directed evolution

[William_T_Powers2]

[From Bill Powers (940908.0830 MDT)]

Martin Taylor (940907.1130) -- (et. al.)

RE: What is the problem.

Through this post and several others today there runs a thread that goes
"of course we know that evolution itself is not controlled." I've been
trying for a couple of years to open this assumption to discussion, but
it seems to be a very deeply-seated belief that is not to be casually
dislodged.

My doubts about this assumption arose when I found out about the E. coli
effect, in which a completely random process turned out to be usable by
a control system to achieve a preselected effect with startling
efficiency, in comparison with the same process having effects only
through chance. Suddenly the effects of "blind variation" seemed to fall
into two categories: directed and undirected. Clearly (to me) the
difference between undirected and directed evolution came down to the
selection process: whether there was a process of comparing the results
of variations with a reference level and initiating random changes with
a timing that depended on the acceptability of the results, or whether
the random changes simply occurred at a background rate without any
relationship to the consequences of those changes.

In the latter, the undirected, case, mutations and crossovers occur as
if imposed by some external agency (cosmic rays, statistical
fluctuations). When they occur is independent of their effects on the
organism. Under this concept, the only selection process is whether an
organism dies before reproducing. Organisms so constructed as to resist
deleterious effects from the environment survive and reproduce; those
which are not die. Logically, this will lead to successive generations
in which properties that promote survival will be propagated. I do not
argue with that.

However, the question that arose in my mind after learning of the E.
coli effect was whether the traditional kind of selection process is
sensitive enough to small variations among progeny to account for
evolution as we imagine it to have happened. Such a question would not
arise without an alternative process to compare with natural selection.
It would simply be assumed that conventional natural selection must be
efficient enough to have created all the species and subspecies that
exist; if it were not, they would not exist. That is basically the
reasoning I have seen in writings on evolution. The only alternative
that seems to have been considered is divine intervention, which
scientists rule out as a non-explanation. In fact, it seems that anyone
who suggests that there could be a direction to evolution is
automatically classed as a closet supernaturalist, because it is
inconceivable that any other kind of direction could exist.

But we now have an alternative within the realm of scientific
acceptability, one in which the mutation rate is no longer independent
of the consequences of mutations. This alternative says that mutation is
not something that happens to an organism; it is something that the
organism does. And of late, this principle of mutation has suddenly
appeared in the laboratories of science and is being used to cause
directed evolution on a large scale and with extremely high efficiency.
The demonstration by Cairnes et. al. of spontaneously directed mutation
in E. coli has created instant resistance among biologists, but the mere
demonstration seems to have unplugged a channel of thinking, so that
some biologists and theoreticians have begun to explore methods of
turning random variation into systematic results without explicitly
recognizing what they are doing, or seeing its implications concerning
natural evolution.

Even in the field of simulating evolutionary processes, this directive
selection has appeared without being specifically recognized as showing
us a new principle. In fact, the direction that takes place has not
really been understood as constituting a phenomenon. Rewarding an
artificial creature by allowing it to survive if it makes a move toward
a beneficial change imposes a systematic relationship between selection
and the consequences of random changes, and quite subtly makes the
frequency of random changes depend on detecting a move in the right
direction. A species which is propagating itself while not showing any
improvements is mutated again right away; one which is behaving in a new
way deemed to be in the right direction is allowed to go on propagating
without change. This is how the new laboratory procedures work, and how
many simulations of mutation work.

There are, of course, simulations of evolution that do not contain this
directing feature. They produce some striking effects, but these effects
do not bear on survival in the same way that changes in the organization
of real behavior do. The patterns that result, while interesting, have
no discernible relationship to the welfare of the patterns. They are
simply interesting to look at. They offer no answer to the question,
"What does this have to do with organisms in real environments?" And
most important, they do not answer the question of whether this method
of inducing change is efficient enough to account for the observed rate
of evolution, particularly in complex organisms. One can produce a
million generations of an organism in a few minutes on a desktop
computer if the criteria are simple enough; the question is not whether
some systematic change might occur in a million or a million million
generations, because surely it must. The question is how far evolution
could proceed by that method in only 3.5 billion years.

The E. coli phenomenon reopens all questions having to do with natural
selection. Because we now can see that a far more efficient selection
process can exist, we can raise doubts about whether natural selection,
as traditionally conceived, is actually efficient enough to account for
evolution as we know it. In the past, evolutionary arguments have
avoided this question, because even to raise it was to question
evolution itself; after all, if natural selection is not efficient
enough to produce organisms at the rate we see them appearing, what
explanation for the fact of evolution is left? None at all that a
scientist could accept. But that is no longer true; we do have an
alternative. We can now admit, if it is true, that ordinary natural
selection can't possibly account for the rate at which organisms have
changed from blue-green algae to the species we see today, including
ourselves, in only a few billion years. We can admit this without being
forced to accept divine intervention. And it's really that possibility,
a holdover from centuries of intellectual (and sometimes physical)
battles, that keeps biologists from taking too close a look at natural
selection.

···

---------------------------------------------------------------------
Best,

Bill P.

[_Martin_Taylor]

[Martin Taylor 940908 15:00]

Bill Powers (940908.0830 MDT)

One assumption and one assertion deserving comment.

Assumption: >we can raise doubts about whether natural selection,

as traditionally conceived, is actually efficient enough to account for
evolution as we know it.

We certainly can raise doubts, but the assumption you work on is that the
doubts are justified.

We can now admit, if it is true, that ordinary natural
selection can't possibly account for the rate at which organisms have
changed from blue-green algae to the species we see today, including
ourselves, in only a few billion years.

Granted, you do put in "if it is true," but the tenor of this posting, as
with your other postings and private messages on the topic, is that it
IS true. You have no evidence for that.

Assertion: That variation of the pace of mutation results in evolution
directed to some end other than survival. What I understand Cairns et al
to have been claiming is that when a specific problem arises in the
environment of e-coli (lactose deficiency, I think), then the mutation
rate OF THE RELEVANT GENE is increased. That would be directed evolution.

There is a less strong version, with which I have no problem, and which
corresponds to the mechanism you have been pushing: that when the ORGANISM
is stressed, its mutation rate changes over all its genes (or at least
those that are susceptible to mutation--that's a variable across genes).
Those that result in a greater ability of the organism to survive the
stressful environment tend to increase in the population. Those that
don't, tend to get lost. That is NOT control, since there is no reference
signal of any kind to specify what way the stressed organism should change.
Quite unlike the e-coli demonstrations/experiments, there is no target
for the species to approach.

None of the arguments of the last few days that you criticise in your
opening line:

Through this post and several others today there runs a thread that goes
"of course we know that evolution itself is not controlled."

are affected in the slightest by your following discussion or by whether
evolution uses your proposed mechanism. Whether the rates change as a
consequence of environmental stress or not, the result is "what works
tends to survive better than what doesn't." The result is not "humans
are the target toward which evolution has been controlled for 4 billion
years."

Martin

[William_T_Powers2]

[From Bill Powers (940909.0900 MDT)]

Martin Taylor (940908.1500) --
RE: directed evolution

There is a less strong version, with which I have no problem, and which
corresponds to the mechanism you have been pushing: that when the
ORGANISM is stressed, its mutation rate changes over all its genes (or
at least those that are susceptible to mutation--that's a variable
across genes). Those that result in a greater ability of the organism
to survive the stressful environment tend to increase in the
population. Those that don't, tend to get lost. That is NOT control,
since there is no reference signal of any kind to specify what way the
stressed organism should change. Quite unlike the e-coli
demonstrations/experiments, there is no target for the species to
approach.

Well, I beg to differ with your statement that this is not control. The
reference signal is implied by the fact that there is a zero-stress
state, in which all the variables involved in stress are at particular
values. In this state, mutuation proceeds at some low rate. When there
is error, which is what we mean by stress, the mutation rate increases
(that is, the interval to the next mutation shortens). This limits the
lifetime of changes that increase the error or stress and extends the
lifetime of changes that reduce the error. The reference level does not
specify the direction in which the stressed organism should change. It
specifies a desireable consequence of change; zero stress (or better,
all critical variables in their reference states).

Your version still uses "survival" as the criterion of successful
evolution. The point of the E. coli principle is that the organism does
not have to die in order to improve the direction it is going.

The problem with a concept like "stress on the organism" is that it is a
generalization from something that must always actually be very
specific. "Organisms" are not stressed; some specific aspect of
organismic functioning is disturbed -- that is, specific variables are
caused to change away from their normal states. It is always possible
that these normal states are specified by reference signals (even in
DNA) and that there is a comparator and output function related to each
one (McClintock's transposable elements). Specific random mutations
would then be instituted locally rather than globally, as in one of your
suggestions for local reorganization in the learned behavioral
hierarchy. This may be what is going on in the Cairnes et. al.
experiments.

The same principle applies to "classical" reorganization, except that
the sensed errors are less directly related to the random
reorganizations, and some other solution must be found for making the
reorganizations apply in the right way. In both cases there is a
reference signal involved (many of them), and the resulting changes in
organization are strongly directed to restore certain variables to their
reference conditions -- even though the reorganizing changes are still
random.

What I am proposing is not to do away with ordinary natural selection,
the Darwin's Hammer type of selection. It is to add an (evolved) system
which DOES have reference signals, and acts to control specific
variables carried in the DNA and other structures passed along from
generation to generation. In a stressful environment, these variables
would tend to be forced away from their reference states, and
reorganization would commence. The result of a disavantageous change
would not necessarily be lethal -- ideally it would be nonlethal. It
would simply increase the error, resulting in advancing the time of the
next reorganization and allowing that change to persist (propagate down
the generations) for a shorter time.

Actually, as we have found from our simulation experiments with the E.
coli phenomenon, the best way to set up a system like this is to give it
parameters which always change at a rate proportional to the absolute
critical error, and in which reorganization amounts to random shuffling
of the _directions_ of change. This duplicates the "swimming" of E.
coli, which is continuous, and the "tumbles" which alter the direction
of swimming at random. The spatial coordinates of E. coli are always
changing; it is the direction in which they are changing that gets
reorganized. So I am proposing a system in which the parameters of
control are always changing, but reorganization alters the direction in
which they are changing. Also, as the total error decreases (unlike the
case for E. coli), the "speed of swimming" decreases -- the size of the
increments added to or subtracted from the parameters during each
iteration decreases. This latter is a refinement that is not strictly
necessary; it simply reduces the span of the random movements as zero
error is approached. If the swimming speed of E. coli depended on the
concentration error (not the rate of change), it would become zero or
minimum at the source of the gradient of attractant. There is no
experimental data on this; all the tests were made in gradients far from
the source, or uniform gradients. It is observed that E. coli tends to
aggregate around pipette nozzles from which attractants are coming, but
I have seen no data on whether the swimming speed in such a situation is
less than in a gradient far from the source. Maybe E. coli just swims
toward the source until it bumps into it.

The problem with my proposal is that it ends up predicting random
changes of organization occurring at variable intervals, which is what
we observe anyway in evolution. The only sign that an E. coli type of
control system is working would be that the changes are biased in some
particular direction -- not because the changes are nonrandom, but
because changes in other directions are followed quickly by a new
change, while changes in the "right" direction are followed only after a
longer interval by another change. Naturally, among variants that are
propagating a less advantageous change, the mortality rate will be
higher than in variants propagating a change in a good direction. But
the mortality rate may not be the determining factor; it could be a
side-effect. Which is the main effect and which the side effect, of
course, depends on the relative magnitudes of the effects, which can be
determined only by a quantitative investigation.

If you simply count changes and record their viability, you will miss
the dimension in which control is taking place: you will see only
ordinary natural selection, proceeding perhaps at a surprisingly high
rate. It is possible to detect E. coli type control if it's happening,
but one has to look at the data from the right point of view. The
critical piece of data is the interval between random changes as a
function of the degree of stress. The mortality rate might be an
indicator of degree of stress, but E. coli type evolution can take place
with no change in mortality rate at all.

I should mention that the "direction" of evolution implied by my
proposal is simply _an increase in control over variables that can
affect the state of the organism_. Nothing is implied about producing
certain forms. This direction, translated into specific processes,
amounts to decreasing the error between certain reference states and the
states of critical variables (I'm beginning to like Ashby's term better
than "intrinsic variables").

···

-----------------------------------------------------------------------
Best,

Bill P.

[_Martin_Taylor]

[Martin Taylor 940909 17:30]

Bill Powers (940909.0900 MDT)

I don't know whether I'm following you or not about your views on
evolution, because of one critical paragraph.

Your version still uses "survival" as the criterion of successful
evolution. The point of the E. coli principle is that the organism does
not have to die in order to improve the direction it is going.

This sounds as though you are thinking of an organism improving the
genes that affect it OWN functioning, whereas my understanding is that
the organism's genetic structure is determined at birth. Now it is
true that contemporary medicine sometimes succeeds by introducing cells
with foreign genes, but these normally have to be replenished as they
die out. And any mutations that happen within somatic cells of an organism
would not be passed on to the next generation, anyway. The only mutations
that affect evolution are those that affect the egg and sperm (in a sexual
organism, anyway), and those don't affect the way the parent organism
functions.

If my reading of this paragraph is correct, then I'm afraid that my
interpretation of the rest of your posting will be rather badly coloured.
But if I omit that paragraph, I can accept that what you say is not
implausible. Whether it is right, that's another matter. I must
think awhile as to whether it is likely or workable.

Mind you, I think you are worrying about a non-problem when you say
that simple random variation and selection couldn't work fast enough
to account for the emergence of the complexity you see in only 4 billion
years. That's a loooong time. Did you read the article in Scientific
American (or was it American Scientist) recently, about a real case of
directed evolution by pure selection? The researcher wanted to see if
the evolution of wolf to dog could have happened as some theories said,
by selection of wolves that were a bit less wary of humans than most. So
he started with foxes, selecting for mating foxes that were a bit less
wary than most. In 20 years (what's that--10 generations?) he had a
"race" of foxes with stubby snouts, floppy ears, and that wagged their
tails and sought out human company. That's not training of the individuals,
because the original fox cubs had none of those properties. I could be
wrong in the details--check out the article--but I'm sure about the outline
of the story.

Evolution by selection can work VERY fast when directed, but it works
pretty fast when it isn't. You probably know about the moths (butterflies?)
in England that often rest on tree trunks, and that changed from mostly
white to mostly black after the industrial revolution made the trees
sooty, and have now changed pretty well back again since the anti-smoke
laws of the 1960s were enacted. The selection there is by birds who can
see moths that contrast with the tree trunks better than moths that blend
in with their background. Individual moths are not like chameleons that
can change colour to blend in. But darker ones survive with a bit higher
probability on sooty trees, and lighter ones survive a bit better on
"normal" trees. So, evolution changes the species from light to dark
and back again in decades.

That is NOT control,
since there is no reference signal of any kind to specify what way the
stressed organism should change. Quite unlike the e-coli
demonstrations/experiments, there is no target for the species to
approach.

Well, I beg to differ with your statement that this is not control. The
reference signal is implied by the fact that there is a zero-stress
state, in which all the variables involved in stress are at particular
values. In this state, mutuation proceeds at some low rate.

In your model. But if you like to call this control, I won't argue. It's
the same kind of control as in the reorganization in an individual. But
the same effect occurs without changes in the mutation rate, by selection.
The selected organisms whose genes propagate down the generations are the
ones with lesser stress, both ways. And I wouldn't call that "control."
It just happens. It's like getting 2 and another 2 and finding you have 4,
or performing an arithmetic operation and finding 2+2=4 and then counting.
Same deal, either way. One may be more efficient than the other, but both
function, and the only question at hand is whether one or both mechanisms
are used. Selection is, because there's no way to avoid it (other than
the purest creationism). Rate-selected mutation may also be used, but
whether it is is a matter for research.

···

===================

The problem with my proposal is that it ends up predicting random
changes of organization occurring at variable intervals, which is what
we observe anyway in evolution.

True, but the same effect occurs with constant mutation rates as well.
It even showed up in my trivial little simulation.

======================

I should mention that the "direction" of evolution implied by my
proposal is simply _an increase in control over variables that can
affect the state of the organism_. Nothing is implied about producing
certain forms.

That same "direction" seemed to me to be the essential result of a selection
process, and the reason why I originally argued (and still do) that PCT
is a necessary consequence of evolution, and would be observed in all
living species, no matter where in the universe they might be, or on
what chemistry thewy might be based.

=====================
On a related matter, much of the evolution discussion has taken for granted
that genes survive only through the survival of offspring of a carrier.
This assumption ignores the fact that many of the same genes are in the
siblings, and other relatives.

It is sometimes said, even in quite serious books, that one shares 50%
of one's genes with one's siblings or with one's children. But this is
not true. Almost all of my genes are carried by anyone who can read this
message, and I could help those genes to survive by helping any of my readers.
A gene is a pattern in context, not a specific bunch of atoms linked in
a special way.

If I want my genes to have an enhanced chance of surviving to another
generation, I can make that true for 98.4% of them by saving a chimpanzee
who then goes on to reproduce. Chimps have most of the same genes that any
of us do. I don't know what proportion of my genes (NW European origin)
are the same as those of my wife's (Korean) brothers, but it's bound to be a
lot higher than 98.4%. (I have fewer genes in common with my wife, or with
any other woman, because of the X-Y chromosome difference). So I can help
the survival of almost all my genes by helping ANY human to have offspring.
I would be surprised if I have any genes at all that are not carried by
quite a few other people.

The "selfish gene" notion says that you tend to see at any one time those
genes that build organisms that act so that the gene has a good chance
to propagate. This means that organisms that act to improve the chances
of their own OR related species will be ones that are found. "Altruism"
is almost a given, when one notices that fact. But there is a STRONG
counter-tendency, because it is precisely members of one's own species
and those of related species that will require the same resources as
oneself, and too assiduous altruism will tend to reduce the survival
tendency of genes carried by oneself that are not carried by the one helped.

A bit afield from PCT, but perhaps relevant to the discussion.

Martin