ORIGINAL TEXT FOR BIOTHEORY BROCHURE:
Judith
Rosen’s comments:
The following material constitutes
my father (Robert Rosen)’s rough version of how a brochure might be written
about him by a third party but is, of
course, written by him. He found the exercise extremely awkward. He told
me about it the day he wrote it, saying that he considered it both a conceit
and a form of prostitution or debasement to have to try and “sell” himself in
this way, and yet he recognized the necessity of creating a brochure for
BioTheory (his brainchild, a subscription service) that would illustrate why
his thoughts might be of value to prospective subscribers. I have made it
available to people because there is an interesting perspective being displayed
here, as my father contemplates himself and his career as a stranger might and
tries to give reasons why his expertise was worth buying. He would be somewhat displeased
that I am publishing it at all! But consider this paragraph his disclaimer: He
didn’t inflict this on you. I did. (Sorry, Dad! But it
might help somebody.) Perhaps a new kind of scientific avenue can be created
using my father’s BioTheory idea as a model. If so, then he would at least have
given rise to a way out of the frustrating treadmill that academia has become
and the market driven needs of commercial venues, or the inbred and incestuous
situation that has developed among the “gate-keepers” at many of the mainstream
and/ or commercial scientific journals
today.
A BRIEF SURVEY OF ROSEN'S APPROACH
Robert
Rosen has been for many years one of the world's foremost theoretical
biologists. He has authored some 250
research papers, and a dozen books, concerned with both the development and the
applications of the theory underlying biological processes.
He
very early began to develop the concept that biology should be based on notions
of function rather than structure, and it was function that was of primary
concern in understanding the basis of life and of organism. He has since then explored the possibilities
of building function-based models of biological processes. These have turned out to be very different
from, and far more general than, reductionistic treatments based on structural
ideas.
After
early years, starting from conventional approaches based exclusively on
empirics and experiments, he concluded that these were too abstract to deal
with the basic questions of organisms. It seemed quite inappropriate to
collapse an entire organism down to one or another measurement, carried out on
some localized piece, as is typical of an experimental procedure of, say,
biochemistry. The problem was to tie these empirical abstractions together in
some coherent way, and that meant developing theory for this purpose. Theory,
to Rosen, was always a way to develop a less abstract view of organisms than
could be provided by experimental data alone.
These
ideas led naturally to a study of mathematics, viewed as a way of binding
particulars together, and inter-relating them in a coherent larger structure. Rosen has always maintained that it the
empirical particulars, the data arising from individual experiments on
localized parts, that were the true abstractions, and that it was theory,
expressed in the last analysis in mathematical terms, which was needed to
offset the highly abstract nature of individual particulars.
Thus,
from very early in his scientific career, Rosen inverted the conventional view,
that experiment dealt with reality, and that it was mathematics which was
unbearably abstract. Nevertheless, it was this extended study of mathematics,
with the intention of determining how it could express biological realities,
which preoccupied most of Rosen's student career. It also prepared him to better assess the
contrasts between physical sciences, and their claims to universality, against
the requirements of biology. The stark contrasts between these claims and the
realities of biology reinforced a growing belief that a "theory"
based on reductionistic ideas alone was inadequate for biology.
Rosen
obtained a PhD in l959 at the
In
l966 Rosen moved to the State University of New York at
During
this period, Rosen spent a year (1970-1971) as Visiting Fellow at the Center
for the Study of Democratic Institutions, founded by Robert Hutchins, and
located in
Most
recently, Rosen has published a comprehensive treatment of his general approach
to biological investigations, stressing also its relations with, and
distinction from, traditional reductionistic views, under the title Life Itself
(
Rosen's
approach to theory, based on function and organization rather than on
fractionable structure, is inherently comparative; functional ideas are
inherently exportable from one domain to another, both within biology, and
across interfaces between biology and other domains. Let us give a few examples.
A.
PROTEIN FOLDING:
Traditional
views of genetic "coding" suffice only to determine the primary
structure, or the sequence, of amino acids along a polypeptide. To become active, the polypeptide molecule
must fold into a specific, three-dimensional active conformation; acquire a
tertiary structure. For various
important reasons, it was necessary that this folding be a spontaneous process,
occurring without the need for further "information". In other words, it was necessary that primary
structure entail the tertiary, active, functional structure. The protein folding problem is basically the
attempt to carry out this entailment explicitly; to find an algorithm which can
predict the geometry of a folded protein from its primary structure alone.
Reductionistic
physics, and physical chemistry, claim to be able to address this question
directly, in terms of writing down a free-energy function for the molecule,
from primary structure alone, and then minimizing it. Despite a great deal of effort, this kind of
approach has basically gotten nowhere after nearly 35 years of trying; indeed,
such efforts have only served to expose the weaknesses of reductionistic
approaches to biological function in general.
Rosen,
on the other hand, was led by theory to take a quite different approach to this
problem, based on morphogenetic considerations rather than the free energies of
physical chemistry. It rests on results
arising from what initially seems like a quite different problem; what
embryologists call cell sorting. This
has to do with generation and stability of patterns formed in populations of
motile units, which possess differential affinities for each other. Such populations arise in many different
contexts besides embryology; in physics, for example, they are the basis for
"phase separations", and even phase transitions. In the l960's, Rosen developed the general
models which solve such problems, at least phenomenologically.
He later realized they could be applied to protein folding problems by tying
together some of the sorting elements with inelastic string.
Experience
with such an approach has been quite interesting. First, instead of the
hundreds or thousands of independent variables, which the conventional
approaches based on physical chemistry mandates, there are only a few; perhaps
half a dozen. This means tertiary
structures can be generated quickly; even with sub-optimal algorithms, a
polypeptide can be folded in half an hour, with accuracies comparable to any
other approach, according to conventional measures of such things. Moreover, we can begin to approach
"inverse problems", which will involve generating primary structures
which fold to display pre-specified activities using these ideas, which are
simply inaccessible to other approaches.
Aside
from its inherent interest, in biomedical and pharmaceutical applications, the
approach itself is broadly applicable far outside these domains. It basically shows folding to be an example
of what is often called a synergetic process, in which only a few degrees of
freedom control hundreds or thousands of others. Biology is replete with such processes, which
are only obscured, perhaps beyond salvation, by reductionistic approaches to
them. We would desire to (but are
presently incapable of) incorporating such synergetic capabilities into our
technologies (say in the design of robots).
A study of protein folding from the above perspective has already
revealed a few necessary conditions, of general applicability, for such
synergies to be manifested.
Indeed,
this entire story itself exemplifies the benefits of beginning from an adequate
theoretical foundation, and of how that foundation itself provides the vehicle
for transporting ideas far beyond the original context for which they were
developed. In particular, ideas from
morphogenesis (developmental biology) were initially exported to problems of
spatial organization (folding) in large molecules, and then, by virtue of the
synergies manifested in them, to the general control problem of coherently
manipulating many degrees of freedom with a few controls.
B. ANTICIPATION:
Traditional
ideas of biological control and regulation have been entirely based on
cybernetic ideas of feedback. These in turn depend upon error signals, which tell a controller
that system behavior has departed from a nominal value, or set-point, and
generates a corrective action, based on that deviation, to reduce that
error signal to zero. Technologically, such a cybernetic system is fractionable
into two functional parts, which control theorists call a controlled system and
a controller. If you do not make this
separation, the result is a single big dynamical system, with a locally stable
point attractor (steady state). That is
why the theory of cybernetic control is, according to theory, a reformulation
of a part of the stability of dynamical systems. But the converse is not true.
Moreover,
there are other styles of control besides error-based, cybernetic
controls. One of them is based on
anticipation, in which a present action is taken, not to correct an error which
has already occurred, but to pre-empt an error which will occur in the absence of
that action. Control here is based, not
on the past, but on the future; in the last analysis, through the agency of a
predictive model, which converts present information into (predicted) future
consequences; these provide the basis for present actions. There are many deep
reasons for regarding the style of control manifested by organisms, not as
cybernetic, reacting to error signals which have already occurred, but as
anticipatory, a model-based transduction of a present situation into a
predicted future. This change of
viewpoint has many profound consequences; it makes a great difference, for
instance, to view a hormone (say) as embodying an error signal, or as a
predictor.
A
control system working on the basis of predictive models, rather off cybernetic
feedbacks, can itself be viewed as a little theorist. And the anticipatory
style of control has many advantages over a cybernetic style. For one thing, it
is much easier for anticipatory controls to manifest synergistic behaviors
reliably; behaviors of the type we have just considered in the folding of
proteins.
On
the other hand, anticipatory systems can exhibit a down side, which is itself
illuminating. Namely, if one's model is
not quite in correspondence with what it models, the control actions it
generates may become inappropriate; maladaptive. Such a system will appear to be failing, but
in a novel, global way, rather than through a lesion of some localized
part. Such global failures in fact
characteristically occur in organisms; roughly, they constitute senescence. In
fact, one of the profound difficulties in studying senescence empirically is
precisely that, in a senescent organism, each localizable part is apparently
behaving normally; unlike our machines, there is no localized lesion, whose
repair will rejuvenate the entire system.
In this view, then, rejuvenation involves the recalibration of
predictive models, rather than replacement of parts.
It
might be remarked that many of those concerned with social organization (e.g.
Toynbee, Spengler, to name two) view social collapses
as a form of senescence, and blame it precisely on a failure of the predictive
models which underpin the society as a whole.
A Dark Age is thus a global failure of precisely the kind we have
discussed. This exhibits, in itself, how
theory allows biological ideas to move across interfaces traditionally
separating biology from other realms.
There
is no reason to expect that anticipatory systems of this type can be
accommodated within the same dynamical language manifested in conventional
cybernetic control, or in generalized form, in the stability of ordinary
dynamical systems. Roughly, in such
model-based behavior, one must make room for the models which drive it.
Something in the system must be doing "double duty", as an ordinary
material constituent, and as a predictor for something apparently quite
unrelated in material terms. This kind
of "double duty" is very hard to accommodate in conventional dynamical
terms, but it is of the essence in biology.
In fact, it is closely related to the phenotype-genotype dualism, which
is so characteristic of organisms in general, but which is essentially absent
(or at least presumed absent) in the inorganic world.
Thus,
the possibilities of anticipatory control open a wealth of possibilities, and a
corresponding wealth of applications, not only in biology itself, but in its
impacts on both technology and on the way we view social systems.
III. SIMILARITY:
Experimental
data pertains, strictly speaking, only to the system on which it was
collected. Nevertheless, it is essential
to believe that such data actually has a far wider currency; that it can be
exported, far beyond the system on which it was collected. Or, in other words,
that the specific system on which the data was collected is only a surrogate
for a far wider class of systems, and that the data itself survives replacing
such a surrogate by another. This notion of surrogacy is very closely related
to the concept of model. The
significance of data rests precisely on how large is the class of systems of
which the one being observed is such a model i.e. on how far that specific data
can be extrapolated to such other systems, and be regarded also as data about
them. Beliefs about the extrapolability of data must underlie the study of, say, a
laboratory rat, when the actual interest is in humans; the rat must be
considered as a model or surrogate human, and thus rat data becomes human data.
Although people collect such surrogate data with great care, no such care is taken
with the conditions under which an underlying surrogacy actually holds. As a result, there is simply a tacit
presumption that surrogate data may be extrapolated ad lib, subject only to a
few heuristic rules of thumb.
In
fact, such notions of surrogacy are not empirical things at all. They have, in fact, been studied for a long
time, under many separate headings. One
of the more familiar forms of these ideas is found in the concept of scaling,
or scale modelling in engineering. Here, one studies
a system of interest, a prototype, by first building a convenient scale
model. The model is related to the
prototype by specific scale transformations; explicit rules which convert data
from the model into corresponding data about the prototype.
Such
scale transformations can be regarded as follows. Think of the prototype as a deformation of
something about the model. The
corresponding scale transformations serve to undo, or compensate for, the
effect of this deformation on data pertaining to the model. Thus, the transformed data stands in the same
relation to the prototype as the original data did to the scale model.
The
study of the effects of deformations, or perturbations, on data, falls within
the province of a branch of mathematics called stability, or more precisely,
structural stability. It deals with
questions of the form: when is a deformed system's data a transform of the undeformed system's data?
That is: when are the two systems, to that extent, similar?"
This
similarity thus provides the basis for every kind of surrogacy relation. Surrogate systems are deformations of one
another, which allow data pertaining to one of them to be transformed into
corresponding data pertaining to the other. This intertransformability
of data, unfortunately, will not generally hold for arbitrary
deformations. Thus, the question of when
a deformation of a system is also a surrogate of that system is in fact a very
deep and difficult one. Let us give a few examples of these ideas, to show how
widely they manifest themselves:
i. Simple ideas of scale modelling underlie the employment of
"dimensionless" units, like moles in chemistry, which seem
independent of any particular substance.
In fact, this apparent independence of particularity is an illusion; that
particularity is incorporated into the scaling rules themselves.
ii. In
mathematics, the concept of congruence in geometry provided an early
illustration of, and motivation for, the study of similarity in general. Ideas originating in geometry have, over the
years, been generalized in all directions.
As we have already mentioned, it is the very substance of structural
stability. But it also animates things
like the Theory of Categories, which is concerned with modelling
relations in general.
iii. In physics, the Special
Theory of Relativity asserts that all "observers" are similar; that
their data is inter- transformable (the similarity transformations here are the
famous Lorenz Transformations). Elsewhere
in physics, we find assertions that all ideal gases are similar, or more
generally, that all gases obeying a given Equation of State are similar, etc.
iv. In biology, a very early attempt to come to
grips with these notions of similarity was embodied in D'Arcy Thompson's
"Theory of Transformations" in l9l7.
Roughly, Thompson asserted that all closely related species are also
similar. This principle is highly
nontrivial, because "closely related" is a genetic notion;
"similar" pertains to phenotypes.
Here, then, we consider deformations generated by genetic changes (call
them "mutations"); the assertion is that the deformed or
"mutated" phenotype is a transform of the original one.
This
last example is, in the last analysis, why people believe a rat, say, is a
phenotypic surrogate for a human. It provides a basis for unlimited interspecific data extrapolation. But on the other hand,
D'Arcy Thompson's assertion cannot be universally true; if it were, there could
be no such thing as macroevolution. It
can be seen that theoretical aspects of similarity, pervade the sciences in
general, and biology in particular, in many deep ways. They have many profound
implications, not only for biology itself, but for the way empirical biology is
done, and for technologies like genetic engineering, which seeks to extrapolate
biological ideas to a wider class of surrogates.
The
extended examples we have discussed above are themselves samples from a much
longer list, but they are fairly characteristic of the breadth and scope of the
ideas, and the richness of their applications across a
broad spectrum of domains. All this can only be hinted at in this short space.
Of course, these examples are themselves interdependent, and so too is the
scope of their applications. These latter far transcend mere attempts to
generalize from empirics and observation and the seeking of serendipity. Of course theory is never intended to replace
data and experiment, just as the latter can never replace or constitute theory. The two are complementary activities, not
adversarial ones. It is, however, always
the intent of theory to render the events of our world more transparent, more
comprehensible, by making such data less abstract than it would otherwise
necessarily be.
The
above examples, and many others, have been pioneered by Robert Rosen. His approach is uniquely characterized by a
refusal to throw away the biology and keep only the physics and the chemistry,
which is asserted in the reductionistic approach, and by approaching an interface
between biology and another area-- from
the biological side. Thus, he
regards a molecule as a (possibly overlapping) array of sites and epitopes, which play functional roles determined by larger systems with which these arrays
interact; these cannot simply be identified with, or reduced to, the internally
bonded arrays of atoms with which chemistry traditionally deals. One
manifestation of this outlook is found in the approach described above, in
connection with the very concrete problems of protein folding; an approach
based on morphogenetic principles rather than on conventional considerations of
thermodynamic free energies and their minimization.
Rosen
brings to such theoretical studies a number of unique perspectives. One of them is an intimate familiarity with,
and indeed, a deep respect for, the empirical facts about organisms...if not
for their conventional interpretations.
Another is with both the historical development, and current status of,
theory in other scientific realms; what it can do, and what it cannot do. A third is an insight into mathematics,
something that conventional biology regards as entirely irrelevant. But in fact, mathematics deals with systems
of (inferential) entailments no less compelling than the systems of causal
entailments we call the Natural World, and which it is the business of science
to study. Seen in this light,
mathematics is not primarily a body of theorems to be applied or not, as other
considerations allow, but rather in its structures almost as alternate
realities, which provide contexts for the one with which science must
deal. It is precisely this aspect which
accounts for what the physicist Wigner once called
the "unreasonable effectiveness" of mathematics as an instrumentality
in the natural sciences. Rosen has based
his general theory of modelling on the establishment
of congruencies between causal entailments in the material world, and
inferential entailments in appropriate mathematical ones. Perhaps no one else has ever utilized
mathematics in quite this way.
In
our society, a scientist is called on to be many things in addition to his
science. He must do research, he must be
an author, an educator, often an administrator, an editor, a consultant, an
expositor, a lecturer, and perhaps many other things. Rosen has done all these things, and a few
others, over the years. For instance, he
holds an international patent on a novel kind of oral drug delivery system
suitable for delivering materials like insulin, or heparin, or vaccines, which
hitherto could only be delivered by injection.
But it is his research for which he is best known, research which,
despite its unconventional and often controversial character, has nevertheless attracted substantial and constantly
growing international attention.
********************************************************
Copyright,
Judith Rosen, 2004