User:Reshma P. Shetty
(→What's the difference between parts, devices and systems?)
|Line 51:||Line 51:|
===What's the difference between parts, devices and systems?===
===What's the difference between parts, devices and systems?===
See [[hierarchy | ]] from a discussion with [[Jason Kelly]] and [[Ilya Sytchev]].
Revision as of 15:44, 3 January 2006
32 Vassar Street, Room 32-311
Cambridge, MA 02139
rshetty AT mit DOT edu
My goal is to engineer transcription-based combinational digital logic in Escherichia coli cells.
Synthetic Biology seeks to intentionally design, fabricate and operate biological systems. There are three primary areas in which synthetic biological systems are of immediate utility: chemical energy, materials and information. To harness these systems to either generate new energy sources or synthesize new materials, it is necessary to develop the necessary infrastructure such that cells can be engineered to sense information, process that information using some form of logic and effect a response. Ideally, the parts and devices used to carry out information processing in vivo would have the following characteristics:
- Well-characterized: Device behavior should be quantitatively measured under standard operating conditions.
- Composable: Devices should be designed such that the output of one device can drive the input of another device. In other words, devices should be well-matched.
- Engineerable: It is difficult to imagine every context in which a device might be used. Therefore it is helpful if devices can be tuned such that they work well in larger systems.
- Numerous: Currently the size of the systems we can construct is severely limited by the lack of well-characterized devices. Therefore, it will be important to develop libraries of devices such that more complicated systems can be assembled.
My thesis work seeks to address these goals by developing a new type of transcription-based logic that uses modular, synthetic transcription factors. I derive the DNA binding domains of these transcription factors from zinc finger domains so that arbitrary DNA recognition sites may be used. I use leucine zippers as the dimerization domain so that these repressors are also capable of heterodimerizing increasing both the number and functionality of available dimerization domains. This implementation change adds modularity to the repressors so that domains are interchangeable and may be fine-tuned independently. Also, since there are large sets of both of these domains kinds available, this design enhances the scalability of transcription-based logic. Another key benefit of my proposed transcription-based logic is that by changing the fundamental event that occurs in the device from a single protein dimerizing on the DNA and repressing transcription to two proteins heterodimerizing on the DNA and repressing transcription, faster and more compact logic may be developed.
Some of the questions that I am interested in addressing as a part of my thesis work include the following.
- How do you evaluate device performance? When is the performance of a device good enough? What does good enough mean?
- How do you engineer a device to deliver good performance?
- How should device behavior be characterized and quantified? How little work can we get away with and call a device characterized?
- How do you insulate devices from one another? How do you design them to be orthogonal? How many devices can you put inside a cell?
Frequently asked questions/objections
Disclaimer: These are personal opinions developed as a result of my own thinking on this subject and interactions with others (which I have tried to note as applicable). They are highly likely to evolve over time. If you have a comment/question on something here, send me an email.
But life isn't digital!
The most common question I receive when talking about my work with others is "Biology isn't digital, so why concentrate on digital logic"? There's an answer on the Synthetic Biology FAQ but I'll give my own two cents here. In my mind, there are two valid responses to this question.
Maybe not, but we are better at thinking digitally.
This is the answer I usually give. After thinking about this question a fair amount and talking with others in the MIT Synthetic Biology Working Group (especially Tom Knight), I came to the conclusion that the first "digital" devices in electrical engineering probably weren't very digital either. In fact, even today's devices aren't 100% digital. Calling a device "digital" is really an abstraction we place upon a physical object that behaves according to certain specifications. By carefully determining those behavior requirements and carefully engineering the device, the digital abstraction holds up sufficiently well for the device to work as desired. A lot of work has already been done in electrical engineering in terms of engineering digital circuits from analog electrical components. We should be able to leverage this expertise to design digital devices from analog, biological components.
Sure it is.
Information in biology is encoded by DNA which consists of strings of 4 kinds of nucleotides. Therefore life is fundamentally digital. I first heard this argument in a talk given by Leroy Hood at MIT in 2003 but I am sure that others use it as well. If biology at its core is digital, then it is no longer so unreasonable to design digital devices from biological parts.
BioBricks assembly is too cumbersome.
Agreed. I would really like to be able to email an arbitrarily long DNA sequence to a synthesizer sitting in my lab and have it given me a tube with that DNA in it for very little cost. It would make my Ph.D. go much faster. Unfortunately, we aren't quite there yet (though some might justifiably disagree). So as a stopgap measure, BioBricks assembly allows us to construct systems from parts in a *relatively* cheap, efficient and reproducible manner.
However, in my mind there is a clear difference between the concept behind BioBricks itself (and by BioBricks, I am referring to the parts in the MIT Registry of Standard Biological Parts) and the method used to assemble BioBricks together. There are various ways that one could imagine to improve assembly or even eliminate it all together by using DNA synthesis instead. Regardless of the technique used to fabricate a system, it is still useful to maintain a library of reusable, well-characterized biological parts from which systems can be made. This is how engineers avoid "reinventing the wheel" so to speak. In my mind, it is this concept (not the particular assembly technique) that is the key idea behind BioBricks. Thus, I think BioBricks and the Registry of Standard Biological Parts will still be useful even if/when long DNA synthesis is easily available.
What's the difference between parts, devices and systems?
Here's some thoughts that I decided to post. Feedback is welcome.
Diatribe on Escherichia coli strains
One thing that surprises me is how difficult it is to find an Escherichia coli strain that meets a certain set of specifications. For instance, I want a strain with the lactose permease knocked out and the arabinose permease under the control of a constitutive promoter so that I can get linear induction with both lactose and arabinose. I don't think one exists (if it does please email me!). I also can't find a strain that is lacIq and has the lactose permease deleted (again email me if you have this strain!). I find this situation mildly frustrating.
I also find the nomenclature of Escherichia coli genotype information to be unnecessarily confusing but I am willing to let it slide as a historical artifact. (See the attempt to decipher the code.)
However, in my mind what is truly astonishing is the dearth of information available on existing strains and the fact that some of this information is wrong! Case in point: I was interested in using a strain of Escherichia coli with the lacIq mutation. I found various strains that are supposed to have this mutation: D1210, JM109, BW26434. Then, since I was getting some anomalous experimental results, Tom suggested that I sequence verify the fact that my strains were lacIq. So I did and lo and behold, none of my sequences had the lacIq mutation on the genome. Now based on my anomalous experimental results (which are no longer so anomalous) and reading of some papers, I think that D1210 really is lacIq but that it just has lacIq on the F plasmid rather than on the genome. But JM109 and BW26434 ... or at least the versions that I sequenced ... are not lacIq as documented. I don't understand how people use these strains without having correct genotype information. I also don't understand that with all the sequencing centers there are and how many people work on or with Escherichia coli, why all the common lab strains at least don't get sequenced. Some claim it is a combination of the lack of resources and the fact that this isn't an interesting thing to do. Quite possibly this is true, but nevertheless, I find this situation unbelievable. Anyway, it was these experiences that led me to populate the standard strain page.