[From Rick Marken (2012.07.20.1715)]
Here's an article from the NY Times about what purports to be a
description of a program that emulates the "lifespan" of an entire
organism:
http://www.nytimes.com/2012/07/21/science/in-a-first-an-entire-organism-is-simulated-by-software.html?emc=eta1
The article describes an article by entitled "A Whole-Cell
Computational Model Predicts Phenotype from Genotype" that was just
published in the journal Cell. The article itself is behind a paywall
and I probably wouldn't understand it anyway. But if they are
simulating an organism wouldn't there have to be some control loops
involved?
Anyone know anything about this?
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[From Bill Powers (2012.07.20.1835 MDT)]
Rick Marken (2012.07.20.1715) –
The article describes an article
by entitled "A Whole-Cell
Computational Model Predicts Phenotype from Genotype" that was
just
published in the journal Cell. The article itself is behind a
paywall
and I probably wouldn’t understand it anyway. But if they are
simulating an organism wouldn’t there have to be some control loops
involved?
I think we have to start distinguishing between a simulation and an
animation. An animation is what results from taking observational data
and/or observed relationships among variables and converting them to
sequences of visual images that show the observations or relationships as
if they’re taking place.
This is often used in the computer imaging industry. You record the
movements a person makes in performing some behavior like walking or
jumping or shouting, and then you draw your graphic elements so as to fit
the observer movement as closely as possible. To animate tracking
behavior, you would record all the mouse movements taking place while a
person does a run of a tracking task. Then you can draw pictures of
meerkats or monkeys or alien creatures doing the tracking task using the
recorded data frame by frame.
In a simulation, you start with a generative model that has components
mathematically like those in the real system. You connect them together,
computationally, and then start evaluating the equations every 1/60
second or so, as behavior is generated from scratch with the same inputs
or other disturbances that were present during the real behavior. You can
then adjust the parameters of the model to get the best fit between the
model’s behavior and that of the real person. That is the only time when
you use the data about the real behavior. In a simulation, you must start
with some hypothesis about how the behavior is generated, and from
that, generate some actual behavior without reference to the real
behavior. Only then do you compare the actual and the simulated
behaviors.
There isn’t enough information in the New York Times article to show
whether the reported simulation is a true simulation (I’d be surprised if
it were) or an animation. You might try writing them a polite letter
asking them which it is – and educating them a little while you do
that.
Best,.
Bill P.
[From Richard Kennaway (2012.07.21.0833 BST)]
I have access to Cell and have just downloaded a copy. The paper is:
"A Whole-Cell Computational Model Predicts Phenotype from Genotype"
Jonathan R. Karr et al
There's also an editorial summary in the same issue:
"The Dawn of Virtual Cell Biology"
Peter L. Freddolino and Saeed Tavazoie
Anyone want copies of these?
There's a YouTube video here:
http://www.youtube.com/watch?v=au4sl9CjKFU
And the project has its own web site, with downloads of their software available:
http://wholecell.stanford.edu/
···
--
Richard Kennaway, jrk@cmp.uea.ac.uk, http://www.cmp.uea.ac.uk/~jrk/
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.
[From Rick Marken (2012.07.21.0835)]
Richard Kennaway (2012.07.21.0833 BST)--
I have access to Cell and have just downloaded a copy. The paper is:
"A Whole-Cell Computational Model Predicts Phenotype from Genotype"
Jonathan R. Karr et al
There's also an editorial summary in the same issue:
"The Dawn of Virtual Cell Biology"
Peter L. Freddolino and Saeed Tavazoie
Anyone want copies of these?
Yes, I'd like a copy please.
Best
Rick
···
There's a YouTube video here:
http://www.youtube.com/watch?v=au4sl9CjKFU
And the project has its own web site, with downloads of their software available:
http://wholecell.stanford.edu/
--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
From Bill Powers (2012.07.22 MDT)]
Richard Kennaway (2012.07.21.0833 BST)
"A Whole-Cell Computational Model Predicts Phenotype from Genotype"
Jonathan R. Karr et al
There's also an editorial summary in the same issue:
"The Dawn of Virtual Cell Biology"
Peter L. Freddolino and Saeed Tavazoie
Anyone want copies of these?
I guess I do, too, though looking at the web site revealed an terrifying collection of little bits of unexplained information. One benchmark to apply: does their model predict the negative feedback control systems that operate inside the cell or relate to the relationship between cell and environment? Or is it just reproducing the average number of items of different kinds?
I note that the YouTube page refers to the result as an animation, not a simulation. I will need a lot more help than that. How can anyone possibly understand all this (and I include the author)?
Best,
Bill P.
[From Bill Powers (2012.07.22.1735 MDT)]
Richard Kennaway
(2012.07.21.0833 BST) –
I have access to Cell and have
just downloaded a copy. The paper is:
“A Whole-Cell Computational Model Predicts Phenotype from
Genotype”
Jonathan R. Karr et al
There’s also an editorial summary in the same issue:
“The Dawn of Virtual Cell Biology”
Peter L. Freddolino and Saeed Tavazoie
Anyone want copies of these?
Got the copies, and thanks. After reading the Karr et. al. paper, or the
30% of it that I am competent to read, I am convinced that these people
understand simulation/modeling and are doing it the right way. I am also
tremendously impressed by the capacity of the authors to comprehend and
model an incredibly large and complex body of knowledge. The only way for
me to understand this paper completely would be to go back to about age
20, and spend 10 or 20 years learning the biology. The authors must be
encountering some difficulties even within their own field, because their
colleagues would have to revise their educational histories in a
comparable manner to grasp the engineering principles of
modeling.
I think that in the Karr paper, in the segments where changes in
variability are described, there is some remote possibility that the
biology of reorganization theory might be found, at least as it applies
to evolution, but even as it might apply to learning in a single
lifetime. This approach also may lead to finding control systems, simply
by observing the interactions between the variables that they explore.
“Gene expression” is a pretty vague concept,. and might look
clearer in terms of control systems setting reference levels for other
control systems involved in protein synthesis.
I especially like, and have my own grounds for agreeing with, the
authors’ recognition of the preliminary nature of their explorations, and
of the guesses they have found necessary to get the model to run in a
plausible way. The important thing is that they have a running model, and
they are very clear about the idea that the purpose of this model is
mainly to find data showing how and where to correct its flaws. It
has already yielded new predictions and explanations of phenomena, but
most importantly it shows holes in our knowledge that we can address and
try to fill in. Well, that they can address. I can only watch and
admire.
So, back to my piddling little model of the reflexes, which seems very
elementary right now. I guess we all have to work within our limitations,
but it’s a bit uncomfortable to realize that some people are a lot less
limited than others.
Best,
Bill P.
[From Rick Marken (2012.07.23.1610)]
Bill Powers (2012.07.22.1735 MDT)]–
Got the copies, and thanks. After reading the Karr et. al. paper, or the
30% of it that I am competent to read, I am convinced that these people
understand simulation/modeling and are doing it the right way. I am also
tremendously impressed by the capacity of the authors to comprehend and
model an incredibly large and complex body of knowledge.
…
I especially like, and have my own grounds for agreeing with, the
authors’ recognition of the preliminary nature of their explorations, and
of the guesses they have found necessary to get the model to run in a
plausible way. The important thing is that they have a running model, and
they are very clear about the idea that the purpose of this model is
mainly to find data showing how and where to correct its flaws.
I just read it myself and agree with your evaluation completely. It reminded me that one of the cool things to do with a model is testing it’s predictions against observation. Of course, that’s what we’re doing all the time when we test the basic control model in various types of tracking tasks. But it would be fun to develop a more complex model – like Richard Kennaway’s model of a six legged insect – to see if the model predicts the effect of various manipulations on the behavior of the modeled organism. The closest I came to such a model was my two level model of bimanual coordination (http://www.mindreadings.com/Coordination.html). I got that model to match the basic two handed coordination results but I kind of gave up on it because I couldn’t get it to mimic some of the spontaneous reversal results that were found in the actual experiment.
I think your “simple” model of the muscle reflexes might actually be a nice little lab for testing how muscle/movement control works.
Best
Rick
···
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com