Transcript POPSIM was commented by Dag at the meeting Uppsala May 14

POPSIM etc
Note that this is discussion, not
firm statements, and written very
late yesterday evening without
checking. But as we have one
hour to fill...
Dag Lindgren
Uppsala May 14
Possible reasons why Breeding
Cycler favors a 2-step strategy
more than POPSIM
• Cycling is penalized by 3 years budget and a
drop in gain for increase in coancestry. (Here
Breeding cycler may be to harsh to cycling)
• Optimization for timing gives higher weight and
longer time in field to the more efficient progenytest.
• Even other optimisations may be more efficient
in a 2 stage strategy.
• POPSIM does not credit extra gain for a two
years longer phenotypic test in two step.
POPSIM vs. BREEDING CYCLER
• Two steps, first phenotype when progeny testing
of pre-selected has been run by BREEEDING
CYCLER and when POPSIM. POPSIM runs
indicate they are similar in annual gain, while
BREEDING CYCLER found it 20% better in
main scenario. What could the reasons be??
• BREEDING CYCLER considers cost for crosses
and generation turn over and optimizes more,
including duration for field testing and cycling,
POPSIM does not. This is a possible reason for
the different results.
Timing considerations
Program
Strategy
Breeding Cycler POPSIM
Phenotypic prese Phenotypic
Progeny
Combined
Followed by progeny
Basic Intensive
Strategy
4
4
3
1
2
5
Cycling
time, total
43
37
20
33
26
20
Sel age
15
Phenotype
15
13
Sel age
progeny
20
15
12
12
15
Use of cycle
for testing
81%
81%
65%
36% 46%
75%
• Two stage is an efficient use of the
breeding cycle
• Progeny testing and phenotypic testing
uses the breeding cycle much more
effective for testing if used 2 stage wise
than if used as two independent
subsequent cycles. Thus it is surprising for
me that 2 stage does not appear still better
than phenotypic in POPSIM.
Timing considerations
Program
Strategy
POPSIM
Phenotypic prese Phenotypic
Progeny
Combined
Followed by progeny
Basic Intensive
Strategy
4
3
1
2
5
Cycling
time, total
37
20
33
26
20
Sel age
Phenotype
15
13
Sel age
progeny
15
12
12
15
Use of cycle
for testing
81%
65%
36% 46%
75%
Efficient use of breeding cycle
• Note that progeny-testing utilizes the
breeding cycle for field testing very
inefficient
• Two stage and combined utilizes most of
the breeding cycle for field testing
• Phenotypic is rather efficient but not in top
POPSIM
•
•
•
•
•
Does not adapt well to testing time. Longer field testing does not increase
gain. Most strategies use different testing time and can thus not be fairly
compared. This is not trivial to mend and need some thinking.
Does not consider costs for e.g. grafts, but cost is easily adjusted for that,
and needs no change in the code. The difficulty is actually that not much
effort has been done to estimate costs.
Calculates gene diversity loss and can make gene diversity loss equal in
different scenarios, but has no mechanism for ranking scenarios with
different gene diversity loss. But a cost and genetic penalty of gene diversity
loss can easily be made and needs no change in the code.
Invests rather much energy in creating the actual base line situation. That is
history and just set to base line. It is a bit unfortunate that the creation of the
founder population depends on a variable among scenarios, so this trivial
factor may cause false differences among scenarios. The latest will
probably be remedied but at the cost of an additional input variable.
It is bloody difficult to set up mating schemes with desires about a high
degree of PAM and restrict against half-sib mating. To restrict for half-sibs is
easier for a real breeder, but a real breeder would also find it hard to make
perfect PAM except for SPM.
Strategy 5, combined
• Development of this part seems to be very worthwhile and
for me looks as the most important and promising
development. However we have differences in opinion
about concepts and terminology, and I suggest that we
discuss that more and try to reach a consensus before
writing the actual code. The manus Dag, Darius and Ola is
one of the sources for this discussion. As it also tries to
exploit basically this strategy, it would be better if the
concepts converged. E.g. I feel more comfortable with
“breeding population” rather than “pre-selections”, and now
call Olas pre-selections for candidate parents.