Hi

This is a very stimulating discussion.  It is always good to hear from
people who struggle with similar challenges and I find the attempts to
better model the genetic impact on crop growth and development very
interesting.

At the South African Sugarcane Research Institute we have been trying
to "unravel the genetic and environmental aspects" by re-defining
genetic parameters and then comparing simulations with experiments for
given genotypes at various environments.  It appears that we continually
have to redefine parameters at lower levels (simpler traits) in the
physiological processes to be able to remove yet another layer of
environmental variation that reveals itself at each new redefinition.  A
challenge related to this is that often the data to directly measure
these parameters are not available or are very difficult to obtain, with
the result that their validity  is assessed by comparing simulated and
observed values of higher level variables (complex traits).

A simple but maybe a good example is the method of simulating plant
development using the concept of thermal time and a base temperature
(defined as the temperature below which a process rate is zero).  For
sugarcane canopy development we believe that this parameter has a strong
genetic component to it - some genotypes seem to have values that are
quite different from others.  However, to complicate  matters, we also
found that the apparent base temperature (as derived from periodic
measurements of canopy stage and daily temperature) was different for
different environments (we got the best fit by changing the base
temperature).  I believe that in this specific case this was due to the
inadequacy of the thermal time model (effective temperature increases
linearly with temperature and it has no maximum limit).  When we refine
this model we may see that the apparent base temperature varies less
with environments.

Work and comments by Graeme Hammer, Gerrit Hoogenboom, Jeff White,
Stephen Welch and others  have inspired us as to try and correlate QTL's
of a sugarcane mapping population with simple traits, such as unstressed
stalk elongation rate per unit thermal time and the size of the first
few leaves. It is still early days and we recognize that these traits
are still far removed from genes and that there a multitude of traits to
be considered, but we have to start somewhere, I believe.  We do not
have any results yet to report.

I agree with Stephen that this approach " permits much more discerning
tests of model validity than what was possible before".  Although I have
not read Matthias's paper yet, I think that results from simulations
with different values for genetic parameters for one genotype over
different environments will have to be interpreted very carefully.  I am
looking forward to reading Matthias' paper.

Forgive me if I have bored you, I enjoyed the discussion and I am
interested to hear any further views on this topic.

Thanks




Abraham Singels
Principal Agronomist
S.A. Sugarcane Research Institute
Private bag X02,
Mount Edgecombe 4300
South Africa

Tel: 031-5087446         Fax: 031-5087597
[log in to unmask]

>>> Stephen Welch <[log in to unmask]> 08 March 2008 01:13 >>>
Hello List:

While I have read the later posts in this thread I would like to
address
one below as it touches on two topics that are of importance:  (1) the
relationship between “genetic coefficients” and “real genetics”
and (2)
the constancy (or lack thereof) of genetic coefficients.

The genetic (read also “metabolic”) systems of plants are, indeed,
complex; Arabidopsis, a relatively simple plant, has over 25,000
genes.
 However, (1) gene networks have enormous redundancy, (2) individual
genes that are far (in network distance) from major developmental
switches may have limited individual effects, (3) a great many genes
are involved in producing tissues that are common across all species
and that
 therefore need not be counted against those that determine
the
idiosyncrasies of one particular crop, (4) others mitigate against
stressors not present in even marginally optimized agricultural
settings, (5) etc.

Viewed from a cybernetic perspective, gene networks have huge innate
capacities to process information.  If the totality of a plant's
theoretical genetic computational power were routinely utilized to
determine its phenotypic outcome, then no physiological crop model
currently in existence would have any predictive skill whatsoever.
As
this is clearly not the case, there is no reason to be intimidated by
apparent genomic complexity, although, there is, to be certain, much
interesting work to be done.

Thus, “using a handful of so-called ‘genetic parameters’ is
[indeed, an]
.. approximation” but (1) how “very crude” it is and (2) to what
extent the adjective “genetic” is justified are questions both
susceptible to direct study and of major interest to both public and
private sector labs interested in exploiting crop models within
breeding programs.

A review relevant to this topic is Hammer et al (2006), Trends in
Plant
Science, 11:587-593.

Consider some combination of a genetic coefficient and a plant trait
that it influences.  The short story is that there may be identifiable
genomic regions containing genes that influence the trait in the same
way that the genetic coefficient affects model predictions of that
trait.  But not always.

Such regions (quantitative trait loci; QTL) can be searched for by
calibrating the model across one or more sets of environments to each
of the genetic lines in some mapping population.  Then the parameter
of
interest is interpreted as a quantitative trait of each line and
mapped
to the genome using standard methods.  Such region(s), if found,
justify the appellation “genetic coefficient” for the parameter
involved.

With one caveat.  Crop modelers often assert that their products
“disentangle GxE interactions”.  If QTL for a parameter are found,
but
the contributions of those QTL to the trait of interest are
environmentally dependent, then the GxE interaction has not been
properly (or at least fully) partitioned.  Such a finding implies a
need for model improvement.  Specific leaf area is one “genetic
coefficient” that has failed this test (Reymond et al, 2003 as cited
in
the Hammer et al paper).

As noted in other posts, the genome of a particular line is fixed.
“Genetic” coefficients that “vary with time” are, in actuality,
varying
by environment and therefore point up the need for improvements
somewhere in the modeling process (in model formulation, in
calibration, or in both, or somewhere else entirely).

A positive feature of all of this is that evaluating the constancy
across environments of quantities asserted to be genetically
determined
constants may, in combination with existing goodness-of-fit measures,
permit much more discerning tests of model validity than have been
possible to date.

Stephen M Welch
Professor of Agronomy
Kansas State University
USA

Quoting Matthias Langensiepen <[log in to unmask]>:

> Dear all
>
> the recent concern of Dr. Andarzian about our
> calibration of CERES-Wheat requires a response:
>
> It was not easy to critize a model which has
> taken a lot of efforts to construct and
> which is still in widespread use as demonstrated
> on this list server. I deeply admire the authors
> of DSSAT who contributed significantly to modern
> crop modelling and provided a wealth of inspirations
> for advancing crop research.
>
> Our motivation to carry out this study and the
> discussion of its results are described in the
> paper which is the reason for not quoting
> them again. However, I would like to respond
> to the genetic coefficient issue:
>
> Bahram Andarzian is right in a way that the genome
> of a plant is fixed. 30 years after the CERES
> model was formulated we are able to decipher the
> genome of a
 plant and can potentially get fascinating
> insights into its metabolism. Practically, however, this
> is like getting a book which we have waited for for a long
> time, but are unable to read. Millions combinations
> of metabolic pathways are possible and we are still
> very far away even from crasping the complexity of plants.
>
> Using only a handful of so-called "genetic parameters"
> is a very crude approximation of this complexity which
> is necessary to allow for a practical application of
> the model. We do not critize this pragmatic approach.
>
> What we do critize, however, is that the majority of
> DSSAT-users do not allow for their changes over time.
> A farmer who has cultivated a field for 30 years
> knows very well that no crop season is like the other
> and that plants respond to these fluctuations in flexible
> manners. The underlying biological mechanism is differential
> gene activity which results in numerous adaptation strategies
> which can differ greatly between seasons. The extreme
> seasonal differences of weather conditions at
> Schleswig-Holstein (ranging between approx. 250
> and 1000 mm rainfall per year with no regular
> distribution, for example) forced us to calibrate
> the model for each season separately. I hope
> this clarifies the issue (We strictly
> followed the user guidelines by the way.)
>
> Plants are more clever than we often think.
> Francis Halle, a well known botanist from France,
> quoted the French writer Michel Luneau in this context,
> "who knows how to make the tree speak: For us, say the
> trees, all is connected so that there is no need for
> any particular centralization. Our internal organization
> recognizes neither God nor a master. It is a free association
> of elements of different and complementary organs.
> These obey nobody but themselves and ask of
> their followers a simple and essential agreement:
> growth. Each organ is free in the means by which it
> attains growth. To each according to its
> inspiration..." (Halle F. 2002. In Praise of Plants.
> Timber Press. Page 99).
>
> Do we need separate crop coefficients for each plant organ
> and season ?
>
> Have a nice weekend.
>
> Matthias
>
>
>
>
>
>
> Dear my friends DSSAT servers
>    Hi
>    Please if possible, see the article entitle" validating
> CERES-wheat
> under North-German environmental condition" by M. Langensiepen and
> co-authors in Agricultural Systems Journal (article in press). This
> article challenges the using and performance of CERES-wheat model to
> simulate grain yield and biomass production under different water
and
> nitrogen conditions.
>    In my idea genetic coefficients are cultivar-dependent and should
> not
> be changed over years, but in their calibration procedure, they
> assume
> genetic coefficients are environment-dependent which according to
> environmental condition of each year have changed! If so, what is
the
> mean of the yield variability? In my idea, yield variability is the
> yield of a crop over different years or in the other word, running
> the
> model with fixed genetic coefficients over different years.
>    Apparently, it seems that they did not calibrate water and
> nitrogen
> modules in their work!
>    This case may be a good discussion topic for DSSAT servers and
> sharing information about strength and weakness points of the DSSAT
> models with each others.
>
>    Best
>    Bahram Andarzian
>    Ph.D in Crop Eco-physiology
>    Agricultural and Natural Resources Research Center of Khuzestan
>    Ahvaz-Iran
>
>
*********************************************************************
This email and all contents are subject to the following disclaimer:
Http://www.sasa.org.za/emaildisclaimer.asp
*********************************************************************