Size of breeding population

1
Size of breeding population

• Dag Lindgren and Darius Danusevicius
• DaDa
• March 2004

For some reason all slides are not visible from
my portable, but from other computers
2
Start with two other matters

• The Swedish long-term breeding is still rather
  open as most of it has not really started yet,
  thus easy to reconsider.
• A number of possible limitations on breeding
  cycler has to be mentioned and discussed

3
Swedish long-term breeding - Intentions

• 20 subpopulations with BP50 (Gpop and Tpop)
  for spruce and pine
• Norway spruce
• Karlsson Rosvall (1993) suggest 4014 ramets
  per parent in a 20 year cycle
• Scots pine
• Wilhelmsson Andersson (1993) made a suggestion,
  which depends on the success of progeny-testing
  F1 uncertain cycling time

4
Swedish spruce breeding status 2002 per
subpopulation
5
Standard pine strategy
Picture stolen from Bengt Andersson
2001http//www.metla.fi/tapahtumat/2001/nordicgen
etics/Bengt_Andersson.ppt
6
Swedish Scots pine breeding status 2002 per
subpopulation
7
Scots pine - addition

• Tpop 11 has run a whole cycle to F2, seems
  phenotypic in DaDa language.
• Tpop 17 OP seeds have been collected from 20
  year old F1 8 progeny tests with 25000 plants.
  If cycle length is 25 and BP50, 25000/(5025)20
  , so if that was typical annual cost for pine
  would be gt20 trees per BP parent and year, seems
  phen/prog in DaDa language
• I could find only a single case (Tpop17) of major
  progeny-testing F1 initiated. Things take time.

8
That long-term breeding takes much time is no
surprise. Thanks to the current good
documentation (Annual status rapports are
available), we will better grasp the time-line
and avoid delays. Long-term breeding has not
proceeded far for most Swedish populations, thus
the methods of long term breeding are not
well-established or based on much experience, but
open for discussion. It has been thought that
methods to force early flowering to get
progeny-test initiated really early on pine
should become important. This has not yet been
done for a single Tpop. Our calculations indicate
that this line of breeding is less efficient than
progeny-testing of field-tested F1 genotypes.
9
Alternative pine strategy (cheap)
Top grafting
Polycross
Top grafts (520)
Planting
Planting
Progeny test 20540 PC 30520 OP 7000
Next Cycle BP DPM
DPM
Select 5 best in 20 best fam.s
Select 5 in worse 30 fam.s
Grafting
Graft archive (530)
pollen
Field test 5000 F1 (50100)
30
0
2
12
18
Year (approx)
10
Phenotypic vs progeny

• In future seed orchards tested clones will
  probably be preferrable to somewhat better - but
  untested - fresh 15 year old phenotypic
  selections
• Progeny offers options which phenotype does not
  (like observations on survival, estimates of
  genetic correlations and other parameters).

11
Constraints and limitations of breeding cycler

• Shit in shit out, entries must be chosen with
  care. Sometimes they are not important but
  sometimes they are
• The input values may need some adjustment from
  the most evident for considering factors not
  considered in the math
• Breeding heads for an area and gets information
  from a limited number of sites with limited
  materials, this can be considered by a reduction
  of CVAm
• The test environments may not be considered a
  sample of future forest environments, this can
  also be considered by a reduction of CVAm.

12
Constraints and limitations of breeding cycler
continued 1

• If ever leading for the details for Sweden (or
  elsewhere) I recommend breeding cycler to be
  rerun after a more engaged debate about the
  inputs. We are willing to do the reruns.
• Any decision support tool may need modifications
  for conciderations beyond the model.
• Breeding heads for improvement in many
  characters, we set the goal as one character
  value for forestry and the observed character
  can be seen as an index with as high correlation
  as possible to the goal.
• The observation is an index of observations,
  and JM is thought of as time development of
  observed vs goal.

13
Constraints and limitations of breeding cycler
continued2

• Plant cost is seen as independent of age of
  evaluation. This can cause problems for some type
  of comparisons (e.g. expensive flowering
  induction). This difficulty can be overcome by
  inserting different costs for different compared
  alternatives.
• It is easy to add many types of considerations to
  the EXCEL sheet, but it makes it difficult and
  complex for the user and journal papers. Those
  who want a special feature can often rather easy
  program it into the existing breeding cycler or
  even I can do it if you ask.
• Phenotypic preselection reduces the genetic
  variance somewhat in our pheno/progeny
  calculations, quantitatively the effect is
  negliable (3) in our main scenario.
• The gain by within family phenotypic selection
  may be slightly overexegarated with large
  families. This effect is probably small and very
  depending on as well the experimental lay-out as
  the evaluation method.

14
Constraints and limitations of breeding cycler
continued 3

• The optimal breeding strategy is too chaotic to
  be found by formulas, stochastic simulation is
  needed instead of breeding cycler.

15
Constraints and limitations of breeding cycler
continued 4

• Breeding cycler does not specifically link to
  seed orchards. To cream off seed orchards with
  sufficient diversity and little inbreeding is the
  most important reason for tree breeding! But this
  is taken care of by the choice of the diversity
  coeff (penalty).

16
Constraints and limitations of breeding cycler
continued 5

• Genetic correlation are likely to change over
  generations
• Breeding cycler considers the first cycle, the
  cumulative effect of some cycles can be assumed
  to be additive, but after a number of cycles this
  oversimplifications become unrealistic.

17
Swedish Norway spruce breeding
ConclusionsAccumulative progress for many
generationsOrchard progress follows breeding pop
progress
18
Inbreeding follows group coancestry
Simulation of Swedish Norway spruce breeding
program by POPSIM, BP48, DPM, equal
representation (2/parent)
0.08
0.06
f
Probability of identity by descent
0.04
0.02
ConclusionsAccumulative change over many cycles
0
0
2
4
6
8
10
Generations
Rosvall, Lindgren Mullin 1999
19
Cycling will accumulate gain. Where is the limit?
20
Balanced long-term breeding

• Some unbalance is favourable
• The inoptimality loss seem to be small it is
  tricky to utilize unbalance, and the balance is
  unlikely to affect recommendations much.

21
Imbalance at initiation of breeding

• At initiation of breeding there is no balance.
• Truncating tested plus trees to long term
  breeding has sometimes been done with inoptimal
  imbalance.
• I believe it is more optimal to sacrify the gene
  diversity in the initial selections slower.
• This has been discussed i förädlingsrådet 1999
  (see link on seminar page), and one argument is
  in Routsalainens thesis (2002).

22
Sweden started imbalanced

• Swedish recently decided to decrease the breeding
  population drastically ( very low Ne first
  generation),
• 6000 plus trees ? 1000 founders

23
Flowering when?

• We assume flowering in phenotypes of conifers at
  ages around 11. That may work with top-grafting.
  To make progeny-testing with open pollination
  possible may take 20 years.
• It is important that hard data on flowering time
  and flowering variations in modern progeny trials
  are collected and well documented.
• In our current figures we probably have
  discounted early flowering assuming pollen
  collection, hormone injections and top grafting
  whenever possible. It is also important to
  document and systematize how efficient such
  actions are.

24
More constraints on breeding population number

• An unpublished manus, which was produced during a
  hectic month in the autumn 2003, is not much fuel
• Results are contraintuitive to me, before the
  study I thought the study would suggest larger
  population size
• Decrease of breeding population is something one
  should be very conservative about
• No immediate reason to go downward!

25
Earlier considerations on breeding population
number

• Most considerations (including Swedish) is to
  choose the lowest number, which ensures
  sustainability and conservation and assurance
  against loss of alleles, not optimal trade-off
  with gain.
• It has not been well investigated if the optimal
  number could be higher than applied.

26
Genetic progress as function of Ne
Expected gain after 1, 5, 10, and infinite (?)
generations of selection for different values of
Ne in a model population (Baker and Curnow 1969
from Johnson et al 2001 http//www.fs.fed.us/pnw/p
ubs/journals/Johnson_StClair_Lipow_2001.pdf). Gen
eration Ne 1 5 10 ? 4 3.3 12.4
19.5 28.4 16 3.3 16.0 31.4 114.5 32 3.3
16.8 34.6 177.5 64 3.3 17.2 36.4 220.9 256
3.3 17.5 37.8 240.0 ? 3.3 17.6 38.0 240.0
Ne gt 50 in one subpop and much larger in the
metapop, so we may not constrain future genetic
progress
27
BP sizes reported in tree improvement programs
E globulus CELBI -Portugal 300
E globulus APM - Australia 300
E grandis ARACRUZ Brazil 400
E nitens APM Australia 300
E nitens New Zealand 270
E. regnans APM Australia 300
E. regnans New Zealand 300
E. Urophylla ARACRUZ Brazil 400
Picea abies Sweden gt1000
Picea glauca Nova Scotia 450
P mariana New Brunswick 400
Pinus elliottii CFGRP - USA 900
Pinus elliottii WGFTIP - USA 800
Pinus radiata STBA - Australia 300
Pinus radiata FRI New Zealand 550
Pinus taeda NCSU- USA 160
Pinus taeda WGGTIP 800
Pseudots. Menziesii BC - Canada 450
Pseudots. Menziesii NWTIC - USA 404
Tsuga heterophylla HEMTIC CAN-USA 150
Note sometimes values refer to what is available for a zone but mostly a meta-population for a larger area Note sometimes values refer to what is available for a zone but mostly a meta-population for a larger area Note sometimes values refer to what is available for a zone but mostly a meta-population for a larger area
Pinus banksiana Lake states - USA 400
Pinus banksiana Manitoba -Canada 116
Pinus caribea QFS - Australia 200-300
from Johnson et al 2001http//www.fs.fed.us/pnw/p
ubs/journals/Johnson_StClair_Lipow_2001.pdf).
28
Comment The Swedish breeding population seems
unusually large (this may be justified by the
ecological amplitude covered by the Swedish
breeding program)
29
The current Swedish breeding program is
sustainable for more than 10 generations.Ten
generations downstream will offer new unknown
options. There is gene banking and natural
resources besides breeders activities. The
current breeders responsibility do not stretch
longer.Question 1 Can it be made more narrow to
save money or to boost gain without loosing
sustainability?Question 2 Are there gains to be
made by enlarging BP
30
How many are needed and desired?
31
Summary
Census number 50 per subpopulation and 1000 in
meta-population for pine and spruce is OK to
continue in the coming decades.
32
Optimal breeding population size

• What is more beneficial at a fixed budget larger
  Breeding population (diversity) and smaller
  testing pop (test precision) or vice versa?
• Find the breeding population size which maximizes
  the annual progress in group merit under the
  annual budget

33
How optimally allocate the resources
Test precision (testing pop.)
Diversity (breeding pop.)
34
Main findings

• Spruce, BP lt 50 is beneficial, as clonal test
  higher benefit from the gain-generating capacity,
  
• Pine, BP 50, as the gain-generating capacity of
  the testing strategy is not powerful enough to
  motivate reduction of gene diversity.

35
Parameters and scenarios
36
Testing strategy
Annual progress would benefit from lower BP for
spruce, where testing can be clonal!
37
Testing strategy

• For Phenotype the optimal is above 50, but only
  slightly. However, if budget is higher, it is
  better to increase breeding pop size beyond 50,
  than to increase offspring size.
• For Progeny the low optimum may reflect high
  testing cost, the total cost must be reduced by
  low BP size. Note that progeny is superior to
  phenotype at very high budget!
• For a mixed Phenotype/Progeny philosophy,
  similar low numbers as for progeny appeared.

• For Clone, low BP size probably reflects the high
  gain, which makes the gene diversity loss
  important

38
Heritability
0.80
30
h2 0.5
0.60
Annual Group Merit,
0.40
60
0.20
h2 0.1, main scenario
0.00
0
100
200
300
400
500
Breeding Population Size
High heritability boosts gain and makes loss of
gene diversity less important. Low heritability
justifies higher BP size, but the dependence is
not strong, and I suggest we can regard the
dependence as unimportant.
39
Annual Budget
100
Optimal BP Size
50
0
0
500
1000
1500
2000
Annual Budget (main scenario 500)
Optimal BP size increases with the budget, but
only marginally and unimportant. But if spruce
has a higher budget than pine, it is also an
argument for a larger BP.
40
Optimum share of resource invested in testing
increases with the budget
100
80
60
Testing cost as of budget
40
20
0
0
500
1000
1500
2000
Annual budget
41
Cost of gene diversity
200
150
main scenario 200
Optimal BP Size
100
50
0
0
100
200
300
400
500
600
700
800
900
1000
Diversity cost (inbreeding depression F 100)
Optimal BP is very dependent (close to linear) on
cost of gene diversity. It is critical and is
difficult to assign a value. Main scenario
approach is that the cost is double as high as if
all production were lost if no gene diversity
remains. This seems sufficiently conservative.
42
Genetic variation
Optimal BP size is very dependent on genetic
variation in value for forestry. It ought to be
possible to assign better estimates as trials
grow older. The main scenario approach is that
the CV of value for forestry is 10 within
family, an educated guess based on estimates from
younger trials is 12.5 (which may be adjusted to
10 for uncertanties).
43
Genetic variation
We are now exploring genetic variation in value
for forestry in old trials, if we find that to be
smaller than CV10, that may be reason to
increase breeding population size.
44
Less balance in spruce!?

• Economically more important for Sweden (??BP)
• More plants produced (??BP)
• Higher investment in breeding (??BP)
• Higher site index (??)
• More flexible, present populations may later be
  merged (??BP)
• Can be more efficiently bred by clonal testing
  (??BP)
• A lower BP may be defended. Instead I suggest to
  breed more aggressively, thus less balance in the
  selections.

45
Spruce breeding population

• We have decided that 50 is needed and should not
  be keen on reducing it only a decade later.
• However there is another reason to reduce spruce
  BP, that is that it is more flexible, thus zones
  can be larger.
• I suggest to manage spruce BP more unbalanced
  than pine.
• And be more prepared for a reduction some decades
  ahead by reducing the number of populations.

46
For pine it is more important to go on with
present BP

• Lower investment in breeding (? BP)
• Less flexible, more difficult to draw on adjacent
  zones (? BP)
• Can not be bred by clonal testing (? BP)
• Thus for long term breeding insufficient reason
  to decrease or increase BP and it may be
  desirable to keep breeding rather balanced.

47
End of the slidesShall we have a final
discussion?
Or someone may have tried Breeding Cycler and
experienced a problem?
48
End slide beer
Or just relax?