nature’s Playing to win at DNA computation says:
The effort required to assemble such a complex, functional group of molecular catalysts was extraordinary. Each enzyme had to be designed to interpret the same set of effectors differently. Effectors that might form stable secondary structures were excluded using computational methods, and the multiple deoxyribozymes were all engineered to preclude misfolding. Effector and enzyme concentrations were then empirically tweaked to differentiate signal from noise, and any designs that displayed nondigital behavior or cross-reactivity were further modified or replaced. Ultimately, these arduous efforts culminated in a tour-de-force implementation that included 23 different deoxyribozymes operating simultaneously in 9 different wells with 8 different possible oligonucleotide effectors.
E. Coli edge detectors (2)
What I now want to do is to describe the bacterial edge detector in some detail, in order to display the variety of abstraction levels with which biological engineers have to cope. This will lead on to an account of Biobricks, an important system for standardised modular assembly of biological systems (including computational systems).
Let’s start at the top level with the edge detection logic. As I mentioned earlier, it was the Universities of Texas at Austin and California in San Francisco who entered the edge-detector to iGEM. However, I’ve not found an explanation of the logic on their pages, while ETH do explain in a Wiki page how it might work.
Imagine a “lawn” of light-sensitive bacteria. If illuminated, a bacterium synthesises protein L; if in the dark, it synthesises D. The bacteria are in a moist culture medium, through which L and D will diffuse. However, we’ll assume they diffuse slowly (or break down rapidly), so don’t travel very far. This means the concentration of L and D will vary across the image, both being present only near a light-dark boundary. If we have then engineered the bacteria to synthesis an opaque pigment when L and D are both present, the bacteria will “paint” a picture of the boundaries.
Self-defence in the gut: a bacterial control system
Next, I need go down some levels and reverse-engineer an E. Coli. Living in the gut, this bacterium has no need to sense light. To make it do so, Austin used a light sensor engineered by Anselm Levskaya at UCSF’s Voigt Lab and hooked into the back end of a sensing mechanism that E. Coli does have and that I’ll now explain.
The E. Coli cell is double-walled. The cellular contents or cytoplasm is enclosed by an inner membrane which is surrounded by an outer membrane. The space in between the membranes is called the periplasm (“peri-” is Greek for “around”, as in perimeter and peripatetic). Of the many proteins in E. Coli’s inner membrane are two conventionally written as OmpF and OmpC. These are “porins” or pore proteins: they have open spaces through which small molecules such as waste products and nutrients can diffuse. OmpC has small pores; OmpF has big pores.
Threaded through the inner membrane are molecules of a “transmembrane” protein named EnvZ. When the end on the outer or periplasmic side finds itself in a high concentration of dissolved substances, the protein changes its electronic structure. As a result, the end on the inner or cytoplasmic side of the inner membrane releases a phosphate group, transferring it to another protein known as OmpR. This is one example of a classic type of bacterial control system, in which stimulation of one protein adds a phosphate to a second protein, which may in turn propagate the signal further. We may, if we want, think of the first protein as a sensor, and say that it detects a signal.
In the case of our E. Coli EnvZ-OmpR system, OmpR that has gained a phosphate can then react with a control area on the gene for our small-pore protein, OmpC. This switches on, or “expresses”, the gene, causing synthesis of OmpC to begin. And via another switching mechanism, the phosphorylated OmpR turns off the gene for OmpF. The effect is that E. Coli responds to a high solute concentration by rebuilding its cell membrane to have smaller pores.
In general, despite the handicap of not having a nervous system, bacteria have evolved sophisticated control systems for regulating their genes and hence synthesis of proteins, thus adapting to changes in dissolved chemicals, heat, and all kinds of other conditions. Other organisms have such systems too: cells whose DNA is neatly packaged into a cell nucleus (eukaryotes) have more, and more complex, systems than cells that, like bacteria have no nucleus (prokaryotes). However, bacteria are quite versatile enough, and provide a rich library of parts and systems for us to hack.
“Die Amerikaner haben einen wunderschönen, lebenden Lichtsensor entwickelt”
It’s the E. Coli control system described above that Austin and UCSF used as an intermediary between light and pigment production. Their light sensor was constructed by using part of a “phytochrome” taken from the photosynthetic blue-green algae Synechocystis.
The phytochrome is a protein. Functionally speaking, it resembles the EnvZ protein described above, in that the front end is a sensor, and the back end passes on a signal to other control pathways. But unlike EnvZ, the front end senses light, and the back end signal doesn’t affect gene activity as EnvZ does. Solution: design a new, chimaeric, protein whose front end is the algae’s light sensor, and whose back end is that of EnvZ. Then code this up as a gene, and then insert that into the E. Coli.
In such engineering, there is always a risk that the final system won’t work. For example, suppose the light-sensitive protein had ended up in the middle of the cell rather than in the cell membrane, it wouldnn’t have been so effective a light detector. Or suppose something in the engineered cell had interfered with an existing control system. You wouldn’t want light to shut down sugar metabolism, for example.
Note also that we’re dealing with two entities here: genes and the proteins they code for. This, plus the complexities just mentioned, creates a number of difficulties: difficulties which synthetic biologists are trying to reduce with design tools, simulations, and standardised assembly. We have to know which proteins we want, and if designing new proteins, we’ll probably want to synthesise them and check they work in our intended environment. Here, for example, the researchers tested various possible chimaeras for their sensitivity to light. These were made by varying the amount of the algal front-end protein included, because chain length is known to affect the signalling mechanisms.
Then we have to code the appropriate genes (including their control regions), and work out a method of inserting them into the target cell and its genome. Then, because such methods are never 100% certain, we must verify experimentally that we now have organisms containing the new genes; and that they behave as we want. It’s more work than hacking Perl.
It remains to explain how this wonderful chimaeric light sensor controlled pigment production. The back end of the sensor, as I’ve said, is EnvZ. As I’ve also said, EnvZ phosphorylates the OmpR protein, which affects a control region on the gene for OmpC. Now, researchers have already designed E. Coli in which that OmpC control region is fused to a different gene, one that controls production of the protein β-galactosidase. And β-galactosidase reacts with a certain colourless chemical one can put in the culture medium, producing an opaque product. Putting all this together, and skipping lightly over man-months of lab work, we have our light-to-pigment converter. The logic is simple; the engineering less so.
Genetic inverters, NOR gates, and binary counters
I can now exhibit a genetic inverter. That’s what E. Coli’s OmpR-OmpF/OmpC system is. Its input is the protein OmpR and its output is the protein OmpF. If we could find a gene which produces a third protein, and that OmpF switches off, we would have another inverter.
I don’t know whether there are in fact genes repressed by OmpF - evolution may have used it as a structural unit only. However, that doesn’t matter, because plenty of genetic inverters do exist. They work in the same general way, and can be cascaded together.
One serious problem with the above is that the inverter output is a different kind of signal from the input, making system design more complicated. This is something that Biobricks tries to solve.
If we can make an inverter, we should be able to make a NOR gate, just by giving our genes more than one control region, so that they can be switched off by any one of several inputs. This is what ETH did for their binary counter. They used a family of “zinc finger” proteins as signals, and designed genes which had several control regions, each sensitive to a different protein. Then they built a J-K flip-flop by combining NOR gates.
The Operon Model
It’s worth saying that these control regions are an extremely important concept in biology - so important that Francis Jacob and Jacques Monod won a Nobel prize for uncovering their function. There’s a nice explanation in MIT’s Biology Hypertextbook. Essentially, it’s a roadblock system. Molecules of RNA polymerase land on a “promoter region” (labelled P in the hypertextbook diagrams), latch on to the DNA, and run down the following gene initiating the synthesis of protein until they hit a terminator. However, there may be a “repressor binding site” (labelled O in the hypertextbook) between the promoter and the gene to be transcribed. This site can be empty, or it can have a repressor protein bound to it. In the latter case, the repressor blocks the RNA polymerase from getting past, hence also blocking the synthesis of whatever protein the gene encodes. I like to think of the repressor as a wrestler or security guard, and the RNA polymerase as a little man who tries to get past and copy the information on the DNA. He can do so if there’s no guard; but also - as the hypertextbook page goes on to explain - if the guard’s girlfriend has turned up and become so intimate that guard and girl have rolled off the DNA, leaving the repressor seat empty.
Biobricks: standardised modular engineering of biological systems
On the page Notes on Tom Knight’s Talk of the iGEM wiki, there’s the following quote:
initial tremendous frustation in biology. every experiment turned into two experiments. first, there was the experiment that you wanted to do. second, there was the experiment that you had to do in order to do the experiment that you wanted to do? e.g., will this restricition enzyme work with this DNA, or, is it not methylated, or something else, or something else. Basically, there are too many things to worry about. think about LEGOS. everything is designed to go together. even the flowers snap together.
“Even the flowers snap together”. Replace “experiment” by “function”, and the experience becomes one familiar to every programmer. We’ve developed tools to solve the problem: standard libraries; functions whose internal workings are hidden so that the caller need know only the inputs and outputs; mutually replaceable modules that you can swap without damaging the code using them; abstraction hierarchies. It’s such intellectual tools that Knight is applying to biology.
He describes how in his paper Idempotent Vector Design for Standard Assembly of Biobricks. “Vector” has its customary meaning, a method of carrying DNA into a cell and incorporating it into the cell’s genome. “Idempotent” is a mathematical term for an operator whose effect is the same no matter how many times it’s applied. In Knight’s paper, Biobricks’s standardised assembly technique is idempotent because if we use it to join biological parts A and B into compound part AB, then we can use the same technique to join AB with a third part C; and so on, as many times as needed.
Plasmids and restriction enzymes
This works as follows. As well as the chromosome containing their main genetic material, bacteria have evolved a mechanism for transferring supplementary genetic information between cells. This consists plasmids - the small rings of DNA I mentioned earlier. Genes for antibiotic resistance are transferred via plasmids; and plasmids were the information carriers in Berkeley’s project on Addressable Bacterial Communication.
There are specific patterns of DNA that, if occurring in a plasmid, the bacterial cell will recognise and insert into its genome. We can subvert this mechanism to insert engineered DNA into the bacterium, inducing it to synthesise insulin, say, or spider silk. It’s a standard technique in genetic engineering.
Suppose now that we have one plasmid containing DNA sequence A and another containing B, and we want to make a component plasmid that will insert AB. Because plasmids are rings, if we want to combine their contents, we need to cut the relevant DNA sequences out of each, splice them together, and then reclose the result into a ring, making sure it still contains the “natural” sequences by which the bacterium recognises it. The cutting is done with restriction enzymes: naturally occuring proteins that cut DNA at specific patterns or “restriction sites”. Biobricks components are plasmids equipped with standard “connectors”: standard restriction sites that can be relied on regardless of the component. As long as you cut and join as just described, you’ll have another plasmid with the same “connectors”, which can therefore be used as a component in its own right.
Abstraction hierarchies and modularity
Even with a standard means of joining components, the components' functions can vary widely. An op-amp may have the same pin layout as a set of logic gates, but that doesn’t mean one chip can be substituted for the other. As well as standardised assembly, synthetic biologists need to develop abstraction hierarchies, so that we only think about components in as much detail as we need; and modular implementations of components, so that the physical embodiment of an abstraction doesn’t “leak”.
I’ve already explained how we can build inverters and more complicated components; this gives us a high level of abstraction at which to think while avoiding details of implementation. Biobricks has its own specific systems-devices-parts-DNA hierarchy; my links point at others, amongst them a Biological Network Layer Model based on the Open Systems Interconnection Reference Model.
How do we design genetic components so that implementation details are reliably hidden? In the section on inverters, NOR gates and binary counters, I mentioned the problems associated with using different proteins as signals in different parts of a circuit. Biobricks’s answer to this is a shift of viewpoint - think not of proteins as the signal carriers, but of RNA polymerase. This entails rearranging the inverter design so that RNA polymerase coming in expresses a gene which codes for a repressor protein which acts on a repressor site just “downstream” of the gene, at the inverter’s output. The notion is explained at the end of the Biobricks abstraction hierarchy page, where we learn that PoPS - Polymerase Per Second - may one day become as important as FLoating point Operations Per Second and Logical Inferences Per Second.
Comics fans may prefer the depiction in Drew Endy and Chuck Wadey’s ADVENTURES IN SYNTHETIC BIOLOGY comic strip. Follow the adventures of Bacteria Buddy, Device Dude, and System Sally as they solve the problem of protein signal proliferation and make Bacteria Buddy smile.
Wet, wet, wet
When someone talks about “a system built from a large number of unreliable components of limited life”, I think of my plumbing. Or British Rail. Biological computing systems would qualify too, and researchers are working on programming and understanding collective behaviour in single cells and in systems of communicating cells. I’ve linked to a few relevant pages. It’s worth remembering the patterns that Nature manages to program: zebra stripes, compound eyes, brains, gecko foot hairs.
There is a huge difference between biological and silicon computing. As Jonathan Goler says in a paper on the BioJADE design tool, electronic signals are localised to wires: biological ones are not. They exist in solution, they diffuse, and they will turn up far from where you want them. They are noisy, may be slow, and (at least in bulk) are not at all digital and sharp. ETH modelled the dynamics of their gene circuits using ordinary differential equations.
A problem of a completely different kind, discussed at the end of Goler’s paper, is security. I don’t want to give anyone ideas, so will just note that there are many things to which we’re not immune. Some people believe complete openness about synthetic biology will ensure maximum knowledge and experience is available to reputable researchers who must counter threats. Others disagree.
How the biologist should comprehend the radio
Let me finish. According to Yuri Lazebnik as explained by Sharon Begley in the Wall Street Journal, this is how a team of biologists would fix a radio:
First, they’d secure a large grant to purchase hundreds of identical working radios. After describing and classifying scores of components (metal squares, shiny circles with three legs, etc.), they’d shoot the radios with .22s.
Examining the corpses, the biologists would pick out those that no longer work. They’d find one radio in which a .22 knocked out a wire and triumphantly declare they had discovered the Key Component (KC) whose presence is required for normal operation.
But a rival lab would discover a radio in which the .22 left the Key Component intact but demolished a completely different Crucial Part (CP), silencing the radio. Moreover, the rivals would demonstrate that the KC isn’t so “key” after all; radios can work fine without it.
A clever post-doc then goes on to find a switch which determines whether it’s KC or CP that the radio currently requires. But the biologists still can’t fix the radio, and they haven’t really understood it. They lack the intellectual tools - the equivalent of circuit diagrams and formal languages. Computing and AI can help provide these. Then, as Drew Endy writes in Foundations for engineering biology:
The refining of natural parts to produce engineered biological parts may be similar to nature’s use of negative selection against promiscuous, deleterious molecular interactions within specific cell types, and is analogous to the processing of physical materials in other engineering disciplines. For comparison, microprocessors and other electronic systems are not built directly from chunks of metal and silicon found lying about the countryside.
We shall one day be able to engineer biological computing systems from scratch, without having to rummage around inside intestinal bacteria for our components.
From Etch a Sketch to Bact a Sketch at iGEM
www.eecs.harvard.edu/~rad/igem05/ - The Harvard team’s page for iGEM 2005. Describes their Biowire and Bact a sketch projects, and links to more info, including their wiki and a Harvard Gazette feature.
karma.med.harvard.edu/w/images/7/7c/FinalFinalPresentation_2005-11-05.ppt - Harvard’s Power Point presentation, diagramming the biological implementatation of BioWire and Bact a sketch (a.k.a BioSketch). Includes some info on experimental design, and a picture of Bact a sketch in a nifty blue case.
www.etch-a- sketch.com/ - Etch A Sketch®.
www .howstuffworks.com/question317.htm - How does an Etch-a-Sketch work?, at HowStuffWorks.
parts2.mit.edu/wiki/index.php/Main_Page - Main page for the iGEM wiki.
web.mit.edu/endy/www/igem/iGEM.supplement.pdf - Backgound to iGEM 2003 and 2004 and their research context, including Amorphous Computing, the increasing improvements in DNA synthesis, biological modularisation, and how the Biobricks abstraction hierarchy is implemented. By Drew Endy, MIT.
parts2.mit.edu/wiki/index.php/UT_Austin - University of Texas at Austin wiki page for iGEM, from which I took the “at heart a bunch of hackers” quote.
www.wired.com/wired/archive/13.01/mit.html - Life, Reinvented, by Oliver MortonPage, Wired, January 2005. “Proper engineering, by contrast [with DNA bashing], means designing what you want to make, analyzing the design to be sure it will work, and then building it from the ground up. And that’s what synthetic biology is about: specifying every bit of DNA that goes into an organism to determine its form and function in a controlled, predictable way, like etching a microprocessor or building a bridge. The goal, as Endy puts it, is nothing less than to ‘reimplement life in a manner of our choosing.'”.
E. Coli edge detectors
www.ftd.de/rd/31875.html - Produktion im Dunkeln, by Constanze Böttcher, Financial Times Deutschland, from where I took the “wonderful living light-sensor” quote.
www.utexas.edu/opa/media/photos.php - Austin download page, with some of the bacterial photos.
www.nature.com/nature/journal/v438/n7067/full/nature04405.html - Synthetic biology: Engineering Escherichia coli to see light, by Anselm Levskaya, Aaron Chevalier, Jeffrey Tabor, Zachary Simpson, Laura Lavery, Matthew Levy, Eric Davidson, Alexander Scouras, Andrew Ellington, Edward Marcotte and Christopher Voigt. Nature438, 441-442 (24 November 2005). Explains how the engineered light sensor works. N.B. I’ve linked to several Nature pages, but they may not be universally accessible. Testing them during my final check, I wasn’t able to access them from home - it seems Nature charges for them.
www.rand.org/publications/MR/MR1608/MR1608.appr.pdf - BIOLOGICAL SYSTEMS (PAPER I), by Robert Burlage, University of Wisconsin. RAND report on Microbial Mine Detection.
parts2.mit.edu/wiki/index.php/Davidson - Davidson College Synth-Aces wiki page for iGEM, with the 7-segment chemical display.
parts2.mit.edu/wiki/index.php/Jamboree and parts2.mit.edu/wiki/index.php/IGEM_2005_Awards - iGEM Jamboree and Awards.
parts2.mit.edu/wiki/index.php?title=Penn_StateProjectDes - Penn State wiki page for iGEM, with the bacterial relay race.
parts.mit.edu/wiki/index.php/Berkeley - Berkeley wiki page, with Addressable Bacterial Communication.
www.nature.com/nature/journal/v438/n7067/full/438417a.html - Synthetic biology: Designs on life, by Erika Check, Nature438, 417-418 (24 November 2005). Shortish feature on iGEM.
Silicomimetic Noughts and Crosses with DNA
www.nature.com/nbt/journal/v21/n9/full/nbt862.html - A deoxyribozyme-based molecular automaton by Milan Stojanovic and Darko Stefanovic. Nature Biotechnology 21, 1069 - 1074 (2003).
www.trnmag.com/Stories/2003/082703/DNA_plays_tic-tac-toe_082703.html - DNA plays tic-tac-toe, by Kimberly Patch. Technology Research News, August 27/September 3, 2003.
www.nature.com/nbt/journal/v21/n9/full/nbt0903-1013.html - Playing to win at DNA computation by Jeffrey Tabor and Andrew Ellington. Nature Biotechnology 21, 1013 - 1015 (2003). A short account of the work, with a diagram showing moves on the game board and corresponding states of the logic gates.
www.nature.com/nbt/journal/v23/n11/full/nbt1105-1374.html - Boolean calculations made easy (for ribozymes), by Adam A Margolin & Milan N Stojanovic. Online copy of Nature Biotechnology 23, 1374 - 1376 (2005).
www.ra.informatik.uni-stuttgart.de/~ghermanv/Lehre/Seminar/material/Presentation10/report.pdf - Logic Gates made with DNA, by Maria Belen Canadas Ruiz-Perez, University of Stuttgart. A paper for an Innovative Computer Architectures and Concepts Seminar, 2002. A detailed account of the physics and chemistry behind MAYA-style DNA logic gates, with diagrams of the molecules and processes involved.
E. Coli edge detectors (2)
parts2.mit.edu/wiki/index.php/Edge_Detection - ETH Edge Detection wiki page, with their ideas on its implementation.
Self-defence in the gut: a bacterial control system
www.bio.davidson.edu/Courses/Molbio/MolStudents/spring2005/Champaloux/first.html - Function and Structure of OmpF Porin, by Paul Champaloux, Davidson College. Includes nice diagrams of E. Coli’s cell membranes and of the pore protein OmpF.
nmr.uhnres.utoronto.ca/ikura/1008/calcium/EnvZ/envz.html - E. coli Histidine Kinase EnvZ, by Mitsu Ikura, Department of Medical Biophysics, University of Toronto. How EnvZ regulates pore size.
pub.ucsf.edu/magazine/200305/gross.html - Carol Gross: Feeling the Heat, by Mike Mason, in UCSF Magazine, 2003. The complicated genetic control systems that E. Coli uses to defend itself against heat and stress, related to what I’ve talked about here.
web.mit.edu/esgbio/www/cb/prok_euk.html - Characteristics of Prokaryotes and Eukaryotes. Similarities and differences between these two fundamental types of cell, from the MIT Biology Hypertextbook.
Genetic inverters, NOR gates and binary counters
parts2.mit.edu/wiki/index.php/ETH_Zurich - ETH’s main wiki page, with details of their genetic counter.
The Operon Model
web.mit.edu/esgbio/www/pge/lac.html - The Lac Operon. MIT Biology Hypertextbook page for the Operon Model.
Biobricks: standardised modular engineering of biological systems
parts2.mit.edu/wiki/index.php/Notes_on_Tom_Knight’s_talk - Notes on Tom Knight’s talk wiki page, including the “Every experiment turned into two experiments” quote.
dspace.mit.edu/bitstream/1721.1/21168/1/biobricks.pdf - Idempotent Vector Design for Standard Assembly of Biobricks, by Tom Knight, MIT Artifcial Intelligence Laboratory. Link appears to have died.
parts.mit.edu/ - MIT Registry of Standard Biological Parts. There’s a nice user interface on the parts pages, which enables you to display them in a number of different ways.
austin.che.name/docs/bbpp.pdf - BioBricks++: Simplifying Assembly of Standard DNA Components, by Austin Che. There’s also Austin’s poster BioBricks++: Simplifying Assembly of Standard DNA Components Mindless Module Manipulation by Monkeys, austin.che.name/docs/bbpp_poster.pdf.
Abstraction hierarchies and modularity
parts2.mit.edu/r/parts/htdocs/Abstr actionHierarchy/index.cgi -Abstraction Hierarchy. Describes the design and implementation of the BioBricks abstraction hierarchy, using an inverter as example.
openwetware.org/wiki/BioBricks_abstraction_hierarchy - OpenWetWare discussion on the Biobricks abstraction hierarchy - the proper distinction between part, device, and system.
openwetware.org/wiki/Network_Layer_Model - OpenWetWare page on a biological network layer model based on OSI, the Open Systems Interconnection Reference Model.
openwetware.org/wiki/Dedicated_systems - OpenWetWare on biological virtual machines and dedicated systems: decoupling the function of an engineered biological system from the function of its chassis.
Wet, wet, wet
www.livescience.com/technology/050428_bacteria_computer.html - Scientists Make Bacteria Behave Like Computers, by Robert Roy Britt, LiveScience, April 2005. Short popular feature, with pictures of programmed bacterial patterns.
web.mit.edu/jakebeal/www/Talks/AC-language-overview.pdf - Amorphous Computing’s Programming Languages, by Jacob Beal, 2005. Slideshow presentation on methods and notations for amorphous computing.
www.eecs.harvard.edu/~rad/ - Radhika Nagpal, Harvard. Page on Radhika’s interests: programming and understanding robust collective behavior in biological systems.
dspace-demo.mit.edu/bitstream/1721.2/3328/2/AITR-2004-003.pdf - BioJADE: A Design and Simulation Tool for Synthetic Biological Systems, by Jonathan Goler, 2004. Link appears to have died.
ra.csail.mit.edu/cjt/ProcIEEE-Jan-00.pdf - An Interactive Learning Environment for VLSI Design, by Jonathan Allen and Christopher Terman. The original Java design toolkit JADE.
How the biologist should comprehend the radio
www.cipic.ucdavis.edu/~dmrocke/papers/Can%20a%20Biologist%20Fix%20a%20Radio.pdf - Can a biologist fix a radio? Or, what I learned while studying apoptosis, Yuri Lazebnik. Compares how a biologist would, and should, think about systems.
www.mindfully.org/GE/2003/Systems-Biology21feb03.htm - Biologists’ New Approach: Do Not Shoot the Radio, by Sharon Begley, Wall Street Journal, 21 February 2003.
www.bio.davidson.edu/courses/synthetic/papers/Synthetic_Foundations.pdf - Foundations for engineering biology, by Drew Endy, 2005.
ADVENTURES IN SYNTHETIC BIOLOGY
openwetware.org/wiki/Adventures - ADVENTURES IN SYNTHETIC BIOLOGY, STARRING: Bacteria Buddy, Device Dude, and System Sally. OpenWetWare wiki page linking to various implementations of the comic.
openwetware.org/wiki/Endy:Writing:CCfGDScript - Drew Endy’s script and references for the above.
openwetware.org/wiki/Endy:Writing:CCfGDThoughts - How the comic was designed: finding good visual analogies. See also openwetware.org/wiki/Adventure_Background and openwetware.org/wiki/Endy:Writing:CCfGDAdvice. Comics fans will appreciate the references to Scott McCloud, www.scottmccloud.com/.
www.princeton.edu/~rweiss/papers/weiss-bridge-2004.pdf - Challenges and opportunities in Programming Living Cells, by Ron Weiss. Online copy of The Bridge, Winter 2003.
www.ee.princeton.edu/people/Weiss.php - Ron Weiss’s page. “In my research group, we are exploring mechanisms for harnessing various organisms as computational substrates and micron-scale robots, and extending their behavior by embedding biochemical logic circuitry that precisely controls intra- and inter-cellular processes. This engineering effort of constructing reliable in-vivo logic circuitry with predictable behavior enables a wide range of programmed applications. The application areas include drug and biomaterial manufacturing, programmed therapeutics, embedded intelligence in materials, environmental sensing and effecting, and nanoscale fabrication.
www.dnahack.com/ - DNA Hack. “The website for Amateur Genetic Engineering”.
syntheticbiology.org/ - The Synthetic Biology site. “Synthetic Biology refers to A) the design and construction of new biological parts, devices, and systems. B) the re-design of existing, natural biological systems for useful purposes.”
openwetware.org/wiki/Main_Page - OpenWetWare. “OpenWetWare is an effort to promote the sharing of information, know-how, and wisdom among researchers and groups who are working in biology & biological engineering. OWW provides a place for labs, individuals, and groups to organize their own information and collaborate with others easily and efficiently.”
www.blueheronbio.com/index.html - Blue Heron Bio, the company who made many of the iGEM components. You can paste your DNA spec into a form and order it, say the interesting notes at interconnected.org/notes/2005/03/etcon/tue_dna.txt.
openwetware.org/wiki/Flourless_chocolate_cake - Synthesis anyone can do: Drew Endy’s recipe for flourless chocolate cake.
Links used in this month’s introduction
The Patuxent Whooping Crane Migrators, a feature by Mark Daly, KARE 11 News.
books.guardian.co.uk/reviews/scienceandnature/0,6121,1640650,00.html - Guardian review of The Gecko’s Foot, by Georgina Ferry.
www.me.cmu.edu/faculty1/sitti/nano/projects/geckohair/ - Gecko Hair Manufacture, CMU NanoRobotics Lab.
www.pbs.org/transistor/album1/addlbios/aylesworth.html - Recreating the First Transistor, at PBS.org.
The spookiest moment came when he realised he was doing more than creating little computers. Once he started the process and switched on the genetic sequences which could compound and duplicate the biologic DNA segments, the cells began to function as autonomous units. They began to “think” for themselves and develop more complex “brains”.
His first E. Coli mutations had had the learning capacity of planarian worms; he had run them through simple T-mazes, giving sugar rewards. They had soon outperformed planaria. The bacteria - lowly prokaryotes - were doing better than multicellular eukaryotes! And within months, he had them running more complex mazes - allowing for scale adjustments - comparable to those of mice.
There, very clearly, were the roughly circular lymphocytes in which he had invested two years of his life. They were busy transferring genetic material to each other through long, straw-shaped tubes rather like bacterial pili. Some of the characteristics picked up during the E. Coli experiments had stayed with the lymphocytes, just how he wasn’t sure. The mature lymphocytes were not reproducing by themselves, but they were busily engaged in an orgy of genetic exchange. Every lymphocyte in the sample he was watching had the potential intellectual capacity of a rhesus monkey.
From Blood Music, by Greg Bear. Published as a short story in 1983; expanded to a novel in 1985.
Evolving Computation in Bacterial Collectives
Coincidentally, given my main feature, reader Dr. Erach A. Irani sent me a proposal, developed in collaboration with Dr. Surendra. B. Khadkikar, for creating computational bacterial collectives. The idea is to drive them up the evolutionary curve by equipping them with carbon nanotubes or other nanoparticles.
Many will know that buckyballs, or buckminsterfullerenes, are hollow football-shaped balls composed of 60 carbon atoms linked to form an icosahedron truncated at the vertices, with chemical bonding similar to that of graphite. Carbon atoms can bond into nanotubes as well as nanoballs, a fact discovered in 1991 by Sumio Iijima of NEC Labs when investigating buckyball synthesis.
Nanotubes can be very small in diameter, only a few nanometers long, yet up to a millimeter long. They are also extremely strong. Erach and Surendra propose to use either these, or the gold nanoparticles developed by Dr. Murali Sastry - Google “gold nanoparticles Murali Sastry” for copious references on the latter.
The first stage is to breed bacteria in culture media containing high concentrations of nanotubes (or other nanoparticles). Many bacteria will doubtless die, cell membranes pierced Sebastian-like with a forest of nanotube arrows. Hopefully though, a few will survive long enough to undergo fission, or at least to contribute plasmids to neighbouring bacteria; evolution will then amplify whatever mechanisms were responsible for their tolerance. It is possible that some might even come to depend on the nanotubes.
In the next stage, we induce the bacteria to form colonies, self-organising collectives containing bacterial “specialists”. Each collective is a blob having specialist “defensive” bacteria on the outside. We do this by adding buckyballs to the culture medium and mechanically agitating it, in the hope that the nanotube-rich bacteria will first evolve to defend themselves against the buckyballs, deflecting these with the nanotubes; and then to aggregate and act as symbiotes organised into colonies. The nanotube-equipped bacteria will hopefully be on the outside. As What being in a biofilm means to bacteria explains, many bacteria do form colonies when under environmental stress, and colonies can also form between bacteria of different species. The plaque on our teeth is such a colony.
Although not essential to the project, Erach and Surendra speculate that bacteria in a collective might evolve to signal one another via the nanotubes. If a nanotube inside one bacterium is near a protein which (for example) gains or loses a phosphate group, the protein may transfer charge to or from the nanotube. Should that nanotube be touching one in another bacterium, the latter will also be affected, and could in turn affect biomolecules in its “owner”. Such interactions do exist, as Charge Transfer from Adsorbed Proteins describes, and they’re already being investigated for biosensing: see, for example, Carbon Nanotube Transistors for Biosensing Applications. If they turn out to benefit the collectives in any way, evolution will probably exploit them. It’s an extreme example of what Yoram Gerchman and Ron Weiss talk about in Teaching bacteria a new language, namely the development of novel cell-to-cell signalling methods that we can then use in biological programming.
Finally, we subject the collectives to increasing amounts of stress from chemicals and other insults. Erach and Surendra hope that this will evolve increasingly complex signalling behaviour, both within collectives and between them. If a sufficiently versatile repertoire of signalling methods evolves, we can then “program” the colonies by sending the right signals. This would require the kind of amorphous computing techniques that I linked to in my main feature.
It may also be possible, regardless of the signalling method, to apply regular signal pulses to the collectives and evolve them to exploit the benefits of synchronous, rather than asynchronous, signalling.
One possible application is programming the collectives to seek and destroy cancerous tumours, possibly by attacking them with the nanotubes.
www.labs.nec.co.jp/Eng/innovative/E1/01.html - The discovery of carbon nanotubes - Guided by serendipity, NEC Laboratories. The discoverer of nanotubes, Sumio Iijima, on their history and future.
www.pa.msu.edu/cmp/csc/nanotube.html - The Nanotube Site at Michigan State University Department of Condensed Matter Physics.
www.erc.montana.edu/CBEssentials-SW/bf-basics-99/bbasics-bfcharact.htm - What being in a biofilm means to bacteria, introductory page from the Center for Biofilm Engineering. The page shows a cartoon: one bacterium to another, I just can’t go with the flow anymore. I’ve been thinking about joining a biofilm.
www.thejcdp.com/issue003/overman/08over.htm - Biofilm: A New View of Plaque. Short page with diagrams of dental plaque, including bacterium-bacterium signalling.
star.tau.ac.il/~inon/baccyber0.html - On The Origin of Collectives - Bacterial Evolution, Bacterial Cybernetics Group, Tel Aviv University. Growth and pattern formation in bacterial colonies. The page has some stunning photos, for which see also the Gallery at star.tau.ac.il/~inon/pictures/pictures.html. Last updated 2000. These links and the one below may be dead - I couldn’t access them during my final check, although fragments survive in Google’s cache.
star.tau.ac.il/~inon/wisdom1/preprint.html - Bacterial Wisdom, Gödel’s Theorem and Creative Genomic Webs by Eshel Ben-Jacob, School of Physics and Astronomy, Tel-Aviv University. Online copy of a paper published in Physica A, 248:57-76, 1998.
www.physics.ucla.edu/research/biophysics/pubs/pdf/pub_03.pdf - Charge Transfer from Adsorbed Proteins, by K. Bradley, M. Briman, A. Star and G. Gruner; Department of Physics at UCLA, and Nanomix Inc., 2003.
www.physics.ucla.edu/research/biophysics/pubs/pdf/conf_paper_01.pdf - Carbon Nanotube Transistors for Biosensing Applications, by G. Gruner, Department of Physics at UCLA, and Nanomix Inc.
www.princeton.edu/~rweiss/papers/weiss-pnas-2004.pdf - Teaching bacteria a new language, by Yoram Gerchman and Ron Weiss, Departments of Electrical Engineering and Molecular Biology, Princeton. PNAS, February 24 2004. Vol. 101, No. 8. The engineering of novel cell-to-cell signalling mechanisms.
www.geocities.com/erach27/ConsciousMachinePage.html - Conciousness, conscious bacteria, Gödel’s theorem, and the Turing Machine, by Erach Irani.
“Genetic Programming. Don’t worry yet. This still usually just means programming using the genetic algorithm.” [Not Any More - Ed.] Glossary entry for “GP”, Stammtisch Beau Fleuve!, www.plexoft.com/SBF/index.html.
“Evolution is so notorious for producing quirks that the existence of quirks is a good test for separating evolved systems from rationally designed systems.” Mark Turner, The Literary Mind, 1996.
“It is possible that the designs of natural biological systems are not optimized by evolution for the purposes of human understanding and engineering.” Foundations for engineering biology, Drew Endy, www.bio.davidson.edu/courses/synthetic/papers/Synthetic_Foundations.pdf.
“The goal of the GA-IDS project is to determine whether it is possible to evolve visualizations of computer network and computer systems data that make intrusions or anomalies easier for network or system administrators to detect than existing visualization schemes. Instead of starting with a preconceived visualization model, we start with a language for expressing visualizations and then use genetic programming to produce increasingly refined visualizations based on user feedback.” GA-IDS: Genetic Art For Intrusion Detection, ga-ids.cs.northwestern.edu/.
“The adaptation described below is a classic example of intricate design in evolution. One wonders how it could have arisen through random bit flips, as every component of the code must be in place in order for the algorithm to function. Yet the code includes a classic mix of apparent intelligent design, and the chaotic hand of evolution. The optimization technique is a very clever one invented by humans, yet it is implemented in a mixed up but functional style that no human would use (unless perhaps very intoxicated).” Thomas Ray writing about his Tierra artificial-life system. Quote taken from an abstract at www.talkorigins.org/faqs/tierra.html; Tierra and related software at www.his.atr.jp/~ray/tierra/.
“I shall edit this email slowly, so that you can read it when drunk.”
King Arthur: [about the inscription on the rock] What does it say, Brother Maynard? Brother Maynard: It reads, “Here may be found the last words of Joseph of Aramathia. He who is valiant and pure of spirit may find the holy grail in the Castle of Aaauuuggghhh…” King Arthur: What? Brother Maynard: “The Castle of Aaaauuuggghhhh” Sir Bedevere: What is that? Brother Maynard: He must have died while carving it. Memorable Quotes from Monty Python and the Holy Grail, www.imdb.com/title/tt0071853/quotes.
“The idea of incongruity-resolution has frequently been suggested as an account of many types of joke. However, there is no precise statement either of this ‘theory’ nor of its main concepts (incongruity and resolution), and different authors may disagree on details. We concentrate on two particular variants and attempt to clarify what would be needed to make these into computational models.” Developing the Incongruity-Resolution Theory, Graeme Ritchie, www.csd.abdn.ac.uk/~gritchie/papers/aisb99.pdf.
“More recently the concept of competition among evolving analogies has been introduced by the exploitation of concurrency of processes, that is, each tentative analogy is incrementally built by a separate process which has to compete for resources with other processes which are attempting to construct alternative analogies.” Analogical Reasoning section, Machine Learning & Knowledge Discovery Research page, University of Aberdeen, www.csd.abdn.ac.uk/~pedwards/research/ml_kd.html.
“A PROPOSAL TO CREATE TWO BIODIVERSITY RESERVES: ONE DIGITAL AND ONE ORGANIC.” Thomas Ray, http://www.his.atr.jp/~ray/pubs/reserves/index.html.
“A processor so small even the bugs are hunchbacked.” Stephen Figgins on TuxBot Programming with Python, www.onlamp.com/pub/a/python/2001/03/21/pythonnews.html.
“Unix is like the maritime transit system in an impoverished country. The ferryboats are dangerous as hell, offer no protection from the weather and leak like sieves. Every monsoon season a couple of them capsize and drown all the passengers, but people still line up for them and crowd aboard.” An analogy of Operating Systems, from Hunzeker, Paul Vixie, www.netfunny.com/rhf/jokes/90q3/unixboat.html.
“SegMan is a perceptual substrate that uses computational vision to ‘see’ the Microsoft Windows graphical direct-manipulation interface. SegMan enables other programs to be able to see the graphical interface screen as a human would see it. This enables programs to interact with Microsoft Windows as if it were a user sitting at the console instead of relying on low-level APIs. With SegMan we can create and test more realistic cognitive models of direct-manipulation interface usage, build AI agents that can reason about and use the graphical interface, and write scripts and programs that learn and perform routine tasks in the graphical interface.” Mark O. Riedl and Rob St. Amant, www.csc.ncsu.edu/faculty/stamant/segman-introduction.html.
“With the built-in Prolog interpreter in Windows NT (no kidding! it was used for configuration. I don’t know if it’s still there) there may be more Prolog systems in use than ever before.” Richard O’Keefe, SWI-Prolog maillist, gollem.science.uva.nl/SWI-Prolog/mailinglist/archive/old/0501.html.
“Artificial intelligence is not a term generally used at IBM.” Kathleen Keeshen, IBM spokesperson, 1982. Stottler Henke’s Artificial Intelligence Quotations, www.stottlerhenke.com/ai_general/quotations.htm. Originally from The Tumultuous History of the Search for Artificial Intelligence, Daniel Crevier, 1993.
“Prolog doesn’t have assignment statements. This is deeply upsetting to most programmers. No declarative programming language has achieved popularity.” Richard O’Keefe, ibid.
To iterate is human; To recurse, divine. Programmers' saying. Pojem rekurzie, neuron-ai.tuke.sk/~krankill/ui/rekurzie.html.
Where recursion begins, Sense ends. Student saying. ibid.
Until next month.