FW: Heritability and generative entrenchment

[Nick Thompson]

Hi everybody,

Forgive me for casting such a wide net, but we seem to be skating very
close to what Carl Tollander calls "artificial epigenesis and I want to
keep the conversation as open as possible until I see who is interested.  

David Wilson (attached and below) has taken the discussion in the direction
I hoped it might turn .... that selection might consist of unstable
relations amongst stable arrays. Everybody is talking as if the elements in
the arrays are genes, but there is no particular reason not to include
epigenetic nodes as well.   The implication for my question on inheritance
is that all the chaos in the genetic-epigenitic system is going on a level
BELOW where selection is going on.  This might seem to beg the question
concerning inheritance ... what "force" holds together the stable arrays?
However, at this early stage of my reading, Wimsatt and Schank seem to be
saying that the stable arrays are high  entropy .. i.e., they hang together
because that's where randomization takes them.  

I am very excited about all of this, as you can see, but as you can also
see, I should shut up and go back to reading before I say more.  Thanks for
your patience.  Be sure to read the message below and the attachment if you
are interested.

thanks, all,

Nick

Nicholas Thompson
nickthompson at earthlink.net
http://home.earthlink.net/~nickthompson

> [Original Message]
> From: David Sloan Wilson <dwilson at binghamton.edu>
> To: <nickthompson at earthlink.net>
> Date: 5/21/2006 8:35:45 AM
> Subject: Heritability and generative entrenchment
>
> Dear Nick,
>
> Thanks for your interesting message. I'm sending this reply to you only
> rather than the whole group--I'll let you be the judge of what the
> whole group sees and in what manner.
>
> Consider a parameter space with many local stable equilibria.  When a
> biological system is in the basin of attraction for a particular local
> equilibrium, it is generatively entrenched and here is a problem for
> heritability. However, there can still be selection among multiple
> basins of attraction, providing a concept of heritability. I discuss
> this in the following paper titled "Natural Selection and Complex
> Systems: a complex interaction."
>
>

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ATT00002.TXT
Url: http://redfish.com/pipermail/friam_redfish.com/attachments/20060521/c604a765/ATT00002.bat

[James Steiner]

Maybe this is a non-sequitor, but here goes: When I look at the
recently published first 1000 lines of chromosome 1, it looks just
like executable code looks in a hex editor: gibberish, but with a
definite sense that it is meaningful gibberish. Lots of patterns,
blocks with definite "character" differeing from neighboring blocks,
long stretches of "empty" space, or buffers, or uninitialized storage,
etc.  I'm sure I'm not the first / only person who looked at the
sequence and thought that DNA code represents biological machine
language. So, there's no reason to exclude from that incredibly vast
and highly developed program anything that we know computer programs
can do or have. For example: data validation routines, code validation
routines, code-repair routines, self-altering code, and code that
restricts (or attempts to restrict) where in the code or data changes
may occur, and the scope of those changes. I expect there are chunks
of DNA that are instructions, data, and meta-versions of those. The
assorted machinery of the cell interprets that code (which may be
"written" in many different "languages").

So, all that being said, I would conjecture that most of the code is
data, and reletively little is instructions. Further, the instructions
may have ways of validating the data. So, the effect of mutations is
limited to changes in non-validated data [like arguments to a
function, there's no knowing what parameters should be--e.g. number of
times to repeat (make arm routine), the field that defines (thickness
of a hair) or (length of left leg)], changes to validation values
[e.g. the field that defines the maximum allowable change in the
(thickness of hair) field)]. changes to validated data that happen to
match the checksum, and changes to both data and checksums that happen
to "add up". So, invalid, life-ending mutations may occur very often
during the gamete dance, but are corrected, and so never percieved by
us.

There being so little in the way of actual instructions the kinds of
mutations that change the instructions happen only very rarely. With
the added possibility that there is code that validates and corrects
the instructions, too, the likelyhood of disasterous mutation occuring
from run-of-the mill mutative effects is very small.

I suppose now that the genome has been enumerated, the next step is to
throw some cyrptographers and machine language programmers into the
mix, so they can find and decode the "machine-language of life".

~~James

On 5/21/06, Nicholas Thompson <nickthompson at earthlink.net> wrote:

> Hi everybody,
>
> Forgive me for casting such a wide net, but we seem to be skating very
> close to what Carl Tollander calls "artificial epigenesis and I want to
> keep the conversation as open as possible until I see who is interested.
>
> David Wilson (attached and below) has taken the discussion in the direction
> I hoped it might turn .... that selection might consist of unstable
> relations amongst stable arrays. Everybody is talking as if the elements in
> the arrays are genes, but there is no particular reason not to include
> epigenetic nodes as well.   The implication for my question on inheritance
> is that all the chaos in the genetic-epigenitic system is going on a level
> BELOW where selection is going on.  This might seem to beg the question
> concerning inheritance ... what "force" holds together the stable arrays?
> However, at this early stage of my reading, Wimsatt and Schank seem to be
> saying that the stable arrays are high  entropy .. i.e., they hang together
> because that's where randomization takes them.
>
> I am very excited about all of this, as you can see, but as you can also
> see, I should shut up and go back to reading before I say more.  Thanks for
> your patience.  Be sure to read the message below and the attachment if you
> are interested.
>
> thanks, all,
>
> Nick

~~James
_____________________________________
turtlezero.com -- its turtles, all the way down!

[James Steiner]

Another thought:

I think its important not to get locked into the view of basins of
attraction as a two dimensional construct. When there are mutiple
dimentions of change, a change in one dimension may alter the
landscape of other dimensions. In other words, any local minimum in
one direction may be coincident with a flat zone of equally viable
possibilities in another dimension of change, but drifing through that
other dimension of change alters the shape of the original landscape,
possibly changing what was a basin to a slope or peak. So "escaping"
from any one basin of attraction is really not so much of a problem,
if there are other dimensions of change still moving.

For a simple example, suppose an organism has optimized the "number of
limbs"--any mutation that changed that number would be a less viable
creature. But the "length of limbs" is still changing--minor changes
in length having little effect for this number of legs, there is no
basin, but a plateau of equally viable possibilites, with some peaks
around the edges, where limbs that are too long or too short lead to
unviable options. But, suppose that as the legs get shorter, the
usefullness of having more of them increases, so as the legs get
shorter, what was a basin in the "number of legs" landcape becomes a
point on a slope toward "more legs". Now mutations leading to more
legs are more likely to be viable. So, the organism escapes the basin,
not by a radical jump from one state to another, but by gradually
changing the shape of the landscape surrounding the basin until it
isn't a basin any more.

As the number of interrelated dimensions of change increases, so does
the variability of the landscape of any particular dimension.

So, in a system, like the DNA code that produces an organism, that has
millions of dimensions, I should think the viability landscape is a
very, very, fluid thing, indeed, and all without any "annealing" at
all.

If you add in the idea from my prior post that at this point in time
DNA may have evolved some ways of protecting its own integrity, so
reducing the scope and impact of "everyday" mutations, we create a
model where the overall multi-dimensional lanscape of change is for
the most part like a flat meadow, where the effects of most everyday
mutations are viable, punctuated with narrow spikes of unviability,
with wild, treacherous formations along the edges, where radical
random mutaion (like that caused by radiation, for example) for the
most part ruins everything, but can sometimes create something that
works.

I should also read more--I expect I'm not saying anything new, as it
seems so simple and obvious to me, and perhaps I am missing something.

Thanks,

~~James
________________
www.turtlezero.com

[Phil Henshaw-2]

I'd agree entirely that a 'fitness landscape' could be complex, that
small changes in starting position could have large effects on what
pathways open up, that taking one path could alter it, etc..   I offer
what I think are good potential examples in my new paper offering a
curiously obvious radical new mechanism for evolution that you
complexity guys might (or might not) like.
http://www.synapse9.com/GTRevis-2006fin.pdf   The question is how to get
an *efficient* mechanism for a genome to explore the fitness pathways.
I think there needs to be some way for one change to result in a
multiplication of changes from it, i.e. some kind of feedback mechanism.
If there were branching structures in the genetic code or related
organizational systems of growth, which there are, that might do it I
think.  If a step in one direction is followed by any other kind of step
in any other kind of direction it seems inefficient at best.  


Phil Henshaw                       ????.?? ? `?.????
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
680 Ft. Washington Ave
NY NY 10040                      
tel: [phone]                
e-mail: pfh at synapse9.com          
explorations: www.synapse9.com    

> -----Original Message-----
> From: Friam-bounces at redfish.com
> [mailto:Friam-bounces at redfish.com] On Behalf Of James Steiner
> Sent: Monday, May 22, 2006 8:46 AM
> To: nickthompson at earthlink.net; The Friday Morning Applied
> Complexity Coffee Group
> Subject: Re: [FRIAM] FW: Heritability and generative entrenchment
>
>
> Another thought:
>
> I think its important not to get locked into the view of
> basins of attraction as a two dimensional construct. When
> there are mutiple dimentions of change, a change in one
> dimension may alter the landscape of other dimensions. In
> other words, any local minimum in one direction may be
> coincident with a flat zone of equally viable possibilities
> in another dimension of change, but drifing through that
> other dimension of change alters the shape of the original
> landscape, possibly changing what was a basin to a slope or
> peak. So "escaping" from any one basin of attraction is
> really not so much of a problem, if there are other
> dimensions of change still moving.
>
> For a simple example, suppose an organism has optimized the
> "number of limbs"--any mutation that changed that number
> would be a less viable creature. But the "length of limbs" is
> still changing--minor changes in length having little effect
> for this number of legs, there is no basin, but a plateau of
> equally viable possibilites, with some peaks around the
> edges, where limbs that are too long or too short lead to
> unviable options. But, suppose that as the legs get shorter,
> the usefullness of having more of them increases, so as the
> legs get shorter, what was a basin in the "number of legs"
> landcape becomes a point on a slope toward "more legs". Now
> mutations leading to more legs are more likely to be viable.
> So, the organism escapes the basin, not by a radical jump
> from one state to another, but by gradually changing the
> shape of the landscape surrounding the basin until it isn't a
> basin any more.
>
> As the number of interrelated dimensions of change increases,
> so does the variability of the landscape of any particular dimension.
>
> So, in a system, like the DNA code that produces an organism,
> that has millions of dimensions, I should think the viability
> landscape is a very, very, fluid thing, indeed, and all
> without any "annealing" at all.
>
> If you add in the idea from my prior post that at this point
> in time DNA may have evolved some ways of protecting its own
> integrity, so reducing the scope and impact of "everyday"
> mutations, we create a model where the overall
> multi-dimensional lanscape of change is for the most part
> like a flat meadow, where the effects of most everyday
> mutations are viable, punctuated with narrow spikes of
> unviability, with wild, treacherous formations along the
> edges, where radical random mutaion (like that caused by
> radiation, for example) for the most part ruins everything,
> but can sometimes create something that works.
>
> I should also read more--I expect I'm not saying anything
> new, as it seems so simple and obvious to me, and perhaps I
> am missing something.
>
> Thanks,
>
> ~~James
> ________________
> www.turtlezero.com
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

[Phil Henshaw-2]

My apologies if your approach does not include worrying about the
disconnects between our maps and territories, etc.  I thought it would
be of interest.   As well as having a physics background I've been a
long time student of natural system form, using sound methods of
observation.   I'm also a very well regarded technical architect in a
first rate firm.  Architecture and Physics have a great deal in common
in terms of a deep interest in abstract form you know, and my designs
work.  

I'm also glad to ask and listen to dumb questions, the dumber the better
most times, and to smash any idea to bits and rebuild it from scratch,
including the valid seeds of any discovered conflict or complaint.  I
just don't see how we can connect with this profound new reality of ours
if we can't explain why humanity's main plan for the world system is so
profoundly disconnected from reality.   I think we just don't realize
how extensively we all build different worlds for ourselves, treating
the world as imagination, a point of view that also implies constructive
strategy.   Is that possible?


Phil Henshaw                       ????.?? ? `?.????
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
680 Ft. Washington Ave
NY NY 10040                      
tel: [phone]                
e-mail: pfh at synapse9.com          
explorations: www.synapse9.com    

[Carlos Gershenson]

Nick,

I don't know if somebody has mentioned already neutrality (search  
Google Scholar for Kimura): neutral mutations that have no effect in  
the phenotype allow the exploration of the parameter space without  
losing fitness, so it is a way out of local optima/entrenchment...
And actually there's lots of neutrality in genomes, not only to  
explore the genome space, but also as a robustness measure against  
random mutations... If every mutation you could do would change your  
phenotype/fitness, it would be very difficult to evolve anything...

Best regards,

     Carlos Gershenson...
     Centrum Leo Apostel, Vrije Universiteit Brussel
     Krijgskundestraat 33. B-1160 Brussels, Belgium
     http://homepages.vub.ac.be/~cgershen/

   ?Tendencies tend to change...?

[Phil Henshaw-2]

Part of the problem with these conceptual models is that it's very hard
to correlate them with observations.  I do entirely agree with your
proposition, though, that given the idea of fitness landscapes, that
neutral genetic change could accumulate and provide a way of 'tunneling'
your way out of a canyon.   That's particularly because of the
phenomenon of random walk that applies to neutral variation alone as far
as I can tell.   Paleo's for years have tried to apply random walk to
non-neutral change to explain unusual progressions, failing to make the
critical observation that random walk only operates when the individual
steps have no effect.  

To test that as a dominant mode of evolution would probably take data
that does not exist (the gaps problem) but could potentially be done if
large changes could be shown to not also demonstrate flow.  Flow and
random walk are anathema to each other, the structural difference
between continuity and noise.  Some of my math helps for that, but
better is needed.

The reason for my initial caution in agreeing, though, is that the whole
concept of fitness landscapes, however useful as a thought experiment,
has dubious physical existence.  The question, after all, is where form
comes from and if you just postulate that form comes from a 'template'
for everything to just jostle around and fit into, we're begging the
question entirely.  Then there's the infinite dimensionality problem,
and that innovation has to be treated and the discovery of new
dimensions of relationships.  Perhaps fitness landscapes are a mental
note pad or projection and not part of the actual subject of study.
They also presume universal random variation in the genome and what you
find is form branching in history with more change at the tips than in
the trunk.  They don't explain the kind of stability observed.




Phil Henshaw                       ????.?? ? `?.????
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
680 Ft. Washington Ave
NY NY 10040                      
tel: [phone]                
e-mail: pfh at synapse9.com          
explorations: www.synapse9.com    

> -----Original Message-----
> From: Friam-bounces at redfish.com
> [mailto:Friam-bounces at redfish.com] On Behalf Of Carlos Gershenson
> Sent: Sunday, May 28, 2006 11:53 PM
> To: nickthompson at earthlink.net; The Friday Morning Applied
> Complexity Coffee Group
> Cc: ppgb at cam.ac.uk; Jaan Valsiner; rsokol at clarku.edu;
> lrudolph; dwilson at binghamton.edu; echarles; jogreen;
> jcschank at ucdavis.edu; w-wimsatt at uchicago.edu; sbarr; Gbarker; elescak
> Subject: Re: [FRIAM] FW: Heritability and generative entrenchment
>
>
> Nick,
>
> I don't know if somebody has mentioned already neutrality (search  
> Google Scholar for Kimura): neutral mutations that have no effect in  
> the phenotype allow the exploration of the parameter space without  
> losing fitness, so it is a way out of local
> optima/entrenchment... And actually there's lots of
> neutrality in genomes, not only to  
> explore the genome space, but also as a robustness measure against  
> random mutations... If every mutation you could do would change your  
> phenotype/fitness, it would be very difficult to evolve anything...
>
> Best regards,
>
>      Carlos Gershenson...
>      Centrum Leo Apostel, Vrije Universiteit Brussel
>      Krijgskundestraat 33. B-1160 Brussels, Belgium
>      http://homepages.vub.ac.be/~cgershen/
>
>    ?Tendencies tend to change...?
>
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

[Carlos Gershenson]

On 29 May 2006, at 07:09, Phil wrote:
> Some of my math helps for that, but better is needed.
>
A recent paper in that direction...
http://dx.doi.org/10.1016/j.physd.2006.03.015
Evolving towards the hypercycle: A spatial model of molecular evolution
Camille Stephan-Otto Attolini and Peter F. Stadler
Physica D: Nonlinear Phenomena
Volume 217, Issue 2 , 15 May 2006, Pages 134-141

Abstract
We extend earlier cellular automata models of spatially extended  
hypercycles by including an explicit genetic component in the model.  
This allows us to study the sequence evolution of hypercyclically  
coupled molecular replicators in addition to considering their  
population dynamics and spatial organization. In line with previous  
models that considered either spatial organization or sequence  
evolution alone, we find both temporal oscillations of the relative  
concentration of the species forming the hypercycles as well as the  
formation of spatial organizations including spiral waves. We also  
confirm the greatly increased robustness of the spatially extended  
hypercycle against various classes of parasites.

We determine how the sequence evolution of each of the  
hypercyclically coupled populations proceeds (after an initial  
selection-dominated phase) in a drift-like manner that can be  
described by a diffusion process in sequence space. Kimura?s theory  
of neutral evolution is therefore applicable on long timescales  
despite the fact that the hypercycle exhibits extreme periodic  
changes in population sizes that are governed solely by frequency-
dependent selection.


Best regards,

     Carlos Gershenson...
     Centrum Leo Apostel, Vrije Universiteit Brussel
     Krijgskundestraat 33. B-1160 Brussels, Belgium
     http://homepages.vub.ac.be/~cgershen/

   ?Tendencies tend to change...?