|
Author
|
Topic: The design of robustness
|
Art
Member
Member # 179
|
posted 02. July 2002 23:52
Hi James,
Thanks for the interesting pointer. I'll add a few short comments and sit back and read some more.
First, continuing with my simple model of robustness, an important question (one that James may allude to in his comments about natural selection) is that of the "amenability" of the simple system to meaningful change. IOW, how do we go from a state where none of the steps can be changed so as to yield, say, evolutionary change? Can such a highly robust system ever change so as to become one more responsive to environment or, I daresay, natural selection?
The answer is, not surprisingly, yes. (Well, it is not suprising to me.) And we do not need the engineer's practiced and exacting touch to accomplish this. Consider, for example, the effects of adding a single branch in the linear pathway, so that one intermediate can now be passed on to F, or alternatively be broken down (and discarded). Now, if the first order rate constant that governs the breakdown of the intermediate is much greater than that which controls its further use towards production of F, then we have rather a new state - one in which the production of F is proportionally changed by changes either of the alternative pathways of the intermediate. (I'll let the reader "do the calculation" to affirm my description of things.)
Of course, my model (and the other models that have been pointed to in this thread) is admittedly very simplified. But it suffices to make a few points. One is that moving from one robust state to another is not an exceedingly difficult "act", and it requires no inauguration or modification of massive interlinked components or controls. Indeed, no direct contact between any of the hypothetical enzymes, and certainly no active or elaborate controlling mechanisms other than those inherent in the scheme, are needed. (Of course, one can debate the accesssibility of the change I introduce in "real-life" situations, and I would be glad to argue my case. But perhaps in another thread.)
A second is a point I make in these boards (mainly on ARN) from time to time - that we must not forget breakdown, turnover, the discarding of perfectly good and useable intermediates, when we think of biological systems.
I'll close for now with a question for James (and anone else who can stand the topic) - do you think the HOT model (or any other) can be related to my idealized one? Do any of the lessons we have learned, and can learn, from my model apply to other quite different ones?
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 03. July 2002 10:11
Art:
I am loath to answer your question directly, because I simply do not possess the technical expertise to compare the HOT model to yours effectively. I wonder if, instead, you would mind if I asked you a couple of questions. I feel that I have a lot more questions than answers!
(1) Do you agree that the HOT model seems more applicable to biological cases than the Bak SOC model, Kauffman's "edge of chaos" models, and other similar ideas? I found this part of the Doyle et al. argument particularly convincing---namely, that we need "integral feedback control" to make sense of biological robustness, and that this is lacking in the SOC-style models. But, again, I am more interested in hearing what others think of this, than in pushing my own intuitive impressions.
(2) Even if HOT is a superior model for biosystems, we still need somehow to get to the origin of life. So, does the idea of a possible transition from SOC to HOT make sense? Can the two types of model be fruitfully combined in some way? Or is something else altogether different still required?
(3) I sort of had the impression---as I said before---that one needed to get more complexity, and specifically, nonlinearity into the system in order to get robustness. So, I guess what I would really like is to understand how your system is able to achieve a similar level of robustness (if it does) without the feedback loops. This is beyond my technical reach, and I am puzzled as to how this can be. If you could explain it to me in a simple way that I could understand, I would really appreciate it.
(4) Do you see the robustness in your model as an abstract mathematical property that could in theory be instantiated in any sort of material "substrate," or is it rather essentially tied to the assumptions being made about the underlying physical processes?
This is the absolutely crucial question, in my view, but again, while I have a gut feeling on this issue, I do not have a deep enough understanding of the relevant physics to be able to articulate it persuasively. I assume the latter is the case---that the robustness depends crucially on assumptions about the underlying chemistry and/or physics---but I would be interested to hear your view of the matter.
Thanks for your help.
---James
IP: Logged
|
|
Art
Member
Member # 179
|
posted 11. August 2002 17:10
Hi James,
I’m finally getting back to this subject.
Trying to move away from the trend (started by moi) of answering questions with questions:
You asked:
quote:
(1) Do you agree that the HOT model seems more applicable to biological cases than the Bak SOC model, Kauffman's "edge of chaos" models, and other similar ideas? I found this part of the Doyle et al. argument particularly convincing---namely, that we need "integral feedback control" to make sense of biological robustness, and that this is lacking in the SOC-style models. But, again, I am more interested in hearing what others think of this, than in pushing my own intuitive impressions.
Certainly, Table 1 from the review by Carlson and Doyle (PNAS 99, 2538-2545, 2002) would lead one to conclude this. But I have not read anything of a response to this suggestion from supporters of SOC, so I’ll have to defer for now.
Well, not entirely. My own intuition is that the HOT model is fairly descriptive, as far as such a simple model can be. But it yields an awfully obvious result, one that has been taught in ecology classes for decades. And I am skeptical that the underlying structure of the model that Doyle and his coworkers use to reach their conclusion is capable of lending new insight into biological processes that reach the same conclusion. Rather, my view of the HOT model is that it is an indication (as is my own model in this thread) of how robustness can be attained from very simple underpinnings.
quote:
(2) Even if HOT is a superior model for biosystems, we still need somehow to get to the origin of life. So, does the idea of a possible transition from SOC to HOT make sense? Can the two types of model be fruitfully combined in some way? Or is something else altogether different still required?
I think that some chemical manifestation of SOC is likely to be productive in understanding the origin of life. And, as has been (tentatively ![[Smile]](smile.gif) - there seems to be a silent skepticism as to my claims) shown in this thread, once we begin to understand the properties of metabolic pathways at the steady state, we see that the super-imposing of other ideas is not needed to “attain” the requisite degree of robustness. Rather, a fairly constant metabolic flux is the major requisite.
(I wonder, as an aside, if robustness is even necessary for the origin of life?)
quote:
(3) I sort of had the impression---as I said before---that one needed to get more complexity, and specifically, nonlinearity into the system in order to get robustness. So, I guess what I would really like is to understand how your system is able to achieve a similar level of robustness (if it does) without the feedback loops. This is beyond my technical reach, and I am puzzled as to how this can be. If you could explain it to me in a simple way that I could understand, I would really appreciate it.
I don’t know about non-linearity, but feedback “loops” are inherent properties of any series of first-order chemical reactions (such as I have illustrated). Briefly, the flux through any step of the pathway is a property of the rate “constant” and of the concentration of the substrate for the reaction. Thus, if one increases the rate constant, then at the steady state, the concentration of the substrate will be decreased by a corresponding factor. Which means that the overall flux through the step will be unchanged.
quote:
(4) Do you see the robustness in your model as an abstract mathematical property that could in theory be instantiated in any sort of material "substrate," or is it rather essentially tied to the assumptions being made about the underlying physical processes?
IMO, the robustness that we see in my model is much more than a mathematical abstraction – it is the foundation, for example, of the robustness that is seen in metabolism in cells, and it affords important insight into more general mechanisms by which robustness may be attained, and different robust states might be derived from others. As importantly, it (probably) can be easily demonstrated with materials that can be bought from any scientific supply company.
All of this raises an intriguing question that I will close this post with – in a sort of metaphysical sense, should we make anything of the inherent robustness of a series of first-order chemical reactions?
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 13. August 2002 09:30
Hi Art:
Sorry I answered your last question with a question. I'll do my best to give you an answer this time, but remeber that I am not a scientist, merely a philosopher trying to soak up as much science as I can to use for my own purposes (i.e., understanding what purpose is).
I find everything you say about the robustness and adaptability (two sides of the same coin?) of metabolic networks very encouraging. This is precisely what I would like to hear, given my interest in replacing selection theory with a dynamical theory of the living state as a means of grounding the immanent teleology of life.
Your remarks comparing and contrasting SOC, HOT, and your own model were helpful to me. Again, I cannot comment on the technical details, but I find the generality of your own approach attractive.
Once we can agree that networks are robust, though, I have the additional concern about whether there is any intrinsic difference between the kinds of systems that we can impose a robust network on from the outside (neural nets) and those in which robust network processes seem to arise spontaneously from the inside (cells).
In short, there are two issues for me, not just one. The first is whether a robust dynamics can be shown to exist at all (I take it that this has now been shown to be not a problem). The second is what sorts of material constitutions are required for such dynamics to emerge spontaneously (as in the protein-ordered water gel), as opposed to being imposed on matter from the outside as a set of arbitrary boundary conditions (as in silicon wafers).
[ 13 August 2002, 09:32: Message edited by: James A. Barham ]
IP: Logged
|
|
Art
Member
Member # 179
|
posted 16. August 2002 21:54
Hi James,
Moving on in our dialogue:
You said: quote:
Once we can agree that networks are robust, though, I have the additional concern about whether there is any intrinsic difference between the kinds of systems that we can impose a robust network on from the outside (neural nets) and those in which robust network processes seem to arise spontaneously from the inside (cells).
An interesting question (as opposed to a concern ) that is related to my playful comment regarding the possible metaphysical signifiance of the sort of robustness I have tried to describe. This is a toughie for me, and seems (at first glance - I’ll be pondering the matter for many days to come) to be akin to asking, say, if there is any connection (in a design sense?) between fine polished pottery and stones that have been "polished" in fast-moving streams. They’re both smooth and shiny, but what does this really mean (especially when it comes to the stones)? I’m pretty comfortable in leaving these things disconnected, and I am probably inclined to leave the different routes to robustness in a similar state.
I guess a more interesting question would be to ask if the macroscopic robustness that is seen living things (e.g., in developmental pathways) is more akin to an intrinsic characteristic or an imposed property. I’m not quite ready to extend my model in this direction, but I would appreciate input from others in this regard. (Even if its just to tell me that I have really bitten off more than I can chew.)
IP: Logged
|
|
Frances
Member
Member # 169
|
posted 19. August 2002 23:36
From Tong Zhou*, J. M. Carlson*†, and John Doyle‡
Mutation, specialization, and hypersensitivity in highly optimized tolerance
quote:
Nevertheless, the process of purposeless mutation and selection in our model, like biology, creates the impression of a clear direction in evolution, with results very similar to what would arise from purposeful engineering design for high yield.
Seems that the evolutionary mechanisms can explain apparant design in many aspects. In this case apparant direction. Doyle's work also suggests that complexity as found in biological systems can be quite well understood in the context of mutaiton and selection.
Fascinating work which shows how 'design' can be quite productive in recognizing that we may be using similar techniques and achieve similar results to natural mechanisms.
Doyle seems to have joined forces with the Santa Fe institute
This site may be quite helpful since it contains definitions of robustness.
Robustness seems to be quite an interesting approach and generating quite a few results in enhancing our understanding of complexity.
A fascinating resource indeed. I will have some reading to do.
[ 20 August 2002, 00:57: Message edited by: Frances ]
IP: Logged
|
|
James A. Barham
Member
Member # 50
|
posted 29. August 2002 20:04
Frances:
I've been hors de combat for a while for personal reasons. I just wanted to thank you for the link to the great site on robustness at SFI. I was not previously aware of it. I appreciate the tip very much.
IP: Logged
|
|
L. R. B. Mann
Member
Member # 331
|
posted 22. September 2002 07:32
Art wrote:
>how does one design the system so that the production of end product - F - is
> unaffected by changes in the properties of the enzymes involved?
...
>it turns out that the very simplest model for the pathway is in fact
>highly robust! The steady-state concentration of [F] is completely
>unaffected by changes in any of the intermediate steps in the pathway.
From the viewpoint of an academic physical chemist & biochemist
(rtd), both the question and the answer contain defects.
1 A large uncountable variety of 'changes in the properties of the enzymes
involved' will slow the metabolic pathway A to F. Some comparably huge,
largely unexplored variety of mutations can cripple one of the enzymes so
that it doesn't work at all.
Which changes will fail to do any such harm is only vaguely
predictable. There is a class of mutations whose base-change in the DNA
leaves the changed codon still coding for the same amino-acid, and indeed
there's only one amino-acid (tryptophan) for which no such 'neutral'
mutations cannot occur as there is only one codon for that a-a. But in
general it is not possible to predict accurately what changes in the
enzymes' catalytic action will be caused by changing one a-a to another,
especially if they're similar e.g valine to leucine.
2 The concept of 'rate-limiting step' in any such pathway is very
important in any discussion of the effects of enzyme changes on the
pathway's thruput.
3 The final quoted sentence is extremely incorrect. The proximal reason
for this disastrous outcome is hinted at by the previous sentence - the
very simplest model is too simple.
cheers
R
IP: Logged
|
|
Art
Member
Member # 179
|
posted 24. September 2002 12:17
Regarding my hypothetical and robust pathway, R wrote: quote:
From the viewpoint of an academic physical chemist & biochemist (rtd), both the question and the answer contain defects.
I think one needs to revisit the issue - it's robustness, and how to get there. I don't understand how this question can contain defects. My "answer" is likewise beyond reproach, as it is primarily concerned with robustness and its design. The broader issues - how the simple model relates to real-life situations - are matters of debate, but no more so than, say, the question of how HOT can be used to understand to real situations.
Nonetheless, some commentary is in order:
quote:
1 A large uncountable variety of 'changes in the properties of the enzymes involved' will slow the metabolic pathway A to F. Some comparably huge, largely unexplored variety of mutations can cripple one of the enzymes so that it doesn't work at all.
True enough - if one blocks any hypothetical pathway, then flux would cease. I'm more interested in the fact that the illustrated pathway is quite insensitive to all changes that yield non-zero values for fluxes through individual steps. This better describes the effects of the vast majority of non-neutral point mutations (if we are speaking of a biochemical pathway, and enzymes).
quote:
Which changes will fail to do any such harm is only vaguely predictable. There is a class of mutations whose base-change in the DNA leaves the changed codon still coding for the same amino-acid, and indeed there's only one amino-acid (tryptophan) for which no such 'neutral' mutations cannot occur as there is only one codon for that a-a. But in general it is not possible to predict accurately what changes in the enzymes' catalytic action will be caused by changing one a-a to another, especially if they're similar e.g valine to leucine.
Again, the remarkable feature of the simple series of first order reactions is that none of this matters.
To be sure, this may be somewhat removed from real life. But perhaps not so much as one might imagine (as indicated below).
quote:
2 The concept of 'rate-limiting step' in any such pathway is very important in any discussion of the effects of enzyme changes on the
pathway's thruput.
I couldn't agree more. In my simple model, the rate-determining steps for flux are the entry and exit points of the pathway. (Ever wonder why feedback control is often or usually effected at the first step in a pathway? The answer lies, IMO, in my very simple model.)
quote:
3 The final quoted sentence is extremely incorrect. The proximal reason for this disastrous outcome is hinted at by the previous sentence - the very simplest model is too simple.
What is "too simple"? Certainly, my model helps one understand some pretty amazing experimental results. James gave us one. Another recent case is the study by Koebmann et al. (J. Bacteriol. 184, 3909-3916, 2002). The abstract finishes this post, but the bottom line is that glycolysis in E. coli seems to be rather robust, but can be significantly altered by increasing the rate with which one end-product (ATP) is utilized.
For now, I think the bigger issue remains - is there any relationship between robustness as I have shown can be derived in simple ways and the robustness that is seen in much more complicated systems? Are there basic, simple rules than extend to many different scales? Can we differentiate (or even identify) "intrinsic" and "extrinsic" robustness?
------------------------------------
J Bacteriol 2002 Jul;184(14):3909-16
The glycolytic flux in Escherichia coli is controlled by the demand for ATP.
Koebmann BJ, Westerhoff HV, Snoep JL, Nilsson D, Jensen PR.
Section of Molecular Microbiology, BioCentrum-DTU, Technical University of Denmark, Lyngby, Denmark.
The nature of the control of glycolytic flux is one of the central, as-yet-uncharacterized issues in cellular metabolism. We developed a molecular genetic tool that specifically induces ATP hydrolysis in living cells without interfering with other aspects of metabolism. Genes encoding the F(1) part of the membrane-bound (F(1)F(0)) H(+)-ATP synthase were expressed in steadily growing Escherichia coli cells, which lowered the intracellular [ATP]/[ADP] ratio. This resulted in a strong stimulation of the specific glycolytic flux concomitant with a smaller decrease in the growth rate of the cells. By optimizing additional ATP hydrolysis, we increased the flux through glycolysis to 1.7 times that of the wild-type flux. The results demonstrate why attempts in the past to increase the glycolytic flux through overexpression of glycolytic enzymes have been unsuccessful: the majority of flux control (>75%) resides not inside but outside the pathway, i.e., with the enzymes that hydrolyze ATP. These data further allowed us to answer the question of whether catabolic or anabolic reactions control the growth of E. coli. We show that the majority of the control of growth rate resides in the anabolic reactions, i.e., the cells are mostly "carbon" limited. Ways to increase the efficiency and productivity of industrial fermentation processes are discussed.
-----------------------------------
IP: Logged
|
|
|
Printer-friendly view of this topic