Re: Data and meaning (3)

At 08:26 AM 09/10/2002, Pedro wrote:

>As for Jerry points, I find rather vague his approach to biological
>optimality. Nowadays there are pretty clear biological treatments of the
>theme. Stuart Kauffman 'rugged landscapes' (see At home in the universe,
>1995; Investigations, 2000) are perhaps one of the most interesting
>EVOLUTIONARY views about how genomes have managed to explore the
>multidimensional optimization problems confronted by their phenotypes.
>There is not only one maximum or a minimum, but scores of relative max-min
>'peaks and valleys' that have to be charted by micro-and macro-movements
>of the genetic parameters. Depending on the ruggedness of the landscape,
>and the 'movements' of the environment, the ways to explore these peaks
>and valleys have to change (eg, prokaryots vs. eukaryots; plants vs.
>vertebrates). It is the modern version of van Valen 'Red Queen'
>hypothesis. Although I disagree with Kauffman in other extremes, more or
>less follow him in this overall approach to biological (and
>technological!) exploration of optimality when the environment presents
>conflicting demands... Besides, evidences of optimality in biological
>STRUCTURES, very well studied ones, pile up in the classical literature
>(eg, D'Arcy Thompson): skeletons, bones, shells, internal organs, brains,
>neuronal connections, plants, leaves, flowers, widespread Fibonacci series
>and golden mean ratios in morphologies, etc. In the late 80's, quite many
>studies of MOLECULAR optimality were done (eg, glycogen molecular design,
>pentose cycle, Kelvin cycle, metabolic control, etc.), and quite many
>mutational analysis were performed --with the result that most natural
>enzymes are just 'optimal' to their molecular function... FUNCTIONALLY or
>physiologically (pretty more difficult terrain) there also lots of studies
>and controversies on allometric relationships, grounded on optimal
>efficiency when metabolism, transportation, maintenance, degradation, etc.
>are considered together. The prevalence of the 3/4 power law relations in
>most animal species is understood as a hallmark of the systemic quest for
>optimal efficiency.

Although I would agree that there are cases in which optimality IS
relatively well-defined, I doubt that fitness landscapes can give us
optimality criteria in general. The main problem is that fitness is a
relation between an organism or population of organisms and environment. If
the parameters of the two are independent (as required by the autopoiesis
account of autonomy or the closure view of efficient causation inn
organisms), then there is no problem, but if the two are dynamically linked
such that function depends on both (as I suggested), then changes in the
organism can change the fitness landscape (and vice versa, of course) so
that non-linear relations between the two can affect fitness measures.
Effectively, this means that moves towards optimality can change the
landscape, which becomes a moving target. This is just the interaction case
that Jerry invokes, quite correctly and insightfully, in my opinion. This
objection would also apply to optimality explanations of enzymes -- the
question still arises, why this molecule rather than that one -- the
optimality space is not obviously well defined for sets of alternative
molecules, even if the molecular shape space is dependent only on the
molecule. If their is some global criterion (global extremal principle),
optimization within this constraint is possible. But, as I said, there are
reasons to think that at least two proposed extremal principles related to
ordering and organization are not generally true. This isn't to say that
there are not others. 3/4 power law relations might be an alternative. I
have looked at these only superficially. I doubt that they are general
enough, from what I have seen so far, and there has been some criticism of
the data fit to the power law. The explanation in terms of branching
distribution and collection patterns is intriguing (the 4 space network
filling the 3 space volume), but it seems to me that we really should have
1 space transport systems filling a three space volume, and we should get a
3-x/3 law. Some critics have said that laws of this form are consistent
with the evidence. Anyway, I remain sceptical.

>As for John, I appreciate the very elegant synthesis presented about the
>different approaches (mostly non-biological ones) to extremal principles.
>I am particularly intrigued by the MaxEntPro. The only additional
>suggestion I would make about that principle is, What would happen if
>enzymes had efficiently decoupled their 'informational' work from the
>thermodynamic entropy (& free energy) requirements?

I haven't thought about this much for enzymes, about which my knowledge is
embarrassingly small. Some people (e.g., Stan Salthe) have tried to apply
MaxEntPro to informational space when information and energy are largely
decoupled. I would suppose that the same considerations abut extremes would
apply to the informational space (we should expect neither maximal
information preservation nor maximal information dissipation). I have
argued various cases with Stan for some time, but though our positions have
hardened, I don't think either of us has a knockdown argument. In any case,
for enzymes, assuming decoupling, MaxEntPro would imply maximal elimination
of alternatives. Stan, in recent discussion with me, justifies this with
selection from competition for energy gradients, but it could also arise
from competition for information (I take it that this was his earlier view,
which he has not rejected, but he tends now, as he has at times in the
past, to assimilate information and energy considerations). For my part, I
doubt that competition for resources is decisive, mostly because I think
that the notion of resource is a moving target. I am not clear how to
specify this for the case of enzymes. I just don't know enough. If it
turned out that enzyme competition is strong compared to rates of change in
the values (resources) for which there is competition, Stan might be right
for the enzyme case. There may be a range of cases, or typical cases may be
irreducibly complex, and non-optimizable.

I think it is an interesting question, which I intend to keep in mind.

Cheers,
John
Received on Thu Oct 10 05:08:51 2002