Transcript Optimisation of long term breeding including grandparental balance

Combining genetic gain, gene
diversity, time components, cost
components when optimizing
breeding strategies
Two parts:
Dag Lindgren
Seminar at the research school of
forest genetics and breeding
Umeå 08-09-15
1. Published 2002-2005 (many heard the essentials 2004, but not
the research school doctorands)
2. In press (galley proof received) or preparation (breeders heard
parts of it in May, but it is improved calculations, also for
meeting tomorrow).
Key issues
• Strategy;
• Genetic Gain;
• Gene Diversity;
• Time;
• Cost.
Consider all!
The program “Breeding cycler”
studies what happens during one
complete breeding cycle
Mating
Selection
Long-term
breeding
Testing
Breeding cycler explores selection of
selecting the best within a full sib family!
Acknowledgement: Large thanks to Swedish breeding for the
justification to construct a reasonable simple breeding cycler!
Breeding population size is set to 50 and each is parent to
two full sib families and from each the best individual is
recruited to the next breeding population! Thus - to optimize
Swedish breeding - it is usually sufficient to consider a single
full sib family. That is the first part.
Acknowledgement also to Ola Rosvall, who thought outside
the box, and is spiritual father of thinking outside the 50 and
a single generation. That is the second part.
Inputs
• Genetic variance components like:
– Additive genetic variation in goal character (goal
character is value for forestry but numerically volume
production at mature age other characters unchanged)
– Juvenile-mature correlation
• Time components (duration for different actions)
like:
– Testing (age at selection)
– Crossing
– Wait for flowering
• Cost components like:
– Cost for additional plants
– Cost for cycling the breeding population
– Cost for additional parents used in crosses
• Strategy: Structuring components e.g.:
– Wait for flowering - mate - establish field test - select
Combined concepts
• Genetic Gain (breeding value) and Gene
Diversity combined into “Group Merit” (the
sum).
• Cost expressed per founder and cycle
• Time and Cost combined to “annual cost”
(yearly budget)
•“Group Merit progress at annual cost”
combines all issues into a single measure,
which can be studied and maximized.
How it works…
Results
• Input parameters
• Get results
• Use “trial and error” to find what is best
Inputs – commonly used
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Chosen to be relevant for to discussions about breeding of Swedish (=Nordic)
Norway spruce and Scots pine (we have also made studies on poplars and
Eucalyps)
Many alternative scenarios evaluated to home in on optimum
Annual budget 10 “test plant equivalents per year and founder”
Cycling cost 30 per founder
Crossing possible at age 10-15 for pine and 20 for spruce
Clone testing possible for spruce but not pine
CVAm = 14% (additive variation in value (volume) among trees at mature age)
Dominance variance ¼ of additive
Heritability almost 0.2 (within family heritability =0.1)
Note than in breeding cycler papers 2000-2005 is the population considered a
single full sib family, thus variance components are within family. That is
explained and correct, but may still be misleading. In coming papers we will
give it for whole population.
Selection gain is created by selecting the best individual from each full sib or
through one offspring per grandparent.
Diversity loss with balanced 50 founder breeding, “group coancestry”, is
punished as percent of forest production (probably high estimate)
Rotation age 60
Juvenile-mature correlation according to Lambeth (1981).
Comparison of main testing strategies
– best genotypes are selected based
on
• Progeny – trees with good progeny
• Clone – trees with good vegetative copies
• Phenotype – trees with good appearance
0.6
0.5
Annual Group Merit, %
0.4
Clone
0.3
Phenotype
0.2
Progeny
0.1
0.0
0
0.1
0.2
0.3
0.4
0.5
0.6
Narrow-sense within family heritability
Clone test strategy is best
• Clone was superior for all realistic scenarios.
• Swedish breeding uses clone testing with near optimal
design (acc to breeding cycler) for Norway spruce and
has initiated it for lodgepole pine. Development with
clone testing is initiated for Scots pine. These
approaches are strongly supported by BREEDING
CYCLER results.
• The optimal scenario for Norway spruce suggests later
selection age in field trials than Swedish breeding heads
for (20 instead of 15 years). The late sexual maturity of
Norway spruce is an argument.
Phenotypic selection may be
slightly superior to Progeny-testing,
for the most realistic scenarios
Time for initiation (flowering)
Annual Group Merit , %
0.20
Progeny
0.15
0.10
Phenotype
0.05
0
3
6
9
12
Delay before establishment of
selection test (years)
15
18
Early flowering improves the efficiency of progeny-testing only
marginally.
Early flowering may make progeny-testing marginally superior to
selection on phenotype
Annual budget
Annual Group Merit , %
0.3
0.2
Phenotype
0.1
Progeny
0.0
0
5
10
15
20
25
Budget per year and parent
Phenotype get more superior at low budget, progeny
slightly superior at high budget
As phenotypic and progeny are
similar, it seems logic to combine
them in a two-stage strategy.
Pre-selection of phenotypes, which
are progeny-tested and the best
reselected.
Waiting time to sexual maturity is
used for phenotypic testing!
Phenotype/Progeny strategy
Mating
Stage1: Phenotype
test and pre-selection
Reselection
based on
performance of
the progeny
Stage 2:.Sexual
propagation of
pre-selected
individuals
Testing of
the progeny
Phenotype/Progeny strategy
superior to either single stage
strategy for most relevant cases
At high budget, successful flower
stimulation and low heritability, progeny test
is superior.
At low budget, high heritability, low cycling
cost and low penalty for diversity,
phenotypic selection is preferable.
• Early flowering at
age 3 can make
progeny-test
compatible with
2-stage
• Sexual maturity at age
10 seems optimal, but no
marked loss if sexual
maturity occur first at age
15.
Annual Group Merit (%)
Result: 2-stage seems preferrable
to Scots pine
Main scenario
0.6
Pheno/Progeny
0.3
Progeny
0.0
0
5 10 15 20 25
Age of mating for progeny
test (years)
Suggested good scenario for Scots pine
(for one family at given annual budget)
Cycle time= 26
Gain=8%
Select the best
when progenytest is 10 years
Stage 2. Progenytest with 30
offspring
Mating
3 years
Stage 1: Test
70 phenotypes
Long-term
breeding
Cross
Polymix
(3 years)
“Preselect”
Select 5 at age 10
Size of the breeding population
• The size of the breeding population (50)
has been chosen because it is “sufficient”.
We tried to optimize with breeding cycler!;
• 50 (as used in Sweden) seemed about
right for spruce and pine in Sweden;
• May be a little lower for Norway spruce;
• Maybe a little higher for Scots pine;
The rest of the stuff is not published
(one study in press, one in manus)
• Idea: Keep balance by monitoring
grandparents instead of parents. Each
grand-parent give the same contribution
but parents/grandparent is an input.
• “Grandparents” can be regarded a
synonym to “founders” for the situation in
Sweden (50 founders/bpop).
Main result - short
•
Scots pine in central Sweden with 50 founders (=grandparents)
New method (balance among grandparents, 300 parents)
Earlier assumed best strategy (phenotypic selection strategy, 50 parents,
Hannrup et al 2007)
Comparison assuming the same budget and gene diversity
Annual Genetic Gain, %
•
•
•
0.3
}
0.25
0.2
= 54%
0.15
0.1
0.05
0
New
Current
Why phenotypic selection?
• Well-documentet (CJFR by Swedish
breeders
• Advantage of 2-stage not large
• Special case of granparental balance with
parents/grandparent = 1.
SPM with parental balance
(almost current Swedish program)
Grand parents
(=founders), F0
Mating grand parents
Select and mate 2 best sibs
(…)
F1
(…)
(…)
F2
(…)
Green trees
show pedigree
Multiple SPMs
Grandparents
F0
=founders
Cross 4 best sibs
(…)
F1
F2
(…)
(…)
1st rank family
(…)
(…)
3rd rank family
nth rank family
Cross e.g. 4 best sibs in the 2 best
families (2 parents per grandparent)
(…)
2nd rank family
• Note that retrospectively SPM and multiple
SPM have identical pedigrees, thus
identical increase of coancestry.
• Simple SPM (phenotypic selection) is a
special case of multiple SPM (“combined
selection”) with 1 parent per grandparent.
Standard scenario
• Heading to be relevant for Scots pine in central Sweden
• Costs, field trial plant is the unit, cost components are
derived from Hannrup et al. 2007. Genetic variance
components derived from Rosvall et al 2001.
• Annual budget 50/grandparent
• Scots pine, cycling cost 100/grandparent
• Added parent 50
• 6 parents/grandparent (near optimal)
• Rotation time 70 years
• Field test selection at 15 (optimal slightly lower but
marginal reduction)
• Breeding cycle length 20 years.
• Juvenile-mature correlations for Scots pine derived from
Jansson et al (2003).
0.5
Jansson et al. 2003
Gwase et al. 2000
Lambeth & Dill 2001
Lambeth 1980
Annual genetic gain, %
0.4
0.3
0.2
0.1
0.0
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Time, years
Jansson et al. 2003 is for Scots pine in southern Sweden, which is most
relevant. 15 year testing time seem near optimal.
Genetic gain at different parents
per grandparent
Annual gain, %
0.30
0.25
Parent cost=50
Optimum at 6
54 %
better
0.20
Phenotypic strategy
0.15
0 1 2 3 4 5 6 7 8 9 10111213141516
Number of parents per
grandparent
0.30
Annual gain, %
Cparent=0
Optimum is not
strongly
dependent on
parent cost
0.25
0.20
Cparent=100
Cparent=50
0.15
0 1 2 3 4 5 6 7 8 9 10111213141516
Number of parents per
grandparent
Optimum number of parents per
grandparent
16
Cparent=0
14
12
10.5
Cparent=50
10
8
6
6
Cparent=100
4
4.5
2
0
10
50
100
Annual budget per grandparent
Optimum
P/GP rises
linearly with
budget
This study supports that the
suggested strategy 5 (which will be
more discussed 090816) is the best
way to breed Scots pine, and
suggest the strategy is substantially
better than other alternatives.
The details of the strategy as
suggested in end of December
2007 seem optimal.
This is however further
investigated.
Note that the start (“phenotypic
preselection) is identical to the two
stage strategy, final decision need
to be done first at crossings
end