Re: DNA carries information


"r norman" <r_s_norman@xxxxxxxxxxxx> wrote in message news:0b6pf21r2t8hf722dt200f2f22jm9rqaq8@xxxxxxxxxx
On 4 Sep 2006 13:37:47 -0700, "Friar Broccoli" <EliasRK@xxxxxxxxx>
wrote:

Perplexed in Peoria wrote:

[extensive snipping]

And, as a final caution, I would urge you to be clear on
the distinction between three things, all of which are
sometimes called 'information'. One is information as
capacity, one is information as quantity of contents, one
as information as distinctive quality of contents.

The information (I am speculatively asserting) is the degree to
which a user (an IPU) can act efficiently based on the
measurement in whatever form presented. Thus there should be
no need to account for such distinctions.

Well, then it seems that you are maybe defining "fitness" rather
than information. Take a look at this paper:
The Fitness Value of Information
Carl Bergstrom and Michael Lachmann
http://arxiv.org/PS_cache/q-bio/pdf/0510/0510007.pdf
which shows how fitness and information can be related.

Right now, I am feeling particularly uninspired by this topic,
but I absolutely wanted to get back to you on this.

During the last 2 weeks I have swung back and forth on the
question of whether their "fitness value" of information is the
same thing that I mean by "efficiency value" but I now think it
is.

I think I was confused because they treat an IPS (apparently
always a complete organism) as a black box, and never ask
anything about the internal workings. Since we were focussed
on a detail (DNA), it took me some time to develop a coherent
model.

I was disturbed to notice that "fitness value" highlights a
weakness, I had hoped "efficiency" would overcome. If someone
is a good liar, then he can be very fit if the measure of
fitness is making money, or self-replication. Thus, false
information is information.

Although it remains possible to distinguish a Ray Martinez
sentence (which is either unconvincing or just incoherent) from
an rnorman one, the inability to get rid of false info
is annoying.

Hmmm. Several distinctions are in order here. The B&L paper
deals only with the fitness value of information to the receiver,
not the fitness value to the sender. Also, the problem of
false information is not addressed. However, they do provide
the germ of a distinction between significant (or relevant)
information and insignificant information. More on this below.
Perhaps false information (from a used car salesman, for example)
can be swept under some philosophical rug by noticing that
the salesman's claims are not to be considered by the receiver
as significant because the buyer has no reason to trust the
seller. And therefore, he has no incentive to even listen to
those claims.

I am also not too sure what the authors have actually done. Dr
Norman has been arguing that entropy and information are not
necessarily the same thing (or in the same class), just because
they follow the same mathematical rules.

So what does showing that some fitness values follow Shannon
information logic really show?

Is a mathematical model like an analogy? Things are the same
only when the entire model can be overlain from one system to a
second?

On the other hand, I don't think that asking if two different
(even identical things) are the same is a meaningful question.
On that confusing note:

So you made me read the Bergstrom+Lagstrom paper, too. It is an
interesting ecology-type analysis of fitness, but there is essentially
zero in it which might be appropriate to whether DNA carries
information. The information they consider deals with gaining
information about the environment, and is completely abstract,
completely conceptual. There is no notion of looking at, smelling,
tasting, making any measurement on the environment, nothing at all
like that. Only the fact that "knowledge" of the environment given
that the environment is variable, helps evolution select an
appropriate strategy for selecting phenotypes to deal with that
environment.

The paper's introduction opens with a paean to information: "Living
organisms acquire, store, process, and transmit information and as
such, information is a central organizing principle in biological
systems at every scale from the digital coding in DNA to the
long-range calls of cetaceans" citing John Maynard Smith. The idea of
information in biology. Quart Rev Biol, 74(4):395-400, 1999.
Unfortunately, it really doesn't say anything. This is the kind of
talk that seems to drive John Wilkins crazy. I accept it since I
believe in the 'information processing' ability of higher level
complex systems even though the molecular processes run perfectly
fine with the notion.

I presume you meant "without the notion".

As to what drives Wilkins crazy, I am not much deterred by such
considerations, nor do I think that John would want me to be deterred.
I hope that I am not distorting his position too much by caricaturing
it as a belief that a forest is best understood by looking at each
of the trees. Whereas I think that there is something to be gained
by taking a wider perspective. Nothing that can't be also gained by
taking John's approach, to be sure, but John's approach will not
complete the census of the forest in my lifetime, so I like to
supplement it with other methods to answer just some of the
system-level questions more quickly.

However the true state of the paper is represented by its closing
argument in the conclusion section: "... we should take
a closer look at the concept of information: information is the
reduction of uncertainty, where uncertainty measures the number of
states a system might be in. Thus mutual information between the world
and a cue is the fold reduction in uncertainly about the world after
the cue is received." Given this definition, it is not surprising
that Shannon information and the entropy measure emerges because that
is exactly what those computations evaluate. Then, later, "We can
now see why the concept of information is the same across different
disciplines. In biology, fitness refers to the fold increase in the
number of surviving lineages. In communication theory, information
refers to the fold increase in the number of messages to encode. In
physics, entropy refers to the fold increase in the number of possible
states in phase space."

In this context, the information "encoded" by DNA is the reduction in
uncertainty in the type of organism produced by the expression of that
DNA. If two sequences result in the same phenotype, the differences
between them carry no information And you must look at the actual
final organism in toto, not merely at the amino acid sequence of a
single peptide if you want information in an evolutionary sense. If
two sequences produce essentially the same "kind" of organism (i.e.
with the same fitness in identical environments), then the differences
in their genome carry no information. The problem is that there is an
enormous set of sequences all of which produce essential the same
living product and at this time we have no real way of cataloging
these. We can only identify specific changes that result in
distinctly different phenotypes, but without the complete set of
possibilities, without knowing the "phase space" under consideration I
don't see how to quantify the "information content" that we can
identify.

A very nice summary of the paper. I will add just a few comments.

Elsewhere, I commented that the only equations to be found in an
information theory textbook are definitions. The real content
of the subject lies in a set of inequalities. For example, Shannon
gives a formula for the information capacity of a channel, but
the real theorem here is that Shannon's formula provides an upper
bound on information flow. To approach this bound, considerable
thought has to be given to the encoding scheme, the balancing of
the statistics of 1s and 0s, etc.

Similarly, B&L don't show that effort devoted to acquiring information
about the environment always pays off in fitness. Instead, they
present an inequality. The fitness gain that can be realized by
paying attention to a changing environment is bounded by the
quantity of information acquired. But to approach this bound,
the organism has to optimize its information collection machinery
so that it only expends effort on acquiring useful, relevant, and
significant information. Information is valuable only if you
can make appropriate responses.

As you point out, there is nothing directly relevant here to the
issue of whether DNA carries information. But I think that an
extension of the framework in that direction is possible. A
micro-organism can use either of two strategies to deal with
a changing environment. Either it can invest in sense organs
and phenotypic plasticity as B&L analyze. Or, as an alternative,
it can follow a policy of generating heritable genotypic variation,
and let natural selection serve as the 'IPS' collecting information
about the environment and responding appropriately. My hope is
that some kind of inequality can be written down dealing with
this approach to biological information processing. I also hope
that some adaptationist arguments and empirical measurements can
show that nature actually approaches the theoretical bounds.
Here, then, we have significant information about the environment
encoded in the organism's or species's DNA.

.

Relevant Pages

  • Re: DNA carries information
    ... The Fitness Value of Information ... the germ of a distinction between significant ... tasting, making any measurement on the environment, nothing at all ... information in biology. ...
    (talk.origins)
  • Re: DNA carries information
    ... the distinction between three things, ... The Fitness Value of Information ... always a complete organism) as a black box, ... tasting, making any measurement on the environment, nothing at all ...
    (talk.origins)
  • Re: DNA carries information
    ... The Fitness Value of Information ... the germ of a distinction between significant ... tasting, making any measurement on the environment, nothing at all ... considerations, nor do I think that John would want me to be deterred. ...
    (talk.origins)
  • Re: Felsenstein and reproductive excess
    ... >conspecifics constitutes a deterioration in the environment for you. ... >model which would allow fitness to "increase" over time without population ... we have to distinguish between mutations that make an individual ...
    (sci.bio.evolution)
  • Re: Heritability of fitness
    ... >> the rate at which information 'about the environment' is accumulated ... >> in the genome and the heritability of total fitness. ... >> should be very close to zero. ...
    (sci.bio.evolution)