The breeds and breeding
system best suited to a given production-marketing ecosystem can be determined
most efficiently by a sequence of steps:
Define the production-marketing system or systems most likely to be economically feasible in the foreseeable future in the geographic region involved. If more than one management system is important, interaction with several management systems may need to be included in the experimental design.
Use any pre-existing information concerning performance of candidate breeds and breeding systems obtained under conditions most similar to those intended, to reduce number of breeds and systems to those worthy of further evaluation.
Choose an experimental design suited to the additional information desired and to the availability of breed samples:
Use breed-of-sire topcross design when only sires or semen is available and/or a large number of breeds are to be evaluated. When necessary, even the less efficient, indirect comparisons of breeds or crosses evaluated at different locations can be obtained as deviations from a common sire breed or cross, using semen or embryos to produce the common control.
Use a diallel design if adequate samples of both sexes are available and a sufficiently small number of breeds is involved.
If usefulness of new composite breeds vs recurrent crossbreeding is to be evaluated, include contemporary comparison of parental purebreds with F1, F2, backcross and F3 generations of crossing.
Sires or females sampled from the breeds compared obviously should be as broadly representative of the breed (i.e., unrelated) as possible, and in the form and at the performance level that would be available if the breed were chosen for further industry use.
Choose performance traits to be measured that will permit estimation of economic production efficiency for the alternative breeds and breeding systems evaluated, as discussed earlier under Performance Measures Required.
Pre-analyze the experiment to determine the most efficient data structure (e.g., numbers of sires, dams and progeny per dam) and total scale of experiment necessary to achieve the desired confidence limits for differences of economic importance (e.g., 5% ± 2% or less).
Factors influencing the efficiency of crossbreeding experiments have
been considered by Dickerson (1942, 1969, 1973); Comstock and Winters
(1942); Robertson (1959) and Solkner and James (1990, 1991). Choice of the
genetic groups essential to minimize error
in estimating the desired breed, heterosis and recombination parameters
is more important than the optimum distribution of observations among genetic
groups. Required numbers usually can be estimated from prior knowledge of
heritability and variability of traits to be measured. This is illustrated in Table 5 for a 5% mean
difference between any two breeds siring crosses from the same breed of dam, when the trait measured has a
coefficient of variation (SD/mean) of either 20 or 10% and heritability
of h2 = 10%. These examples for cattle or sheep assume only one (1)
progeny per dam, so that both the dam (D)
and the within dam (W) components of variance in the SE of mean difference
are reduced in proportion to the total numbers of progeny per breed of sire (nG). The sire component (S = 1/4
Vg) is reduced only by number of sires sampled per breed (Ns).
In this example, a difference of 5 ± 2% and P ~ .02 would be expected for
traits with CV = 20% and h2 = 10% when nG = 220 and Ns
= 22. If numbers of progeny per breed of sire are increased to nG =
280, only Ns = 9 would be
required for the same degree of reliability for the estimated breed difference. The desired numbers can be
reached by running the trial with different sire samples over several
years or locations in matings with the same breed of dam.
For traits
with the lower CV = 10% but same h2 = 10%, numbers required for a 5 ± 2% breed difference would
be only nG = 60, Ns = 6. For traits with higher h2,
required numbers would be still lower.
For pigs or poultry, the 
where Nd = number of dams per breed. A
reasonable goal for size of an experiment might be 5 ± 2% for estimated breed difference in the important
trait having the highest CV and lowest h2.
7. Analyze
results to estimate size of breed, heterosis and epistatic effects in
performance traits, and of differences in net production efficiency among alternative
breeds and systems of breed use.
Some of the
informative topcross mating designs are shown in Figure 1. The objective usually is to determine the potential
usefulness of several exotic breeds (B, C) for crossing with one or more
indigenous breeds (A). This requires estimates of individual and maternal
average (gI and gM) and heterotic (hI and hM)
as well as non-allelic gene recombination effects (rI and rM)
for crosses of exotic with indigenous breeds (Table 4). Results are useful for
at least preliminary choices among exotic breeds for possible 1) replacement of
the native breed or breeds, 2) crossing with the native breed or breeds, or 3)
development of new composite
breeds. Information about heterosis in crosses among the exotic breeds would require extension of this
design to include three-breed crosses or crosses of each exotic with the
backcrosses to other exotic breeds (e.g., B×CA or B×
C(CA)), but can be done much more efficiently with diallel crossing including
both males and females of the breeds involved (Table 2). Deviations of F1
crosses from the native pure breed (A) include both average and heterosis
effects - i.e., 
(Table 4). The difference in average transmitted effects of
B from A can be estimated directly from the reciprocal backcrosses:
Thus, 
The linearity of
increases in additive gene effects with "percentage of blood" can be evaluated
by comparing:
Only the
effect increases
linearly with fraction of B in pedigree; all other expected genetic fractions
are unchanged except 
, which is 1/4 for the F2 (BA×BA) but only 1/8 for each
backcross. Thus, BA epistatic recombination effects also can be estimated by
comparing the F2 with the mean of the two backcrosses;
Linearity of average gene effect difference between the exotic breeds B
and C can be estimated similarly from the differences
among their paired backcrosses and the F2 inter se matings:
Any
differences between and 
can be estimated by
comparing (BA)2-(CA)2 with the mean for the two backcross differences = 
because all other
elements cancel (Table 4).
Traits of reproducing females
can be evaluated in a parallel manner, using each sire of a common unrelated
breed in matings with females of all the exotic x native F1, crosses (e.g., D×A,
D×BA and D×CA). Differences
in female performance which include progeny-performance (e.g., progeny
output) will contain offspring average and heterotic gene effects e.g., 
and 
, that are confounded with those for the F1
female's maternal effects (e.g., with 
and 
However, separate
estimates of 
are obtainable from the F2 and
backcross progeny contrasts described above.
Linearity of increase in
maternal breed effects from increasing the fraction of exotic genes also can be estimated from crosses of the F2
and the reciprocal backcrosses with sires of a common unrelated breed, e.g., D×A(BA), D× (AB)2 and D×B(BA). Increases in 
or in 
with change from 1/4 to 3/4
B or C genes will correspond to those for or and 
or 
shown above for the F2 and backcross progeny. The parallel effects
on and of progeny from the
matings with F2 and backcross females will be exactly one-half of
those for and ; but will also include proportional 1/4 to 3/4
increases in 
and 
proportion of total hI heterosis. Again, importance of non-allelic gene interaction effects (e.g., 
or ) can be estimated by comparing
means of D crosses with the F2 vs those with the two
backcrosses of each exotic breed.
Compared
with topcrossing exotic breeds on a common indigenous breed population, a
diallel mating design permits estimation of heterosis among all n(n-l)/2 pairs
of breeds instead of only n exotics with the base breed. However it requires
representative samples of both males and females of each breed. Thus, it is
useful mainly for evaluation of breeds already
indigenous to a region or for a limited number of breeds chosen on the basis of
prior topcross evaluation.
As shown in Table 1, diallel matings involve reciprocal crosses between
each pair of breeds plus the contemporary pure breed matings. This first phase
permits estimation of breed individual (gI)
and maternal (gM) effects as well as heterosis for individual
progeny performance (hI), as deviations from the unweighted
mean (Pn) of the n pure breeds evaluated.
Each pure breed mean includes the general purebred mean (Pn)
plus that breed's genetic deviations for individual 
and maternal 
effects, where 
Individual (hI)
heterosis can be estimated for each reciprocal cross (hij) = 1/2(Xij + Xji -
Pi - Pj) and for all 2(n-l)
crosses involving sires or dams of a given breed:
Mean heterosis for all crosses, of course, is
simply
Effects of any new
non-allelic gene interactions in the first crosses are included in the
estimates of F1 heterosis (hI).
Average maternal
effect of each breed 
is estimated as the
average difference between the dam and the sire effect of reciprocal crosses:
because 
so that
To obtain estimates of heterosis for maternal effects on progeny
performance (hM) requires Phase 2 comparisons of females of each
reciprocal cross and of the two pure breeds both
mated to the same breed of sire (Table 2), and preferably to the same sires
(e.g., 
. Such contrasts for each set of female reciprocal crosses
provide an estimate of 
where 
represents possible
additional non-allelic gene recombination effects in progeny that are not
included in definition of 
. If the effects should be real
and negative, they would cause an underestimate of 
, and vice versa.
The reality
of effects can be
estimated by extending the matings to include comparisons of the F2
generation of each F1, cross with the mean of the two reciprocal
backcrosses, as shown in Table 4 for the topcross designs.
Evaluation of rM requires comparison of F2
females with mean of reciprocal backcross females all mated to the same sire
breed, e.g., D(BA)2–[D(A●BA)+D(B●BA)]/2, as shown in Table 4. The value of such phase 3
matings will depend upon accumulated evidence concerning
the importance of such epistatic deviations from only average plus dominant
gene effects for each species of animals and the traits of interest.
Usefulness
of a new composite breed can be determined most directly by comparing the F3
progeny (from F2 sires and dams) with the weighted means of the
purebreds and of the F1 crosses represented in the composite (Table 3). However, wise choices between systematic crossbreeding vs new
composite breed formation, as well as breed composition of the composite,
require the prior information about individual and maternal breed (gI and gM), crossbred heterosis (hI and hM)
and epistatic recombination (rI and rM) effects on production
efficiency. Use of such information, plus the reproductive rate of the species,
in production system evaluation should clarify
possible justification for forming new general purpose or specialized
maternal and paternal composite breeds. Optimum proportional representation of
breeds in a composite can be predicted from the estimates of breed and
heterosis effects on component traits as illustrated by MacNeil
(1987).
A variety of
less complete comparisons of breeds or crosses are also useful. These include growth, carcass and feed utilization tests
of market meat animals; growth and feed utilization of breeding males;
egg production, quality and feed efficiency of laying hens; meat, wool or fiber
production, quality and feed efficiency of sheep or goats, etc. In such tests,
the entries are samples of specific breeds or strain crosses. The information
is helpful to potential users of the breeding or commercial stocks compared. It
is also used by breeders to compare their stocks with those of other breeders.
In either case, usefulness of the comparisons depends on representative
sampling of each stock and the completeness and accuracy of performance
information obtained. Entry of selected-samples by breeders can bias results obtained. Differences in health
background of entries also can be a problem. Information from such comparative tests can be useful in selecting
breeds or crosses for more complete evaluation experiments.
In several livestock species (i.e., dairy and beef cattle, sheep and
swine), the genetic analyses of field (on farm) records also can provide excellent
preliminary information on breed characteristics.
