Selection of Wheat Ideotype Based on Multiple Traits using
Genotype by Yield-Trait Approach
Muhammad Faheem1*, Khalil Ahmed Laghari1,
Muhammad Khalil-Ur-Rehman2, Saima Mir Arain1 and Mahboob Ali Sial1
1Division of Plant Breeding and Genetics, Nuclear Institute
of Agriculture, Tando Jam, Sindh, Pakistan
2Beijing Advanced Innovation Center for Tree Breeding by
Molecular Design, Beijing Forest University, P. R. China
*For correspondence: mlofaheem@gmail.com
Received 06 January 2021; Accepted 16 February 2021; Published 10 May
2021
Abstract
In plant breeding, a novel genotype-by-yield trait (GYT)
biplot approach was introduced to select superior genotypes based on multiple
traits. The present study demonstrated the application of the GYT biplot model
to evaluate the superior wheat advanced lines from a panel of 24 genotypes to
select the ideotype for end users. Results show that the genotype-by-trait (GT) biplot covered 57% of the
total variation of the data to reveal that grain yield was strongly associated
with 1000-grain weight and grain width. In contrast, the GYT biplot explained
90.2% of the total variation which was significantly much higher than GT
biplot. According to tester vector view of GYT biplot almost all the yield
trait combinations were associated with each other at different degree of
association; whereas the genotypes present within the acute angles of tester
vectors (yield trait combinations) had the trait profile contributed positively
towards grain yield. The polygon biplot of GYT had eight sectors, out of which
only three had the yield trait combinations. The eight genotypes were the
polygon vertex among which the advanced line DF1906 of first sector was
designated as the best genotype for spike length, number of spikelets per
spike, grain weight per spike and number of grains per spike. Additionally, the
DF1912 of second sector was early maturing coupled with high 1000-grain weight
while DF1917 of third sector had short stature and gave the highest harvest
index. The average tester coordination (ATC) biplot grouped 13 genotypes as
superior and nine as inferior genotypes and recommended two advanced lines
DF1912 and DF1917 as ideotype based on balanced traits profile. These findings
strengthened the argument that the GYT biplot analysis is better than other
selection indices and guaranteed the selection of superior genotypes and
rejection of inferior ones based on multiple traits yield combinations. © 2021
Friends Science Publishers
Keywords: Wheat;
Genotype by yield*trait biplot; Grain yield; Multiple traits
Introduction
Wheat (Triticum
aestivum L.) is the most important food security crop in the world which
provides basic calories and protein to 36% of the human population on earth (Li
et al. 2020). In Pakistan, wheat
exclusively fulfills 60% of the caloric requirement of the population (Joshi et al. 2017). Despite the fact, the
grain yield of wheat has increased many folds since the last few decades but
the pace of improvement in yield (0.8–1.2%) is not enough to feed 9–10 billion
human population of the globe by 2050 (Tilman et al. 2011). To meet such a huge demand, wheat researchers are
striving hard to accelerate the genetic gain for further improvement in the
wheat grain yield and its quality.
Like other field crops, wheat also possesses a number of
morphological and agronomic traits besides grain yield which must be considered
during the evolution of high yielding cultivar. However, evaluation and
selection of wheat genotypes based on multiple traits considering grain yield
as the most important one is a tedious job due to the interaction of traits
with the environment and also among themselves (Bernardo 2010). The scenario
becomes more challenging when traits of interest are negatively correlated with
each other (Yan et al. 2007). On the
other hand, the wheat growers always expect and demand that new wheat cultivars
should outclass the old cultivars specifically in grain yield coupled with
improved agronomic traits and disease resistance (Merrick et al. 2020). Under such circumstances, selection based on multiple
traits needs deep insight about genotype trait relationship and association
among traits to combine optimum values of traits of interest in a single
genotype. Yan et al. (2019) argued
that the economic value of any trait in a cultivar depends upon the level of
the main target trait i.e., yield.
For instance, lodging resistance in wheat has the economic value only when it
is coupled with high grain yield. Similarly, a genotype possessing high level
of quality traits but with poor grain yield has low economic value of trait due
to the fact that this genotype cannot be recommended as a cultivar. Hence, the
main purpose of multiple trait selection is to pyramid the desirable traits
with higher grain yield in a single genotype in such a way that all the traits
escalate the economic value of one another.
Different conventional and molecular breeding approaches
have been devised to tailor wheat genotype depicting a balance between grain
yield with other agronomic traits not only under optimal environmental
conditions but also under various biotic and abiotic stresses. Among
conventional breeding approaches, tandem selection (Simmonds and Smartt 1999),
independent culling (Godshalk et al.
1988), and index selection (Bos and Caligari 2007) are well known, effective,
and common practiced tools to breed a cultivar based on multiple traits (Boureima and
Abdoua 2019; Yan et al. 2019). But
all these approaches rely on some truncation points or weightages which are at
the disposal of researcher’s choice, hence resulted in different selection
outputs from the same dataset (Yan and Fregeau-Reid 2018). To remove such
biasness, recently a novel genotype-by-yield trait (GYT) biplot approach was
introduced in plant breeding to select superior genotypes based on multiple
traits (Yan and Fregeau-Reid 2018). Basically, the GYT biplot is extension of
genotype by trait (GT) biplot analysis which relies on combing different traits
with yield (Merrick et al. 2020). The
different GYT biplot views provide the accurate mean to display the ranking of
genotypes according to the superiority of yield trait combinations and also the
trait profile of each genotype to depicts its weaknesses and strength (Yan et al. 2019). After the pioneer study in
oats (Yan and Fregeau-Reid 2018), this model was successfully practiced in
other crops such as Hordeum vulgare
(Karahan and Akgun 2020), Sesamum indicum
(Boureima
and Abdoua 2019) T. durum (Kendal 2019;
Mohammadi 2019), and spring wheat (Merrick et
al. 2020) to select the superior genotypes based on multiple traits. The
same model (GYT biplot) was applied in this study to get better understanding
about the associations of desirable traits in combination with key grain yield
trait in elite wheat germplasm and subsequently to identify the superior
genotypes based on multiple traits selection to speed up the genetic gain of
wheat grain yield improvement in Pakistan.
Materials and
Methods
Plant
material and experimental design
The plant
material included 23 wheat elite advanced lines along with a commercial check cv. NIA-Saarang was used in this study.
These advance lines were developed by wheat breeding group of Nuclear Institute
of Agriculture (NIA), Tando Jam, Pakistan. The check cultivar NIA-Saarang is a
high yielding, rust resistant, and widely adopted wheat variety of NIA. The
trials were conducted in the first week of November at experimental farm of NIA
over two consecutive growing seasons (2018–19 and 2019–20). In 2018–19, alpha
lattice experimental design with two replications was adopted to evaluate these
advance lines against the commercial check while in 2019–20 the same genotypes
were evaluated in randomized complete block design with three replications. The
plot size of each genotype for both the trials was 9 square meter comprising
six rows of 5-meter length. To maintain the healthy growth of wheat plots, the
recommended dose of fertilizers (120 kg ha-1 nitrogen 60 kg ha-1
potassium and 60 kg ha-1 phosphorous) were applied. The crop cycle
for each season was completed with four irrigations while other weed control
standard practices were followed as established in the region.
For both the growing seasons, data regarding agronomic
traits like days to heading (DH) and days to physiological maturity (DM) were
recorded for each plot. At maturity, ten plants were randomly selected from the
central rows of each plot of all the genotypes for each replication. Data
regarding plant height (PH) and yield components viz., spike length (SL), number of spikelets per spike (SPS),
number of grains per spike (GS), grain weight per spike (GWS), and thousand
kernel weight (TGW) were recorded from these plants. To determine the grain
length (GL) and grain width (GW), the cumulative grain length and width of five
grains were measured to compute the respective parameters of one grain. For
biological yield, four central rows were harvested and weighed, subsequently,
the same bundle was threshed to record grain yield per plot (GY) and expressed
as tons per hectare. Harvest index (HI) was computed from biological yield and
grain yield per plot to express in percentage.
Statistical
analysis
The dataset
was subjected to a novel approach of genotype by yield-trait biplot analysis
following the protocol as described by Yan and Fregeau-Reid (2018). In brief,
at the first step overall means and standard deviation for two years for all
the parameters including grain yield were computed. This genotype trait data
table was used to find the association among different traits and also for
genotype by trait (GT) biplot analysis. Besides this, the GT data table was
transformed to genotype by yield-trait (GYT) table by either multiply or
dividing grain yield of each genotype with its respective parameter depending
upon the breeding objectives. So, in the GYT table grain yield was divided by
DH, DM, and PH with a notation of “/” as our objective of developing wheat
advance lines were early maturing with short stature which could resist
lodging. While all the other yield components (SL, SPS, GWS, TGW, GL, and GW) including
harvest index was multiplied (*) with grain yield as larger means of these
traits were more desirable. Eventually, before the final evaluation of
genotypes the GYT table was standardized to remove the differences in the
measuring units of yield trait combinations according to following equation:
Where Pij
represents the standardized value of ith genotype for the trait or yield-trait
combination j in the standardized table, Tij is the original value of genotype
i for trait or yield-trait combination j in the GT or GYT tables, Tj is the
mean across genotypes for trait or yield-trait combination j, and Sj is the
standard deviation for trait or yield-trait combination j. This standardized
dataset of GYT was then subjected to GYT biplot analysis and to calculate the
mean superiority index value of each genotypes. The GT biplot and different
views of GYT biplot were constructed by GGEbiplot software v. 8.2 following
same software setting as suggested by Yan and Fregeau-Reid (2018).
Results
Association among
traits and trait profile based on genotype by trait (GT) biplot
The GT biplot analysis was applied to the standardized
data set of two growing seasons of 24 wheat genotypes of wheat to get deep
insight about the relationship among the traits. The GT biplot presented in Fig.
1 was based on the principles (scaling =1, centering =2 and SVP= 2) laid down
by Yan and Fregeau-Reid (2018) to represent the correlation among the traits
and with the genotype which covered 57% of the total variation of the data by
plotting two main principal components (PC1 =36.4% and PC2 = 20.6%). The GT
biplot revealed that grain yield (GY) formed acute angle with thousand grain
weight (TGW), hence strongly interlinked with one another, whereas, weak
positive association was also found among GY, grain width (GW) harvest index
(HI), grain weight per spike (GWS) and grain length (SL) (Fig. 1). Contrarily,
all other traits either formed right angel or obtuse angles (> 90o)
with GY hence, depicted no or negative association with GY. The Pearson
correlation of the same data set presented in Fig. 2 also confirmed the
findings of GT biplot with some exceptions. For instance, correlation analysis suggests
significant positive correlation between GY and HI while no significant
association was found for other traits including TGW (Fig. 2). Similarly,
another important yield component i.e.,
GWS formed acute angels with TGW, GL, PH, SL, GS, DM and SPS to represent
strong association of these traits with GWS as well as among each other. In
fact, the GT biplot formed a cluster of multiple traits viz., GWS, GL, PH, SL, GS, DM SPS and DH to depict the positive
association of diverse strength among these traits depending upon the magnitude
of angle between any of two given traits (Fig. 1). Additionally, all the
members of this cluster had obtuse angle with HI to represent negative
correlation with HI. The values of Pearson correlation graphically displayed in
Fig. 2, also validated the same trend of association among traits as revealed by
GT biplot.
Graphical display of trait profiles of genotypes is
another unique feature of GT biplot which mainly depends upon the total
variation explained by the biplot. The GT biplot shows that the genotype DF1912, DF1903, DF-1917, and
DF1915 produced better grain yield than check cultivar i.e., NIA-Saarang and other contesting genotypes and also showed better trait profiles for TGW and GW (Fig.
1). Similarly, DF1923 had the maximum HI as compared to rest of genotypes while
other genotypes with good profile of HI were NIA-Saarang, DF1920,
DF1918, and DF1916. The genotypes DF1910, DF1907, DF1908, DF1906, DF1905, DF1904, and
DF1909 were found within the cluster of multiple traits viz., GWS, GL, PH, SL, GS, DM SPS and DH depicted good trait
profile for these traits as these traits had positive association with each
other but performed poor for HI and GW. The advanced lines DF1901, DF1911,
DF1914 and DF1919 clustered away from all trait vectors under consideration, hence, had poor
trait profile for the traits (Fig. 1).
Association
among traits based on grain yield by Genotype by yield-trait (GYT) biplot
To
select the best genotypes among the contesting advanced lines based on multiple
traits, a novel approach of GYT biplot was applied considering grain yield as
the most important economical trait. The GYT biplot analysis represented the
90.2% of the total variation by plotting first two principal components (PC1 = 79.5% and
PC2 = 10.7%) in three exclusive biplot views named as the tester
vector view (Fig. 3), the polygon view (Fig. 4), and the average tester
coordination (ATC) view (Fig. 5). The ATC view of GYT biplot clearly
demonstrated the two clusters of yield trait combinations, one having GY/PH,
GY*HI, GY/DH, GY/DM and GY*GW while other consisted of GY*SL, GY*SPS, GY*GWS
and GY*GS (Fig. 3). In first cluster, GY/PH and GY*HI had very strong
correlation while GY/DH, GY/DM and GY*GW had same level of relationship among
each other. In other cluster, GY*SL and GY*GWS exhibited strong relationship
with GY*SPS and GY*GS, respectively. However, wider angle between the members
of these clusters presented weak relationship among yield trait combinations,
specifically for GY/PH and GY*HI with the members of second group (Fig. 3). Two
yield trait combinations i.e., GY*TGW
Fig. 1: The genotype by trait biplot based on standardized data
of 23 wheat advanced lines and check ‘NIA-Saarang’
GY, grain
yield; DH, days to heading; DM, days to physiological maturity; PH, plant height; SL,
spike length; SPS; number of spikelets per spike; GS,
number of grains per spike; GWS, grain weight per spike; TGW, 1000-grain
weight; HI, harvest index
Fig. 2: Correlation among traits (lower diagonal) and grain
yield trait combinations (upper diagonal) of 24 wheat genotypes
The cross
symbol represents the non-significance association
and GY*GL highly correlated with each other were found
in between these two clusters and showed the adequate level of relatedness for
all the yield trait combinations.
Selection of
the best genotype based on multiple traits
The
polygon views also known as which is best for what biplot was constructed for
all the yield trait combination to graphically present the trait profile of
contesting genotypes (Fig. 4). The GYT biplot showed that the polygon consisted
of eight genotypes which were present at the farthest distances from the origin
at vertex positions to encompass all the remaining genotypes. These vertex
genotypes included DF1906, DF1912, DF1917, DF1923, DF1916, DF1914, DF1909 and
DF1904. Additionally, the eight perpendicular rays radiating from the origin of
polygon distributed the biplot into eight sectors, among which only three
sectors possessed the yield trait combinations. The first sector between 1st
and 8th radiating rays comprised of four yield trait combinations viz., GY*SL, GY*SPS, GY*GW and GY*GS and
6 advanced lines viz., DF1906,
DF1907,
Fig. 3: The test vector view of genotype by yield*trait (GYT)
biplot to represent the association among grain yield trait combinations
Fig. 4: The best view of genotype
by yield*trait (GYT) biplot to represent the vertex genotypes of the polygon
DF1905, DF1910, DF1908 and DF1903 among which DF1906 was
the winner genotype. It implied that the genotype DF1906 had the best trait
profile for these yield components as compared to rest of the genotypes. The
second sector between 1st and 2nd radiating rays harbored
six yield trait combinations including GY/DH, GY/DM, GY*GW, GY*TGW, and GY*GL
with only one vertex genotype (DF1912) while third sector between 2nd
and 3rd rays also had only one genotype (DF1917) present at vertex
position which performed quite well for GY/PH and GY*HI. Taken together these
results depict that the advanced line DF1912 and DF1917 were early maturing
short stature genotypes with excellent trait profile for grain weight, grain
yield and harvest index. In contrast, the polygon view
also described that the remaining 16 genotypes including check cultivar were
present in those sectors which had any yield trait combinations, implying that
these genotypes were the poor performer of studied traits in combination to
grain yield as compared to rest of genotypes (Fig. 4).
The ATC view of GYT biplot presented in Fig. 5, best
explained the ranks of 24 genotypes based on the performance of multiple traits
and usefulness. Overall, the ATC biplot plotted 13 genotypes at the right side
of the double head arrow on the ATA axis to represent as better performers than
the average of all the yield trait combinations. Hence, these genotypes could
be ranked as DF1912 > DF1917 > DF1907 > DF1906 > DF1921 > DF1923
> DF1918 >DF1920 > DF1913 > DF1908 > DF1905 > DF1910 >
DF1903 (Fig. 5). The positive values of mean superior index (SI) of all these
genotypes also authenticated the outcomes of GYT biplot (Table 1). The advanced
line DF1915 present at the origin of biplot had the same level of performance
as that of overall average of all the traits. In contrast, 10 genotypes
including check cultivar NIA-Saarang appended on the left side of double head
arrow had the poorer performance than the average of all the traits. All these
genotypes had negative mean SI values (Table 1), which implied that these
genotypes had not the suitable combinations of traits as desired. It is evident
from the ATC biplot that the four advanced line DF1912, DF1917, DF1907 and
DF1908 had the trait profiles better than the overall average of all the
traits, hence, outclassed other contesting genotypes including check cultivar.
Among these lines, DF1912 was at the top position followed by DF1917 which had
the potential to produced maximum grain yield at the expense of studied traits
and signified ideotypes of the testing panel of genotypes to be selected on the
basis of breeding objectives. Both of these advanced lines DF1912 and DF1917
had the maximum mean SI values of 1.50 and 1.44, respectively for which each
trait contributed positively toward grain yield (Table 1).
Discussion
Table 1: The mean standardized genotype by yield*trait data of
two years and the mean superiority index (SI) of 24 wheat genotypes
Genotypes |
GY/DH |
GY/DM |
GY/PH |
GY*SL |
GY*SPS |
GY*GS |
GY*GWS |
GY*GL |
GY*GW |
GY*TGW |
GY*HI |
Mean SI |
DF-1901 |
-1.69 |
-1.25 |
-1.07 |
-0.87 |
-0.92 |
-0.65 |
-1.28 |
-0.85 |
-1.87 |
-1.67 |
-1.39 |
-1.23 |
DF-1902 |
-0.61 |
-0.89 |
-0.68 |
-0.75 |
-0.69 |
-0.81 |
-0.99 |
-0.49 |
-0.69 |
-0.91 |
-0.78 |
-0.75 |
DF-1903 |
0.44 |
0.11 |
-0.33 |
-0.03 |
0.52 |
-0.39 |
0.47 |
0.67 |
0.49 |
1.28 |
-0.35 |
0.26 |
DF-1904 |
-0.82 |
-0.63 |
-0.98 |
0.93 |
0.03 |
1.13 |
0.87 |
-0.59 |
-0.50 |
-0.55 |
-1.18 |
-0.21 |
DF-1905 |
0.19 |
0.20 |
-0.17 |
0.89 |
0.53 |
0.66 |
0.60 |
0.52 |
-0.12 |
0.26 |
-0.20 |
0.30 |
DF-1906 |
0.62 |
0.79 |
0.06 |
0.67 |
2.21 |
2.53 |
2.02 |
1.12 |
0.71 |
0.40 |
-0.10 |
1.00 |
DF-1907 |
0.70 |
0.82 |
0.62 |
1.78 |
1.16 |
0.82 |
1.31 |
1.32 |
1.04 |
1.55 |
0.86 |
1.09 |
DF-1908 |
0.79 |
0.56 |
-0.07 |
0.72 |
0.58 |
0.13 |
0.30 |
0.25 |
0.56 |
0.56 |
-0.13 |
0.39 |
DF-1909 |
-1.40 |
-1.59 |
-1.88 |
-1.14 |
-0.93 |
-0.43 |
-0.61 |
-1.15 |
-1.54 |
-1.39 |
-1.52 |
-1.24 |
DF-1910 |
0.17 |
-0.22 |
-0.60 |
0.77 |
0.23 |
0.32 |
1.29 |
0.36 |
-0.11 |
1.24 |
-0.20 |
0.30 |
DF-1911 |
-0.27 |
-0.53 |
0.47 |
-0.25 |
-0.36 |
-0.57 |
-1.28 |
-0.49 |
-0.74 |
-1.36 |
-0.46 |
-0.53 |
DF-1912 |
1.76 |
1.66 |
1.50 |
1.70 |
1.17 |
1.43 |
1.18 |
1.67 |
2.04 |
1.04 |
1.31 |
1.50 |
DF-1913 |
1.07 |
0.68 |
1.27 |
-0.03 |
0.17 |
0.00 |
0.09 |
0.12 |
-0.02 |
0.56 |
0.39 |
0.39 |
DF-1914 |
-1.86 |
-1.63 |
-1.29 |
-1.96 |
-1.06 |
-1.34 |
-1.23 |
-1.95 |
-1.39 |
-1.21 |
-1.27 |
-1.47 |
DF-1915 |
0.22 |
-0.08 |
-0.10 |
-0.29 |
-0.04 |
-0.58 |
-0.12 |
-0.04 |
-0.07 |
0.42 |
0.70 |
0.00 |
DF-1916 |
-1.33 |
-1.54 |
-1.21 |
-1.40 |
-2.60 |
-1.79 |
-1.47 |
-1.57 |
-1.20 |
-1.02 |
-1.00 |
-1.47 |
DF-1917 |
1.10 |
1.73 |
1.69 |
1.43 |
1.05 |
1.36 |
1.14 |
1.71 |
1.35 |
1.18 |
2.06 |
1.44 |
DF-1918 |
0.61 |
0.74 |
0.86 |
0.15 |
-0.08 |
0.38 |
0.21 |
0.27 |
0.61 |
0.34 |
0.95 |
0.46 |
DF-1919 |
-1.27 |
-1.03 |
-0.03 |
-0.97 |
-0.95 |
-1.09 |
-1.52 |
-1.26 |
-1.06 |
-1.57 |
-1.10 |
-1.08 |
DF-1920 |
0.57 |
0.63 |
0.77 |
-0.40 |
1.10 |
0.34 |
0.05 |
-0.58 |
1.12 |
0.02 |
0.80 |
0.40 |
DF-1921 |
0.81 |
1.20 |
1.31 |
0.75 |
0.30 |
0.54 |
0.10 |
1.02 |
0.70 |
0.43 |
1.09 |
0.75 |
DF-1922 |
-0.61 |
-0.36 |
-0.42 |
-0.59 |
-0.59 |
-0.74 |
-0.52 |
-0.10 |
-0.38 |
-0.08 |
-0.02 |
-0.40 |
DF-1923 |
1.07 |
0.91 |
1.30 |
-0.07 |
-0.10 |
-0.45 |
0.08 |
0.80 |
0.98 |
0.87 |
1.54 |
0.63 |
NIA-Saarang |
-0.25 |
-0.27 |
-1.02 |
-1.03 |
-0.72 |
-0.82 |
-0.70 |
-0.75 |
0.11 |
-0.40 |
-0.03 |
-0.53 |
GY, grain yield;
DH, days to heading; DM, days to physiological maturity; PH, plant height; SL,
spike length; SPS; number of spikelets per spike; GS,
number of grains per spike; GWS, grain weight per spike; GW, grain width; GL, grain
length; TGW, 1000-grain weight; HI, harvest index
Fig. 5: The average tester
coordination (ATC) view of genotype by
yield*trait (GYT) biplot to rank wheat genotypes based on multiple traits
During the development of crop cultivars, plant breeders
usually focus on the selection of a couple of specific traits in connection
with main economical trait. It is because selection based on multiple traits
may disturb the balance among the traits due to negative association among
different traits and the interaction of these traits with the environment (Yan et al. 2007; Kendal 2019). In contrast,
the value of cultivar increases for end users when it is evolved on the basis
of multiple traits (Karahan and Akgun 2020). Therefore, statisticians and
breeders are continuously putting effort to develop an effective model to
select superior genotypes based on multiple traits. In this regard, initially
genotype by trait (GT) analysis was proposed to understand the relationship
among traits and genotypes (Yan and Rajcan 2002) and was utilized by many
breeders in cereals as well as in other crops (Rubio et al. 2004; Oladejo et al.
2011; Legesse et al. 2013; Paramesh et al. 2016). However, this GT biplot
model cannot show the strengths and weaknesses of a genotype, hence deprives of
providing a decisive power of selection or rejection of given genotype (Yan et al. 2019). An extension of GT biplot
analysis named as genotype by yield*trait (GYT) biplot model was recently
proposed by Yan and Fregeau-Reid (2018) to select or reject the genotypes based
on multiple traits (Merrick et al.
2020). The same model following all the steps and principles was applied in the
present study to select the best advanced lines based on multiple traits from
the panel of 24 contesting wheat genotypes. At first, GT biplot was constructed
from the average standardized data set of two years revealed that the GY was
strongly linked with TGW and weak correlation was found for GW, whereas no
other traits had the significant association with grain yield. Additionally,
other yield components like GWS, GS SPS, SL and GL were associated with each
other in different strengths of associations while these traits were negatively
correlated with harvest index and GW (Fig. 1). The Pearson correlation table
also validated the same trend of associations among traits with some exceptions
(Fig. 2). According to correlation table, the grain yield was significantly
associated with only HI not with TGW. This might be due to the fact that not
enough variation was found among the genotypes for grain yield as also revealed
by short vector length of GY in GT biplot. The other reason included the
fitness of good of GT biplot which was quite low as compared to GYT biplot
(90.2%) which means that GT biplot could only explain 57% of the total
variation.
At second step of GYT biplot analysis, the GT table was
transformed to GYT biplot by multiplying or dividing grain yield with other
traits depending upon the breeding objectives. As, we were looking for early
maturing, short stature and high yielding genotypes for the agro-climatic
conditions of Pakistan, therefore, the grain yield was divided by days to
heading, days to physiological maturity and plant height while all other trait
parameters were multiplied to get the standardized GYT table (Table 1). Three
unique views of GYT biplot were constructed (Fig. 3, 4 and 5) from this data
set which explained 90.2% of the total data variation, significantly much
higher than GT biplot. Like other studies, our results also demonstrated that
GYT biplot was better and accurate than the GT biplot to display the actual
variation in dataset for reliable selection (Yan et al. 2019). The tester vector view of GYT biplot demonstrated
that almost all the yield trait combinations were associated with each other at
different degree of association which could not be explained by GT biplot.
Secondly, the genotypes present within the acute angles of tester vectors
(yield trait combinations) had the trait profile contributed positively towards
grain yield which could not be assessed by GT biplot. For instance, GT biplot
depicted that the advanced lines DF1909, DF1904 and DF1905 had the features of early maturing (Fig. 1) but the GYT
biplot placed these genotypes away from all the tester vectors related to early
maturity (GY/DH and GY/DM) advocating that these genotypes may had the early
maturing featuring but not contributed positively towards grain yield which was
the goal of any breeding program. A similar case was with DF1923 which had the
highest HI but the strength of this trait was not in the optimum balance of
other traits to be selected as an ideotype. However, all these genotypes could
be used as donor parents for the improvement of these traits in breeding
programs. Recently, Yan et al. (2019)
also pointed out that the economic value of any trait in a cultivar depends
upon the level of main target trait i.e.,
grain yield.
The polygon views also known as which is best for what
biplot is an excellent graphical presentation of trait profile of genotypes.
The results of polygon view showed that 8 genotypes were the polygon vertex
among which the advanced line DF1906 was the best genotype for main yield
components including spike length, number of spikelets per spike, grain weight
per spike and number of grains per spike, while DF1912 was early maturing
coupled with highest thousand grain weight and DF1917, was low stature with
highest harvest index (Fig. 4). In contrast, the remaining five vertex
genotypes DF1923, DF1916, DF1914, DF1909 and DF1904 were winners of respective
sectors but did not possess the desired level of multiple traits to be selected
as ideotype. The ATC view is the most unique feature of GYT biplot as it
displays the ranks of contesting genotypes based on strengths and weaknesses of
each genotypes which cannot be viewed in other biplots including GT biplot
(Karahan and Akgun 2020; Yan et al.
2019). This view categorized the inferior and superior genotypes groups by
drawing a perpendicular doubled head arrow on ATA axis which were present on
left and right side of the arrow, respectively. Additionally, a small hollow
circle on main ATA axis representing the average of all yield trait
combinations further subdivided the superior group into the most desirable
genotypes (ideotype) and desirable ones (Yan and Fregeau-Reid 2018). In this
study, the ATC biplot grouped 13 genotypes as superior and 9 genotypes as
inferior while one genotype DF1915 was at the boundary of these groups (Fig.
5). All the superior advanced lines outclassed the check cv. NIA-Saarang while four genotypes crossed the small hollow
circle on ATA axis. Among these genotypes, DF-1912 and DF1917 had the best
traits profile as also depicted by GYT table that all the yield combinations of
these genotypes had positive values. Between these two ideotypes, DF1912 was
better and more stable than the DF1917 as it had the shorter projection on ATA
axis. In contrast, the nine inferior genotypes had the poor trait profile when
assessed in combination with yield, hence could be rejected on the basis of
multiple traits.
Conclusion
This research endeavor was meant to demonstrate the
application of GYT biplot approach to select the superior wheat advance line
based on multiple traits to release the high value cultivars for end users. The
GYT biplot analysis clearly categorized the contesting wheat lines into
superior and inferior group and recommended two ideotypes i.e., DF1912 and DF1917 on the basis of balanced trait profile. Our
findings also strengthen the argument that the GYT biplot analysis is better
than other selection indices and guaranteed the selection of superior genotypes
and rejection of inferior ones when evaluated on multiple traits in connection
with yield.
Acknowledgements
The research
project was financially supported by the funds of Pakistan Atomic Energy
Commission.
Author Contributions
MF developed the experimental material,
designed and executed the experiment, recorded the data, and wrote the
manuscript; KAL and SMA developed the experimental material, MK statistically
analyzes the data, MAS supervised and reviewed the manuscript.
Conflict of Interest
Authors declare no
conflict of interest.
Data
Availability
The data will
be made available on fair request to the corresponding author.
Ethics
Approval
Not applicable.
References
Bernardo R (2010). Selection for multiple trait. In: Breeding
for Quantitative Traits in Plants, pp:343‒358. Stemma press, Minnesota, USA
Bos I, P Caligari (2007). Index selection and
independent culling method. In: Selection
Methods in Plant Breeding, 2nd edn, pp:318‒324. Springer, Netherlands
Boureima S, Y Abdoua (2019). Genotype
by yield* trait combination biplot approach to evaluate sesame genotypes on
multiple traits basis. Turk J Field Crops 24:237‒244
Godshalk E, D Timothy, J Burns (1988). Effectiveness of
index selection for switchgrass forage yield and quality. Crop Sci 28:825‒830
Joshi KD, AU Rehman, G Ullah, MF Nazir, M Zahara, J
Akhtar, M Khan, A Baloch, J Khokhar E Ellahi, A Khan, M Suleman, M Imtiaz (2017).
Acceptance and competitiveness of new improved wheat varieties by smallholder
farmers. J Crop Improv 31:608‒627
Karahan T, I Akgün (2020). Selection of barley (Hordeum vulgare) genotypes by GYT
(genotype × yield × trait) biplot technique and its comparison with GT
(genotype × trait). Appl Ecol Environ Res 18:1347-1359
Kendal E (2019). Comparing durum wheat cultivars by
genotype × yield × trait and genotype× trait biplot method. Chil J Agric Res
79:512‒522
Legesse H, N Dechassa, S Gebeyehu, G Bultosa, F Mekbib
(2013). Multivariate analysis as a tool for indirect selection of common bean
genotypes (Phaseolus vulgaris L.) for
soil acidity tolerance under field conditions. Sci Technol Arts Res J 2:07‒15
Li J, I Dundas, C Dong, G Li, R Trethowan, Z Yang, S
Hoxha, P Zhang (2020). Identification and characterization of a new stripe rust
resistance gene Yr83 on rye
chromosome 6R in wheat. Theor Appl Genet
133:1095‒1107
Merrick LF, KD Glover, D Yabwalo, E Byamukama (2020).
Use of genotype by yield* trait (GYT) analysis to select hard red spring wheat
with elevated performance for agronomic and disease resistance traits. Crop
Breed Gene Genomics 2; Article e200009
Mohammadi R (2019). Genotype by yield*trait biplot for
genotype evaluation and trait profiles in durum wheat. Cer
Res Commun 47:541‒551
Oladejo AS, R Akinwale, I Obisesan (2011).
Interrelationships between grain yield and other physiological traits of cowpea
cultivars. Afr Crop Sci J 19:189‒200
Paramesh M, D Reddy, MS Priya, P Sumathi, P Sudhakar, K
Reddy (2016). GT biplot analysis for yield and drought related traits in mung
bean (Vigna radiata Wilczek). Electr
J Plant Breed 7:538‒543
Rubio J, J Cubero, L Martin, M Suso, F Flores (2004).
Biplot analysis of trait relations of white lupin in Spain. Euphytica 135:217‒224
Simmonds NW, J Smartt (1999). Genetic aspects: Populations
and selection. In: Principles of Crop
Improvement, 2nd edn, pp:62‒114. Oxford, Blackwell, USA
Tilman D, C Balzer, J Hill, BL Befort (2011). Global
food demand and the sustainable intensification of agriculture. Proc Natl Acad Sci 108:20260‒20264
Yan W, I Rajcan (2002). Biplot analysis of test sites
and trait relations of soybean in Ontario. Crop
Sci 42:11‒20
Yan W, J Fregeau-Reid (2018). Genotype by yield* trait
(GYT) biplot: A novel approach for genotype selection based on multiple traits.
Sci Rep 8; Article 8242
Yan W, J Fregeau-Reid, N Mountain, J Kobler (2019).
Genotype and management evaluation based on genotype by yield*trait (GYT)
analysis. Crop Breed Genet Genomics 1;
Article e190002
Yan W, MS Kang, B Ma, S Woods, PL Cornelius (2007). GGE
biplot vs. AMMI analysis of
genotype-by-environment data. Crop Sci
47:643‒653