J. Plant Breed. Genet. 01 (03) 2013.

117-129

Available Online at ESci Journals

Journal of Plant Breeding and Genetics

ISSN: 2305-297X (Online), 2308-121X (Print)
http://www.escijournals.net/JPBG

MATING DESIGNS: HELPFUL TOOL FOR QUANTITATIVE PLANT BREEDING ANALYSIS

a,bAthanase Nduwumuremyi*, aPangirayi Tongoona, cSlyvestre Habimana
a African Centre for Crop Improvement (ACCI), University of KwaZulu-Natal, Scottville, Pietermaritzburg, South Africa.
b Department of Natural Resource Management, Rwanda Agriculture Board (RAB), Kigali, Rwanda.
c Faculty of Agriculture and rural development, Higher Institute of Agriculture and Animal Husbandry, Musanze, Rwanda.

ABSTRACT

Selection of parental materials and good mating designs in conventional plant breeding are the keys to the successful
plant breeding programme. However, there are several factors affecting the choices of mating designs. Mating design
refers to the procedure of producing the progenies, in plant breeding, plant breeders and geneticists, theoretically and
practically, they use different form of mating designs and arrangements for targeted purpose. The choice of a mating
design for estimating genetic variances should be dictated by the objectives of the study, time, space, cost and other
biological limitations. In all mating designs, the individuals are taken randomly and crossed to produce progenies
which are related to each other as half-sibs or full-sibs. A form of multivariate analysis or the analysis of variance can
be adopted to estimate the components of variances. Therefore, this review aimed at highlighting the most used mating
design in plant breeding and genetics studies. It provides easy and quick insight of the different form of mating designs
and some statistical components for successful plant breeding.
Keywords: mating designs, plant breeding, genetics statistics.

INTRODUCTION the investigation of polyploidy species provided they

In plant breeding, various mating designs and behave as functional diploids, with disomic inheritance.
arrangements are used by breeders and geneticists to (b) Uncorrelated genes distribution. The genes
generate improved plants. The selection of suitable controlling the character should be independently
parents and good mating designs are keys to the distributed among the parents. (c) Absence of non-allelic
successful plant breeding schemes (Khan et al., 2009). interactions. In the triple test and diallel crosses
However, there are several factors affecting the choices epistasis can be detected and its effects including in
of mating designs. The factors influencing the choice of prediction. (d) No multiple alleles at those loci
mating design are (i) the type of pollination (self- or controlling the character. (e) Absence of reciprocal
cross-pollinated); (ii) the type of crossing to be used differences. Again this assumption can be tested in
(artificial or natural); (iii) the type of pollen several designs and appropriate measure taken. (f)
dissemination (wind or insect); (iv) the presence of a Ideally the diallel cross should be restricted to crosses
male-sterility system; (v) the purpose of the project (for among homozygous lines. Heterozygous can be catered
breeding or genetic studies); and (vi) the size of the for, but it complicated the interpretation of the results.
population required (Acquaah, 2012). (g) Absence of genotype-environment interaction. Their
Before discussing the mating designs, it is very presence merely emphasizes the need for wide scale
important to understand the genetic assumptions (Hill et testing of material in order to determine the extent of
al., 1998) : (a) Diploid behaviour at meiosis; this such interaction.
assumption applies to all designs, but it doesn’t preclude The mating designs have four main importance, (1) to
* Corresponding Author: provide information on the genetic control of the
Email ID: nduwatha@gmail.com character under investigation; (2) to generate a breeding
© 2013 ESci Journals Publishing. All rights reserved. population to be used as a basis for the selection and

117
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

development of potential varieties; (3) to provide different form of mating designs and some statistical
estimates of genetic gain and (4) to provide information components for successful plant breeding; these will
for evaluating the parents used in the breeding program serve as reference to the scientists and students in the
(Acquaah, 2012). In making various crosses breeders domain of plant breeding and genetics.
have interests in discovering the answer to the following MAJOR MATING DESIGNS IN PLANT BREEDING AND
questions: how significant is genetic variation? How GENETICS
much of the variation is heritable? And what types of Mating design refers to the procedure of producing the
gene affecting that significance? However, these are progenies, in plant breeding, plant breeders and
answered by comparing the variances of the segregating geneticists, theoretically and practically, they use
and the non-segregating generations (Kearsey and different form of mating designs and arrangements for
Pooni, 1996). Another interest of the breeder is targeted purpose. However, the choice of a mating
identifying plants with superior genotypes as judged by design for estimating genetic variances should be
the performance of their progeny. Suitable inbreds or dictated by the objectives of the study, time, space, cost
lines are selected based on combining ability effects with and other biological limitations. Thus, several studies
better mean performance. Combining ability depends on (Griffing, 1956b; Kearsey and Pooni, 1996; Hallauer et
the gene action controlling the trait to be improved. al., 2010; Acquaah, 2012) described and contrasted
General combining ability (GCA) is the average different mating designs and six types of mating designs
performance of a line in hybrid combinations and is due have been described so far: (1) bi-parental progenies
to additive genes action. The estimation of GCA for a (BIP), polycross, topcross, North Carolina (I, III, III),
particular line depends upon the mating design, but Diallels (I, II, III, IV) and Line X tester design. In all
essentially, it is the deviation of its progeny mean from mating designs, the individuals are taken randomly and
the mean of all lines included in the trial (Acquaah, crossed to produce progenies which are related to each
2012). Thus, theoretically, differences between maternal other as half-sibs or full-sibs. A form of multivariate
groups measure variation in their general combining analysis or the analysis of variance can be adopted to
ability. Specific combining ability (SCA) refers to estimate the components of variances.
combinations or crosses that do relatively better or BI-PARENTAL MATING
worse than would be expected based on the average The bi-parental design is also called paired crossing
performance of the lines involved, it is therefore due to design and it is reported to be the simplest mating design
non-additive gene action (Falconer and Mackay, 1996, (Mather, 1982). In this design, the breeder selects a large
Acquaah, 2012). The information regarding the number of plants (n) at random and cross them in pairs to
estimates of combining ability and genes actions is vital produce 1/2n full-sib families (Acquaah, 2012). Their
for a successful plant breeding (Panhwar et al., 2008). progeny are tested and the observed variation partitioned
Plant breeding experiments use two types of design, (1) by straightforward analysis of variance into between and
mating and (2) experimental design which should march within families (Hill et al., 1998). Statistically, if r plants
with its statistical components analysis and per progenies family are evaluated, the variation within
interpretation. Therefore, this review provides the (w) and between (b) families may be analyzed as follows:
Table 1. Analysis of variance for BIP design.
Source of variation df MS EMS
Between families ( n)-1 MS1  2 w  r 2b
Within families n(r-1) MS2  2w
Total nr-1
Source: Acquaah, 2012
Where: n and r refer to the number of parents and plants sampled within each cross respectively;
 2b is the covariance of full-sibs (  b = CovFS = 1 VA  1 VD  VEC  1 (MS1  MS2 ) and
2

 
2 4 r
 2 w =  G  CovFS   EW = 1 VA  3 VD  VEW  MS2 ; is the environmental source of variation for variance within the
2 2
2 4
crosses. When you assume that dominance effects are zero, then  b = 1 VA and  w = 1 VA  VEW (Acquaah, 2012).
2 2
2 2

118
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

The simplicity of this design is counterbalanced by its genetic and environmental variance.
inability to yield sufficient information to estimate all POLYCROSS
parameters required by the model (Acquaah, 2012). This design is for intermating a group of cultivars by
Only two statistics are available for estimating VA, VD, natural crossing in isolated block. Term polycross was
VEW and VEC. This is because the progeny from this coined by Tysdal, Kiesselbach and Westover in 1942, to
design are either full sibs or unrelated; no other indicate progeny from seed of a line that was subject to
relationship exists among them. The estimates of the outcrossing with other selected lines growing in the
parameters can only be obtained either by simplifying same nursery (Hill et al., 1998). It is most suited to
assumptions, or if extra statistics become available (Hill species that are obligate cross-pollinaters (e.g., forage
et al., 1998). If dominance is assumed to be absent grasses and legumes, sugarcane, sweet potato), but
(VD=0), and there is no common environment (VEW=0), especially to those that can be vegetatively propagated
that is individuals from the same family do not share the crops such as sugarcane, cassava and sweet potato
same environment. Consequently, if these assumptions (Acquaah, 2012). The design provides equal opportunity
are unjustified, it will lead to an overestimate of the for each and every clone or parent to naturally cross
genetic component relative to the environmental with each other in the block such that self-pollination is
component. These difficulties can be circumvented to a prevented (Saladaga, 1989). However, to achieve this
limited extent in practice. Family plots can be dispensed objective, a proper design in the polycross block is
by randomizing individual plants over the whole critical. It provides an equal opportunity for each entry
experiment. In this way VEC becomes zero. For to be crossed with every other entry. It is critical that the
biometrical geneticist individual plant randomization is entries be equally represented and randomly arranged
a useful device for increasing the precision for estimates, in the crossing block (Falconer and Mackay, 1996). The
but for a plant breeder it may be an unaffordable luxury. Latin square experimental design was suggested to be
Nevertheless, for shrubs and trees, single tree plot used as the most appropriate design to ensure all entries
designs are frequently used. Usually, however, it may be have equal chance of random intermating with each
simpler for the breeder to estimate VEC directly from the other in the polycross nursery (Morgan, 1988).
families X replicates interaction mean square in properly Nevertheless, when the entries number is more than 10,
replicated experiment. the completely randomized block design may be used
In BIP design, extra statistic can be generated if (Acquaah, 2012). In both cases, about 20–30 replications
information on the parents is available, or by inclusion of are included in the crossing block. The ideal
the selfed progenies of parents (Kearsey, 1970). requirements are hard to meet in practice because of
However, in practice both options are of limited value; several problems, placing the system in jeopardy of
because of the problems posed by presence of deviating from random mating. If all the entries do not
genotypes-environment interaction when considering flower together, mating will not be random. To avoid
parent-offspring correlation, and introduction of this, the breeder may plant late flowering entries earlier.
additional parameters. Pollen may not be dispersed randomly, resulting in
The BIP design, though simple to execute, has obvious concentrations of common pollen in the crossing block.
limitations. Because no constraints have been imposed Half sibs are generated in a polycross because progeny
upon the mating there is a lack of relatedness among the from each entry has a common parent. The design is
resultant progeny. Consequently, unjustifiable used in breeding to produce synthetic cultivars,
assumptions many be required if estimated of the most recombining selected entries of families in recurrent
important genetic and environmental components are to selection breeding programs, or for evaluating the GCA
be obtained (Hill et al., 1998). The most limitation of this of entries (Acquaah, 2012).
design is its inability to provide the needed information The polycross design has an advantage of producing
to estimate all the parameters required by the model. synthetic cultivars, recombining selected genotypes in
The progeny from the design comprise full sibs or the recurrent selection procedure and evaluating the
unrelated individuals. There is no further relatedness general combing ability of the parent genotypes
among individuals in the progeny. The breeder must (Falconer and Mackay, 1996; Sleper and Poehlman,
make unjustifiable assumptions in order to estimate the 2006; Acquaah, 2012). The general combining ability

119
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

(GCA) can also be estimated from polycross mating performance of the parental lines and their progenies
design. such as flowering is likely to be affected (Morgan, 1988).
The general combining abilities estimated are basically In addition, the differences in performance of progeny
for maternal parents and the variations measured in a clones could arise from variations in heritability of trait
progeny can be partitioned into within and between measured (Gorz and Haskins, 1971). Consequently these
maternal parents (Falconer and Mackay, 1996) and could lead to inaccurate estimates of GCA; hence
consequently, general combining ability helps in heritability determined needs to be treated with caution.
estimating heritability. The mean performance of the In polycross, the progenies from individual plants are
progenies of any female parent in the polycross is used tested. These progenies are half-sib families. The
to determine the variance components and consequently covariance within the families is Cov( HS )  1  F  2 A
the general combining ability (GCA). The heritability 4
calculated provides decision guidance for usefulness of where the inbreeding coefficient of the genotypes
polycross in breeding programme (Saladaga, 1989). being tested. The following ANOVA table shows the
However, since the parents are of different origin and analysis of many replications for polycross design.
the crop is sensitive to environmental changes, the
Table 1. ANOVA table of polycross design with many replicated.
Source df MS EMS Variance components

Progenies ( )

Blocks - -

error ( )( )  2e   2
Source: Wricke and Weber, 1986
The variance component is an estimate of ; TOP CROSS DESIGN
when the parents are non-inbred, = zero. A Topcross refers to a mating between a selection, line,
comparison of the coefficients with the corresponding clone and a common pollen parent which may be a
coefficients in case of parent-offspring covariance variety, inbred line or single cross. The selected plants
indicates that the precision of the estimate of is lower are crossed with a common tester(s) of known
for the topcross or polycross than for the covariance performance, generally in open pollination. The design
between parents and offspring. The precision is was proposed by Jenkins and Brunsen in 1932 for
increased if the tested genotypes are inbred. It is testing inbred lines of maize in cross-bred combinations
convenient to use polycross design in cross-pollinated and later renamed topcross by Tysdal and Grandall in
species when evaluating a large number of genotypes 1948 (Hill et al., 1998). The tester parent should have
(Wricke and Weber, 1986). The selection is then applied well known genetic background; either narrow- or
based on half-sib progeny means. However, polycross broad-based testers (Aly et al., 2011). The purpose of
design has a number of limitation such as random using top is to increase the chance of obtaining a
mating, insufficient statistics to estimate all the desirable gene or genes from exotic or difficult materials.
parameters, the component of variance are only Exotic refers to lines from other countries which are
estimated from the maternal half sibs; information about generally poorly adapted to local conditions. Difficult
the males is lost, no control over the pollen source; material refers to varieties or lines which are tall, poor
expected genetic gains are reduced by half, the non- combiners, or dominant susceptibles, etc. i.e. lines which
randomness of mating (due to lack of synchronisation of have given poor results (progeny) from single crosses in
flowering, unequal pollen production and position previous crossing cycles. In making top crosses, only
effects in the crossing block). The polycross is ideally single cross F1's are utilized because they are uniform.
suited for identifying mother plants with superior The top cross F1's will be segregating and it is
genotypes from the performance of their progeny impossible to identify superior plants at crossing;
general combining ability (Hill et al., 1998). therefore, they are not used. The F1's are selected for

120
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

desirable agronomic characteristics or for desirable yield trials. This design is probably the simplest model of
parentage. mating design that can provide preliminary rapid
Topcross has been fairly widely used for preliminary screening of genetic stocks as it involves the lowest
evaluation of combining ability of new inbred lines crossing load and simple statistical analysis (Mosa,
(Mosa, 2010). The possible numbers of crosses are n x 1, 2010). In topcross, the progenies from individual plants
given n number of inbreds. Topcross progenies yield are tested; these progenies are half-sib families. The
only GCA information, not SCA. It is also called inbred- covariance within the families is ( )
variety cross (Sleper and Poehlman, 2006). It is a simple
where the inbreeding coefficient of the genotypes is
and efficient system of screening inbred lines for
tested. Table 3 presents the general ANOVA for topcross.
combining ability before pairing inbreds in single-cross
Table 3. Skeleton of ANOVA for half sib family test by topcross.
Source df MS EMS Variance components

Progenies
( )
Blocks -

Error ( )( )  2e   2
Source: Wrickle and Weber, 1986.
The variance component  2
prog is an estimate of criterion. As a nested design (Figure 1), each member of
a group of parents used as males is mated to a different
from   V (m1 )  V (m2 ) ,
2
, calculated prog group of parents. NC Design I is a hierarchical design
when the parents are non-inbred, = zero (Wrickle and with non-common parents nested in common parents
Weber, 1986). However, top crosses require 5 heads per (Acquaah, 2012).
cross; this number is necessary because these crosses
will segregate in the next F1 generation and at least 80
plants to facilitate the selection of desirable plants in the
F1. The design has two shortfalls. First, a single tester
variety may not offer wide genetic background for
testing the inbred stocks. Secondly, the numbers of
crosses become large if the test inbreds are many.
NORTH CAROLINA
North Carolina design was developed after using long
time diallel. However, the later require much labour.
Therefore, in order to obtain more information about
combining ability but without much labour comparing to
full diallel, Comstock and Robinson in 1952, introduced
the North Carolina designs I, II, and III.
North Carolina Design I: It is a very popular
Figure 1. North Carolina Design I (Source: Acquaah,
multipurpose design for both theoretical and practical
2012).
plant breeding applications (Acquaah, 2012). It is
commonly used to estimate additive and dominance The progenies include both full-sibs and half-sibs. Each
variances as well as for the evaluation of full- and half- set of families with the same father in common
sib recurrent selection. It requires sufficient seed for constitutes a half sib family group and a set of families
replicated evaluation trials, and hence is not of practical with both parents in common constitutes a full-sib
application in breeding species that are not capable of family (Kearsey and Pooni, 1996; Hallauer et al., 2010).
producing large amounts of seed. It is applicable to both The model for experiment conducted in one
self- and cross-pollinated species that meet this environment is: Where

121
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

 is the mean, mi is the effect of the ith male, f ij is the (Hallauer et al., 2010). The design is commonly used to
estimate additive and dominance variances (Acquaah,
effect of the jth female mated to the ith male, rk is the 2012).
replication effects, and eijk is the experimental error
Table 2. Skeleton of general ANOVA for North Carolina design I.
Source of variation df MS Expected mean squares
Males ( ) M1  2 w  r 2 mf  rf 2 m
Females ( ) M2  2 w  r 2 mf
Within plots ( ) M1  2w
Total
Source: Acquaah, 2012.
The parameter  w refers to the average variance
2 genetic variance. Also, NCI has been used successfully to
within the full sib families and is given as; tree breeding where mass collection of pollen from
common parents possess no practical problems (Hill et
 2 w  M1  12VA  3 4 VD  VE
al., 1998). Further, the design is applicable to both self-
 2 m  (M1  M 2 ) / rf  1 4VA and cross-pollinated crops. However, this design is most
widely used in animal studies. In plants, it has been
r 2 mf  (M 2  M 3 ) / r  1 VA  1 VD
4 4 extensively used in maize breeding for estimating
The translation of components of variance to covariance genetic variances (Acquaah, 2012).
of relatives permits estimation of components of genetic North Carolina Design II : In this design, each member
variances (  A and  D ) and their interaction with the of a group of parents used as males is mated to each
2 2

environments. Assuming no epistasis, the estimation of member of another group of parents used as females
Design II is a factorial mating scheme (Figure 2). It is
additive genetic variance ( 
2
A) and total genetic
used to evaluate inbred lines for combining ability.
variance (  G ) are obtained from the mean squares of
2
The design is most adapted to plants that have multiple
the analysis of variance (Hallauer et al., 2010). flowers so that each plant can be used repeatedly as both
Dominance variance  D is obtained from the
2
male and female.
difference between the female - within males and males Blocking is used in this design to allow all mating
components of variance. The NCI has advantage over involving a single group of males to a single group of
biparental and polycross designs, because it gives three females to be kept intact as a unit (Acquaah, 2012). The
statistics compared with only two in the polycross and design is essentially a two-way ANOVA in which the
biparental (Kearsey, 1965). variation may be partitioned into difference between
Like the polycross, the main advantage of design I is its males (m) and females (f) and their interaction.
ability to supply a test of significance for the additive

Figure 2: NC II Design (factorial design with paired rows) Source: Acquaah, 2012.

122
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

Table 5. Skeleton of general ANOVA of NC II.

Source of variation df MS Expected mean squares
Replications r 1
Males m 1 M1  2 w  r 2 mf  rf 2 m
Females f 1 M2  2 w  r 2 mf  rm 2 f
Males x females (m  1)( f  1) M3  2 w  r 2 mf
Within progenies mf (r  1) M4  2w
Error (r  1)(mf  1) M5 2
Total rmf  1
Source: Kearsey and Pooni, 1996
Where:  w is the within progenies genetic and
2 This design also allows the breeder to measure not only
environmental variances. In the absence of epistasis and GCA but also SCA (Acquaah, 2012). However, the NCII is
not providing test of epistasis or G X E interaction
common environmental effects,  2 mf is a function of
(Kearsey and Pooni, 1996). In North Carolina II, every
dominance variance VD only (Kearsey and Pooni, 1996). progeny family has half sib relationships through both
If there is environmental variation between FS families, common male and common female. This is accomplished
this could be due to general and specific maternal by systematic crossing programme in which n1 male and
effects. The general maternal effects (VEM ) will appear n2 female are mated in all possible combinations to give
n1n2 progeny families. It is therefore a rectangular
in  2 f  1 V  V , while the specific maternal effects mating design, unless n1=n2. Reciprocal crosses may be
A EM
4
carried out to analyze maternal effects (Hill, et al., 1998).
VEC  VEM will appear in  2 mf  1 VD  VEC  VEM  and will North Carolina Design III: In this design, a random
4
sample of F2 plants is backcrossed to the two inbred
be confounded with VD (Kearsey and Pooni, 1996). If
lines from which the F2 was descended. It is considered
the number of males and females is the same, the most powerful of all the three NC designs.
, we can have a test of maternal effects by
comparing as a variance ratio.

Figure 3. NC III design with conceptual, practical and modified. Source: Acquaah, 2012
However, it was made more powerful by the capable of testing non-allelic (epistatic) interactions,
modifications made by Kearsey and Jinks that add a which the other designs cannot, and also capable of
third tester not just the two inbreds (Acquaah, 2012). estimating additive and dominance variance (Acquaah,
The two parental lines act as testers against which F2 are 2012). According to Hill et al, 1998, it is also called triple
assessed. The parents being progenitors of the F2, are test cross because of inclusion of the third tester. This
very special testers because F2 is segregating at all loci inclusion increase the power of this design considerably,
for which the testers differ but for no other loci (Kearsey because it provides a sensitive and unambiguous test for
and Pooni, 1996). The F2 population is reference non-allelic interactions, a capability which none of the
population for mating NCIII (Hallauer et al., 2010). designs described so far, not even design 3 in its original
The modification is called the triple testcross and is form possess. Moreover, both in its original and

123
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

extended form design 3 has a general utility for second assesses the significance, and provides estimates
investigating any population, irrespective of gene of the additive and dominance components of variation.
frequency or mating system (Hill et al., 1998). In triple The NCIII is a special case of NCII, therefore the ANOVA
test cross a random sample of n individuals from the is similar to that of the NCII although it differs in one
population under investigation is crossed to the same special feature; the two testers are not a random sample
three testers, L1, L2 and L3 to give 3n progeny families. from any population but are two very particular lines,
The analysis of this design may be divided into two the progenitor of the F2.
parts, the first part supplies a test for epistasis, and the
Table 7. Skeleton of NC III ANOVA.
Source of variation df MS Expected mean squares
Testers, p 1 M4  2  r 2 mp  rmK 2 p
Males (F2), m m 1 M3  2  2r 2 m
Testers x parents m 1 M2  2  r 2 mp
Within FS families/error r 12m 1 M1 2
Total 2mr  1
Source: Hallauer et al., 2010
Where; r and m are number of replications and male triple test cross is without doubt the most powerful of
plants, respectively. Although this design had many the available mating designs. But for the plant breeder
advantages, there are some pitfalls which should be this design may be off-piste (Hill et al., 1998).
avoided if its full potential is to be realized. If interest DIALLEL DESIGN
centres on determining whether dominance is present in A complete diallel mating design is one that allows the
particular population, virtually any pair of inbred lines parents to be crossed in all possible combinations
will suffice testers. But if we wish to estimate the total (Schlegel, 2010), including selfs and reciprocals. This is
dominance variation, L1 and L2 should differ at all those the kind of mating scheme required to achieve Hardy–
loci which are segregating in the population. The same Weinberg equilibrium in a population (Acquaah, 2012).
criterion also applies if an unbiased estimate of the The diallel is the most used and abused of all mating
additive component is required. Generally speaking designs in obtaining various genetic information
therefore, L1 and L2 should be high and low selections (Hallauer et al., 2010). Much of its abuse could probably
from the population, when additive and dominance be due to the presence of two models for diallel analysis;
components will be estimated with equal precision. But random and fixed models (Griffing, 1956b). A random
the tester must be derived from the population under model involves parents that are random members of a
investigation as the model developed for this mating random mating population. A random model is useful for
design is only valid within a population. Because L1 and estimating GCA and SCA variances. In contrast, when
L2 represent extreme selections, estimates of narrow parents are considered fixed effects, the aim is to
sense heritability can be obtained only after selection measure the GCA effect for each parent and the SCA
had taken place (Hill et al., 1998). It is reported that the effect for each pair of parents. These effects only apply to
main use of the triple test cross will be to investigate the the set of parents in the diallel. It is also widely used for
inheritance of quantitative traits in natural population developing breeding populations for recurrent selection
and relating this to natural selection and ecology (Acquaah, 2012). In addition, Johnson and King (1998),
(Kearsely and Jinks, 1968).In practice the triple test reported that diallel mating designs are deployed to
cross is viewed as a resource consuming design. provide the maximum opportunity to manage co-
Moreover, the generations required to establish this ancestry in breeding population and maximize selection
design are not those which a breeder would routinely differential. However, in practice, a diallel with selfs and
produce in a breeding programme. For the biometrical reciprocals is neither practical nor useful for several
geneticist, interest in determining the genetic reasons. Selfing does not contribute to the
architecture of a particular character and in estimating recombination of genes between parents. Furthermore,
the genetic component with maximum precision, the recombination is achieved by crossing in one direction

124
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

making reciprocals unnecessary (Acquaah, 2012). 1

Because of the extensive mating patterns, the number of
Yij    g i  g j  sij  rij    ijkl Where,
bc k l
parents that can be mated this way is limited.
 is the population mean, gi , g j is the general
Nursery arrangements for the application vary
depending either complete or partial diallel design and combining ability effect for the ith and jth parents, sij
four methods under the diallel mating design have been is the specific combining ability effect of the cross
so far described. The number of progenies generated
between the ith and jth parents such that sij = s ji rij is
from each method are different, the number of progeny ,
families (pf) for methods 1 through 4 are: pf = n2, pf = the reciprocal effect involving the reciprocal crosses
1/2n (n + 1), pf = n(n − 1), and pf = 1/2n(n − 1), between the ith and jth parents such that rij = rji and,
respectively (Acquaah, 2012).
Method I or full diallel design: The method I or full eijkl is the experimental error due to environmental
diallel design consisted by parents, one set of F1’s and effect associated with the ijklth , which is assumed to be
reciprocal F1’s. The system gives n2 genotypes (Griffing, uncorrelated and normally distributed with zero mean
1956b). The mathematical models for combining ability
and variance, VE
analysis for the fixed and random effects are given by;
Fixed effect model or model I: Random effect model or Model II:
1 1
Yij    g i  g j  sij  rij  
bc k
(bv)  
bc k

l
ijkl
ijk

The table 8 presents the estimates of variances for variance components both fixed and random model.
Table 8. Skeleton of ANOVA for method I diallel design.
Expected mean squares
Source df SS MS
Model I Model II
GCA p 1 Sg Mg 1 2( p  1) 2
 2  2 p( )  g 2i
2   g  2 p 2 g
p 1
p
SCA p p 1 / 2 Ss Ms 2 2( p 2  p  1) 2
2    Si j
2

p( p  1) 2  s
p2
Reciprocal eff. p p 1 / 2 Sr 2
)  ri j
Mr
 2  2(
p( p  1) i j
2
 2  2 2 r
Error m Se Me 2
Source: Griffing (1956b)
The various effects can be estimated as follows: and Var(rˆ )  1 ˆ 2
ij
1 2
ˆ  X .., gˆ i  1  Xi.  X .i   12 X .., Method II: This method includes parents and one set of
p2 2p p
F1’s without reciprocals F1’s. This design gives p(p+1)/2
sˆij 
1
2
xij  x ji   21p X i.  X .i  X j.  X . j  p12 X .., genotypes. The mathematical models for combining
ability for fixed model is:
rˆij 
1
xij  x ji  (Griffing, 1956b). It is important to note 1 i, j  1,..., p,
2 Yij    gi  g j  sij    ijkl
bc k l

k  1,...,b,
these restrictions;  gi  0 and
i
sj
ij  sii  0 for each i . l  1,...,c,


However, the variances of the effects can be estimated Where,  is the population mean, gi , g j is the
using the following equations: Varˆ   1 ˆ 2 , general combining ability effect for the ith and jth parents,
p2
sij is the specific combining ability effect of the cross
p 1 2
ˆ , Var(sˆij )  2  p 2  2 p  2ˆ 2 , i  j ,
1
Var( gˆ i )  between the ith and jth parents such that sij = s ji and
2p2 2p

125
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

eijkl is the experimental error due to environmental The mathematical equation for analysis of combining
ability for random model is:
effect associated with the ijklth . It is important to
1 1 1
remember that
 gi  0 and  sij  sii  0 . Yij    gi  g j  sij  
b k
bk   (bv)ijk    ijkl ,
b k bc k l
i j

Table 9. The analysis of variance for method II.

Expected mean squares
Source df SS MS
Model I Model II
GCA p 1 Sg Mg 1    s  ( p  2) g
2 2 2
 2  (2  p)( )  gi
2

p 1
SCA p p 1 / 2 Ss 2  2 s
 j si j 2
Ms 2
 
2

p( p  1) i
Error m Se Me' 2
Source: Griffing (1956b)
The various effects for method II can be estimated as i, j  1,..., p,
follows, 1 
2
Yij    gi  g j  sij  rij   eijkl , k  1,...;b,
ˆ  X .., bc k l 
p( p  1) l  1,...,c,
gˆ i 
1  2 
X i.  xii  X ..,  is the population mean, gi , g j is the general
p  2  p 
combining ability effect for the ith and jth parents, sij
sˆij  xij 
1
p2

X i.  xii  X j.  x jj 
2

( p  1)( p  2)
X .. . is the specific combining ability effect of the cross
However, the variances of the effects can be estimated between the ith and jth parents such that sij = s ji and
using the following equations: Var( ˆ )  2
ˆ 2 , rij is the reciprocal genotypic effects such that rij =
p( p  1)
p 1 rji and, eijkl is the experimental error due to
Var( gˆ i )  ˆ 2 ,
p( p  2) environmental effect associated with the ijklth
p( p  1) observation (Griffing, 1956b). However, the following
Var( sii )  ˆ 2 and Var(sij )  p2  p  2 ˆ 2 equations help in estimating effects:
( p  1)( p  2) ( p  1)( p  2)
1 p 1
Method III: In this method, one set of F1’s and the Var( ˆ )  ˆ 2 , Var( gˆ i )  ˆ 2 ,
p( p  1) 2 p( p  2)
reciprocals are included. This mating design gives rise to
a  p( p  1) different number of genotypes. As for p  3 2 (i  j ) ,
Var(rˆij )  ˆ 2 (i  j ) .
1
Var( sˆij )  ˆ
methods I and II, also it has both fixed and random effect 2( p  1) 2
models. Random model:
Fixed model:
1 1 1
Yij    gi  g j  sij  rij  
b k
bk   (bv)ijk  
b k bc k

l
ijkl
,

Table 10. Skeleton of ANOVA of Diallel method III.

Expected mean squares
Source df SS MS
Model I Model II
GCA p 1 Sg Mg  2  2 p( p  2)(
1
)  g 2i  2  2 s  2( p  2) g
2 2
p 1
SCA p p  3 / 2 2  2  2 s
2
Ss Ms 2   si j 2
p( p  3) i j
Reciprocal eff. p p 1 / 2 Sr Mr  2  2(
2
)  ri j
2
 2  2 r
2

p( p  1) i j
Error m Se Me' 2 2
Source: Griffing (1956b)

126
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

The variances for variance components can be estimated i, j  1,..., p,

by the following equations, 1
Yij    gi  g j  sij    ijkl  , where
1 1 bc k l k  1,...,b
Var(ˆ g ) 
2
Mg 
2
Ms
2
l  1,...,c
2 p( p  1)( p  2) 2 p( p  2) 2 ( p  3) 
1 1  is the population mean, gi , g j is the general
Var(ˆ s )  Ms  (M e ' )2 ,
2 2

p( p  1) 2m combining ability effect for the ith and jth parents sij
1 1
Var(ˆ r )  Mr  (M e ' )2 , is the specific combining ability effect of the cross
2 2

p( p  1) 2m
between the ith and jth parents such that sij = s ji and
2
Var(ˆ 2 )  (M e ' )2 ,
m eijkl is the experimental error effect unique to the ijklth
2.5.4 Method IV observation (Griffing, 1956b). The variances of the
In this method, only one set of F1’s are included. It is the effects may be estimated as follows:
most common of the diallel crossing systems. There are 2 p 1
Var( ˆ )  ˆ 2 , Var( gˆ i )  ˆ 2 , and
a  p( p  1) / 2 different genotypes evaluated. As for p( p  1) p( p  2)
other diallel methods, there are two models. p  3 2 (i  j ) .
Var(sˆij )  ˆ
Fixed model: p 1
Table 11. Skeleton of ANOVA for Diallel method IV.
Expected mean squares
Source df SS MS
Model I Model II
GCA p 1 1   2 s  ( p  2) g 2
) g
Sg Mg 2 2
 2  ( p  2)( 2

p 1
i
i

SCA p p  3 / 2 Ss 2  2   s2
Ms 2   si j 2
p( p  3) i j

Error m Se Me' 2 2

Source: Griffing (1956b)

This mating design provides information on GCA and provides only half-sibs. It provides SCA of each cross,
SCA (Griffing, 1956b). However, the fixed model of and it is not providing GCA of lines only but of the testers
method 3 or 4 is the most appropriate for obtaining also, as liner and tester both are different sets of
unbiased estimates of combining abilities and gene genotypes (Sharma, 2006). It is therefore most suitable
action (Shattuck et al., 1993). This method is most for animal experiment (Sharma, 2006). In addition, it is
suitable when there are no genotypic reciprocal effects used in estimating various types of gene actions
(Griffing, 1956b). The most of the problems arising with important in the expression of quantitative traits
diallel crosses are essentially due to experimental design (Rashid et al., 2007).
such that analysis of data is complex (Johnson and King, Its statistical model is:
1998). Yijkl    al  bkl  vij  (av)ijl   ijkl , where Yijkl =
LINE X TESTER DESIGN
observed value from each experimental unit; μ =
Line x tester is basically an extension of topcross design
population mean; al = location effect; bkl = block or
in the sense that instead of one tester as used in
replication effect within each location; vij = F1 hybrid
topcross, more than ones testers are used under L x T
mating design. Line x tester mating design was first effect = gi + gj+ sij. However, vij  gi  g j  sij ,
proposed by Kempthorne in 1957 cited by Sharma
where gi = general combining ability (GCA) for the ith
(2006). This design involves hybridization between lines
parental line; gj = GCA effect of jth tester;
(f) and wide based testers in one to one fashion
sij = specific combining ability (SCA) for the ijth F1 hybrid
generating f x m = fm hybrids (Sharma, 2006). It is the
and (av)ijl = interaction effect between ith F1 hybrid and
simplest mating design that provides both full-sibs and
lth location; and Ɛijkl = residual effect.
half-sibs simultaneously as opposed to topcross which

127
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

Table 3. Skeleton of ANOVA for Line X Tester Design

Expected mean squares
Source df MS
Model I Model II
Replication r 1
Lines m 1 M1  2  rf
1
  gi
2
 2  vsca  rf gca( m )
m 1 i
Testers f 1 M2  2  rm(
1
) g j
2
 2  rvsca  rmgca( f )
f 1 j
Line x tester (m  1)( f  1) M3  2  r
 1 
  sij
 2  rvsca
 (m  1)( f  1)  i j
Error (r  1)(mf  1) M4 2 2

Source: Sharma (2006)

The significance of mean square for line x testers Falconer, D.S. and T.F.C. Mackay. 1996. Introduction to
provides a direct test of significance of dominance quantitative genetics. 4th ed. Pearson Prentice Hall,
variance, σ2D while significance of σ2A is provided by Harlow, England.
significance of lines and testers mean squares. Gorz, H.J. and F.A. Haskins. 1971. Evaluation of cross-
CONCLUSION fertilization in forage crops. Agronomy--Faculty
Selection of suitable parents and good mating designs publications, University of Nebraska. Lincoln. p.
are keys to the successful of plant breeding schemes 731-734.
(Khan et al., 2009). However, selection of mating design Griffing, B. 1956. Concept of general and specific
depends on many factors, for instance in early stages of combining ability in relation to diallel crossing
breeding programme, there are a large numbers of systems. Aust. J. Biol. Sci. 9: 463-493.
different lines, clones or accessions to be evaluated. In Hallauer, A.R., M.J. Carena and J.B.M. Filho. 2010.
this case, appropriate designs would include the Quantitative genetics in maize breeding. 6th ed.
polycross and perhaps the topcross but the polycross is Springer, Iowa, USA.
particularly suited to the identification of potential Hill, J., H.C. Becker and P.M.A. Tigerstedt. 1998.
parents of synthetic varieties. A comparative evaluation Quantitative and ecological aspects of plant
of mating designs summarized mating designs in two breeding. Chapman and Hall, UK, London.
ways (1) in terms of coverage of the population and it Johnson, G.R. and J.N. King. 1998. Analysis of half diallel
was reported that BIPs > NC I > polycross > NC III > NC II mating designs: I - A practical analysis procedure
> diallel, in that order of decreasing effectiveness; (2) in for ANOVA approximation. Silvae Genetica. 47: 74-
terms of amount of information, diallel > NC II > NC III > 79.
NC I > BIPs (Acquaah, 2012). However, diallel mating Kearsely and Jinks. 1968. A general method of detecting
design is the most important for GCA and SCA but additive, dominance and epistatic variation for
breeder has a great role in choosing mating design metrical traits: I. Theory. Heridity. 23: 403-409.
instead of relying on the mating design per se. The Kearsey. 1970. Experiment sizes for detecting
proper choice and use of a mating design will provide dominance variation. Heredity 24: 45-57.
the most valuable information for breeding (Acquaah, Kearsey, M.J. 1965. Biometrical analysis of a random
2012). mating population: a comparison of five
REFERENCES experimental designs. Heredity. 20: 205-235.
Acquaah, G. 2012. Principles of plant genetics and Kearsey, M.J. and H.S. Pooni. 1996. The genetical analysis
breeding. 2nd ed. Wiley-Blackwell, Oxford. of quantitative traits. 1st ed. Chapman and Hall,
Aly, R.S.H., E.M.R. Metweli and S.T.M. Mousa. 2011. London, UK.
Combining ability of maize (Zea mays L.) inbred Khan, S.A., H. Ahmad, A. Khan, M. Saeed, S.M. Khan and B.
lines for grain yield and some agronomic traits Ahmad. 2009. Using line x tester analysis for
using topcross mating design. Glob. J. Mol. Sci. 6: 1- earliness and plant height traits in sunflower
8.

128
 J. Plant Breed. Genet. 01 (03) 2013. 117-129

(Helianthus annuus L.). Recent Research in Science Rashid, M., A.A. Cheema and M. Ashraf. 2007. Line x
and Technology. 1: 202–206. tester analysis in Basmati rice. Pak. J. Bot. 39:
Mather, K.J., J. L 1982. Bimotrical geneticsChapman & 2035-2042.
Hall, London. Schlegel, R.H.J. 2010. Dictionary of plant breeding. 2 ed.
Morgan, J.P. 1988. Polycross designs with complete CRC Press,Taylor & Francis Group, Boca Raton.
neighbor balance. Euphytica. 39:59-63. Sharma, J.R. 2006. Statistical and biometrical techniques
doi:10.1007/bf00025112. in plant breeding. 1 ed. New Age International.
Mosa, H.E. 2010. Estimation of combining ability of New Delhi.India
maize inbred lines using topcross mating design. J. Shattuck, V.I., B. Christie and C. Corso. 1993. Principles of
Agricul. Res. Kafer El-Sheikh University. 36: 1-14. Griffing's combining ability analysis. Genetica. 90:
Panhwar, S.A., M.J. Baloch, W.A. Jatoi, N.F. Veesar and 73-79.
M.S. Majeedano. 2008. Combining ability estimates Sleper, D.A. and J.M. Poehlman. 2006. Breeding field
from line x tester mating design in upland cotton. crops. 5th ed. Blackwell Publshing, Iowa, USA.
Pak. Acad. Sci. 45: 69-74. Wricke, G. and W.E. Weber. 1986. Quantitative genetics
and selection in plant breeding.Walter de Gruyter
& Co., Berlin.

129