Document 13561376

advertisement
AN ABSTRACT OF THE THESIS OF
Robert C. Gaynor for the degree of Master of Science in Crop Science presented on
May 27, 2010.
Title: Quantitative Trait Loci Mapping of Yield, its Related Traits, and Spike
Morphology Factors in Winter Wheat (Triticum aestivum L.).
Abstract approved:
_____________________________________________________________________
C. James Peterson
Increasing grain yield in wheat (Triticum aestivum L.) is a challenging task,
because yield is a complex trait controlled by many genes and highly influenced by
environmental factors. The genetic control of yield components and other traits
associated with yield may be less complex and thus more manageable for breeding.
This study seeks to identify quantitative trait loci (QTLs) for these traits. Two new
genetic linkage maps were constructed from recombinant inbred lines (RILs) derived
from crosses between the Oregon soft white winter wheat variety Tubbs and a Western
European hard red winter wheat variety, Einstein. A third linkage map was
constructed from RILs from a cross with Tubbs and a Western European experimental
hard red winter wheat line. A combination of Diversity Arrays Technology (DArT),
Simple Sequence Repeat (SSR), orw5, and B1 markers were used to construct genetic
linkage maps. Two replications of the RIL populations were grown in yield trial sized
plots at Corvallis, OR and Pendleton, OR in 2009. The RILs were evaluated for grain
yield, spikes per m2, fertile spikelets per spike, sterile spikelets per spike, seeds per
spike, seeds per fertile spikelet, average seed weight, growing degree days (GDD) to
flowering, GDD to physiological maturity, GDD of grain fill, plant height, test weight,
and percent grain protein. Composite interval mapping (CIM) detected 146 QTLs for
these traits spread across all chromosomes except for 6D. Thirty six percent of all of
the QTLs detected were in close proximity to four loci: Rht-B1, Rht-D1, B1, and
Xgwm372. The use of factor analysis to aid in QTL mapping for correlated traits
related to spike morphology was explored. Quantitative trait loci mapping of factor
scores for these traits potentially showed an increase in statistical power to detect
QTLs and a decrease in the probability of type I error over mapping the traits
individually.
©Copyright by Robert C. Gaynor
May 27, 2010
All Rights Reserved
Quantitative Trait Loci Mapping of Yield, its Related Traits, and Spike Morphology
Factors in Winter Wheat (Triticum aestivum L.)
by
Robert C. Gaynor
A THESIS
submitted to
Oregon State University
in partial fulfillment of
the requirements for the
degree of
Master of Science
Presented May 27, 2010
Commencement June 2011
Master of Science thesis of Robert C. Gaynor presented on May 27, 2010.
APPROVED:
_____________________________________________________________________
Major Professor, representing Crop Science
_____________________________________________________________________
Head of the Department of Crop and Soil Science
_____________________________________________________________________
Dean of the Graduate School
I understand that my thesis will become part of the permanent collection of Oregon
State University libraries. My signature below authorizes release of my thesis to any
reader upon request.
_____________________________________________________________________
Robert C. Gaynor, Author
ACKNOWLEDGEMENTS
I would like to thank Adam Heesacker for his work in constructing the genetic linkage
maps used in this study. Thanks to Mark Larson, Eddie Simons, Mary Verhoeven,
Dolores Vasquez, and the wheat project’s summer crew for helping collect data. I
would also like to thank Lan Xue, Oscar Riera-Lizarazu, Mike Flowers, Jim Peterson,
and Jeff Leonard for advising on technical aspects of this research. Finally, I would
like to thank Jeffery Stone for serving as a graduate council representative on my
committee.
TABLE OF CONTENTS
Page
1. INTRODUCTION.....................................................................................................1
2. LITERATURE REVIEW.........................................................................................6
2.1. Yield Components and Crop Ideotype........................................................6
2.2. Donald's Ideotype......................................................................................10
2.3. Genetic Markers........................................................................................12
2.4. Linkage Analysis.......................................................................................15
2.5. QTL Mapping............................................................................................17
2.6. Factor Analysis..........................................................................................22
2.7. Crop Models..............................................................................................26
3. MATERIALS AND METHODS............................................................................28
3.1. Plant Populations.......................................................................................28
3.2. Data Collection and Summary Statistics...................................................29
3.3. Linkage and QTL Mapping.......................................................................30
3.4. Factor Analysis..........................................................................................32
4. RESULTS AND DISCUSSION.............................................................................34
4.1. Genetic Linkage Maps...............................................................................45
4.2. QTL Mapping............................................................................................45
4.3. Factor Analysis..........................................................................................56
5. CONCLUSION.......................................................................................................64
6. BIBLIOGRAPHY...................................................................................................66
TABLE OF CONTENTS (Continued)
Page
7. APPENDICES.........................................................................................................73
LIST OF TABLES
Table
Page
Table 1 Measured traits.................................................................................................31
Table 2 Least square means for parents and RILs in TxN population..........................36
Table 3 Least square means for parents and RILs in ExT1 population........................37
Table 4 Least square means for parents and RILs in ExT2 population........................38
Table 5 Summary of across location analyzes for the three populations of RILs........39
Table 6 Pearson product-moment correlations for the Tubbs x NSA 98-0995 RILs...41
Table 7 Pearson product-moment correlations for the ExT1 population......................42
Table 8 Pearson product-moment correlations for the ExT2 population......................43
Table 9 QTLs for yield, grain protein, test weight, and spikes per m2.........................46
Table 10 QTLs for flowering, ripening, and grain fill duration....................................47
Table 11 QTLs for plant height and seeds per spike.....................................................48
Table 12 QTLs for fertile spikelets per spike and sterile spikelets per spike...............49
Table 13 QTLs for seeds per fertile spikelet and seed weight......................................50
Table 14 Pearson product-moment correlations between initial factors and yield,
grain protein, and test weight........................................................................................57
Table 15 Factor loadings for rotated spike morphology traits......................................59
Table 16 QTLs for spike morphology factors...............................................................60
LIST OF APPENDIX FIGURES
Figure
Page
Figure 1 Genetic linkage map for the Tubbs x NSA 98-0995 population (TxN)
with QTL positions.......................................................................................................74
Figure 2 Genetic linkage map for the larger Einstein x Tubbs population (ExT1)
with QTL positions.......................................................................................................78
Figure 3 Genetic linkage map for the smaller Einstein x Tubbs population (ExT2)
with QTL positions.......................................................................................................82
Quantitative Trait Loci Mapping of Yield, its Related Traits, and Spike
Morphology Factors in Winter Wheat (Triticum aestivum L.)
1. INTRODUCTION
Increasing grain yield is a primary goal of wheat (Triticum aestivum L.)
breeders. Accomplishing this goal is a challenging task due to the complexity of the
trait. Grain yield is a complex trait, because it is controlled by many genes and its
expression is heavily influenced by environmental factors. Wheat breeders produce
large numbers of crosses every year to develop higher yielding varieties. These
crosses are in turn used to generate numerous inbred lines with unknown yield
potential that have to be evaluated for a litany of agronomic and quality traits before
any of these lines can become eligible for release as a variety. For example, Oregon
State University's (OSU) wheat breeding program typically generates over 600 crosses
and evaluates around 40 thousand F4 and F5 rows of inbred lines each year.
Accurately measuring yield of these lines is only possible in subsequent generations
when there is sufficient seed to plant replicated yield trials at multiple locations. Thus,
several cycles of selection have to be carried out on those lines before they are ever
evaluated for yield. Once sufficient seed is generated, the high cost of running yield
2
trials severely limits the number of lines that can be evaluated. At OSU, less than 300
of the initial 40 thousand rows will be screened for yield in replicated trials. Many of
these lines will prove to have insufficient yields to be candidates for variety release.
Finding a way to screen lines earlier and at a lower cost could vastly improve a
breeder's ability to generate higher yielding varieties.
Marker assisted selection (MAS) is a potential method for selecting for yield at
early generations. This method uses genetic markers to select for favorable plants
based on knowledge of associations between those markers and traits. Genetic
markers only require a small sample of DNA, so limited quantities of seed are not a
problem. Screening a line with genetic markers is also significantly cheaper than
screening that line in a replicated yield trial. However, there are several limitations to
MAS that have to be addressed first.
The most common way of finding markers for MAS is to identify quantitative
trait loci (QTLs) using QTL mapping (Lande and Thompson, 1990). Statistical
limitations of QTL mapping for complex traits, like yield, can make such an approach
impractical (Doerge, 2002). However, QTL mapping for traits that influence yield
may be more successful, because they often have simpler genetic control and are less
influenced by environment (Yin et al., 2000). MAS for traits that influence yield may
then be an indirect means of applying early generation selection for yield.
The traits that influence yield include disease resistance, stress tolerance,
maturity, and plant architecture. A great deal of research has gone into elucidating the
3
genetic control for disease resistance (e.g. Singh et al., 2000), stress tolerance (e.g.
Zhou et al., 2007), maturity (e.g. Yan et al., 2003) and plant architecture, particularly
yield components (e.g. Li et al., 2007). This has lead to successful, albeit limited,
implementation of MAS techniques for disease and stress response in wheat (e.g.
Murphy et al., 2009; Baerstmayr et al., 2009; Kumar et al., 2010). Markers are
available for vernalization genes, genes that partially influence maturity, which can be
used in MAS (e.g. Sherman et al., 2004). Dwarfing genes, major determinants of
plant height, also have markers that can be used for MAS (Ellis et al., 2002). Besides
plant height, there is little literature to suggest that successful MAS strategies for yield
components and other architectural traits have been implemented.
Successfully incorporating QTLs for these traits into a MAS program will
likely not be accomplished by simply pyramiding genes. This is due to the lack of an
obvious “more is better” philosophy for these traits that is present with disease
resistance and stress tolerance. Breeders will instead have to manipulate QTLs for
yield components and other architectural traits to achieve a crop ideotype. An
ideotype is an idealized structure of a plant that obtains the maximum performance
under production conditions. Finding optimum values for yield components and
architectural traits that obtain this ideotype will require consideration of their
interactions with each other, related traits, and the environment (Loomis et al., 1979).
Combining knowledge of these interactions with knowledge of the QTLs that control
these traits can be accomplished with “QTL-based ecophysiological modeling” (Yin et
4
al., 2003). This approach uses a statistical model to estimate yield of a variety based
on its composition of QTLs and the environmental and production conditions that
variety is grown under.
Before such an approach can be considered, markers for QTLs that are relevant
to germplasm that a breeder is using have to be available. The QTLs that have already
been identified may not be relevant to OSU's germplasm. These QTLs may simply
not be present in the germplasm. QTL effects are dependent on genetic background,
so the QTLs already identified may perform very differently in OSU's germplasm than
they did in the studies that identified them (Babu et al., 2004). This could mean that
the already identified QTLs for traits related to yield have no practical value to OSU's
breeding project.
The goal of this study is to identify QTLs for traits related to yield in crosses
that are representative of elite germplasm in OSU’s wheat breeding program. The
genetic mapping populations used herein contained crosses with an Oregon variety,
Tubbs, and varieties from Nickerson, a private breeding program based in Western
Europe. Material from Nickerson has been a source of well adapted lines for
enhancing germplasm at OSU and has also been a source of lines for direct release as
varieties. Knowledge of important QTLs in the genetic background of these crosses
could be applicable to a wide range of crosses in OSU’s wheat breeding program.
Furthermore, this study examines using QTL mapping of factor scores as a
way of dealing with numerous correlated traits in QTL analyzes. The use of factor
5
scores would create new variables that are not correlated and can reduce the number of
variables that are analyzed with QTL mapping. This should result in a reduction of
type I error compared to analyzing traits individually. It may also be possible to
increase the power to detect QTLs with this method. Factor scores will be calculated
for traits related to spike morphology. The results of QTL mapping for these factors
will be compared to mapping done on the original traits to see if any of these potential
benefits are realized.
6
2. LITERATURE REVIEW
2.1. Yield Components and Crop Ideotype
The study of yield in cereals has often been based on exploring the
relationships between yield and other traits. In particular, traits that are referred to as
yield components have been the subject of considerable study. Engledow and
Wadham (1923) may have been the first to partition the yield of cereal crops into its
components. Their list of components included, among others: plants per unit area,
number of ears per plant, number of grains per ear, and weight per grain. They
proposed that if the favorable combinations of these traits were known, producing
higher yielding lines is possible by generating crosses that obtain these combinations.
Although they didn't phrase it as such, this favorable combination of traits can be
considered an ideotype. An ideotype simply refers to the idealized appearance of a
plant variety. Subsequent research has produced some support for Engledow and
Wadham's proposal by showing evidence that changes in the distribution of yield
components in the highest yielding varieties can partially explain increases in yield
over time (Calderini, 1995). Changes in maturity and plant height have also been
attributed to increased yield and were the driving forces behind the “Green
7
Revolution” (Borlaug, 1983). Thus, determining a wheat ideotype requires
consideration of various components of plant architecture and maturity.
Varieties are ultimately judged on their performance in production settings.
Therefore, studies that evaluate performance in markedly different conditions may not
be meaningful for determining an ideotype (Donald, 1968). For example, many
studies on yield components and plant architecture have been conducted on individual
plants or thinly seeded plots (e.g. Quarrie et al., 2006; Li et al., 2006). The
performance of these individual plants and thinly seeded plots is not likely to be
representative of how the plants will perform under production settings (Nerson,
1980). For this reason, these studies may have limited to no applicability when it
comes to constructing an ideotype. Only studies of yield components as they behave
in a community of plants are likely to further the creation of a crop ideotype.
It is widely known that the optimum maturity for wheat in one environment
can be different in another environment. Thus, the ideotype for one region can be
different from the ideotype in another region. This difference may extend beyond
maturity. For example, seeds per m2 and tillering have been shown to be the two most
correlated components with yield in a study conducted in Kansas (Donmez et al.,
2001). This is consistent with the widely accepted view that gains in yield are
primarily due to increases in seeds per unit area (Slafer et al., 1996). However,
selection based on kernel weight has been shown to be a more effective way of
breeding for higher yielding lines in a study conducted in Oklahoma (Alexander et al.,
8
1984). Calderini et al. (1995) examined a historical accession of varieties in
Argentina and found an increase in seeds per unit area as the driving force for
increasing yield in varieties released before 1987 and found seed weight to be more
important in varieties released after 1987. These discrepancies could be the product of
the different populations studied, but it may also represent differences in ideotypes for
these regions. If that is the case, findings from one region may not be of any value to
another region.
In addition to variation between environments there is also variation between
years that have to be considered. The variation in wheat yields from one year to the
next in one field can be large (Washmon et al., 2002). Thus, data from a single year
may not be representative of an average year. This year to year variation will have to
be taken into account when formulating an ideotype. First, care must be taken to not
give too much weight to a single year’s data. The ability of an ideotype to have stable
yields over multiple years will have to be considered.
Large degrees of correlation between yield components and other traits such as
maturity and height have been observed in several studies (e.g. Li et al., 2007; Kato et
al., 2000). These correlations can impact studies that attempt to determine the
structure of an ideotype. For example, multiple regression of yield on its components
is an approach that previously was often used for studying yield components (Walton,
1971). The results from these analyzes may be misleading, because the interpretation
of the calculated coefficients is how much a change in one trait will change yield when
9
all other traits are held constant. Since the traits are correlated, it is not likely and
perhaps not possible to change just one while keeping the others constant.
Furthermore, the estimates of standard errors for those coefficients are not accurate
when traits are correlated. Thus, consideration of these correlations is crucial. This
has lead to research into yield components that include a statistical analysis involving
the correlation matrix such as path analysis (e.g. Fonseca and Patterson, 1968) or
factor analysis (e.g. Walton, 1971). These sorts of analyzes give information on how
changing one yield component effects yield and the other traits simultaneously. How
much an individual trait can be manipulated and how consistent the effects of those
manipulations are over a large range of changes is not determined with these analyzes.
This drastically limits the ability of these approaches to predict the performance of
varieties not studied.
More recent studies into yield components to identify QTLs have largely
ignored correlations between traits (e.g. Börner et al., 2002; Marza et al., 2005). Even
when the correlations are considered, the extent of the examination is often limited to
reporting a correlation matrix and suggesting that collocated QTLs for different traits
represent close linkage or pleiotropy (e.g. Kato et al., 2000; Li et al., 2007; Cuthbert et
al., 2008). These correlations could partially be examined with multiple-trait QTL
analyzes, and such an approach would increase the power of QTL detection (Mi et al.,
2008). The massive computational requirement of analyzing many traits at once has
been a limiting factor to the adoption of this technique. An alternative is to use
10
principal component analysis to reduce the correlated traits into latent, uncorrelated
variables. This approach has been used with tassel morphology in maize (Upadyayula
et al., 2006) and for morphological factors in wheat (Addisu et al., 2009). The study
on tassel morphology found QTLs for the principal components that were not
identified with single trait analysis. This might have represented increased power
from the use of principal components. The study in wheat didn’t find any additional
QTLs. This could be due to the inclusion of too many traits in the principal
component analysis. This large number of traits were probably controlled by many
genes, so there was no realized benefit of increased statistical power.
2.2. Donald's Ideotype
Donald (1968) presented a model for a wheat ideotype in an article titled “The
Breeding of Crop Ideotypes”. He attempted to draw on the research available at the
time to predict what a wheat ideotype would look like. His proposed ideotype would
have short, strong stems to withstand lodging. He pointed to the work being done by
others with dwarfing genes and the general trend towards shorter wheat as evidence
supporting this view. The subsequent “Green Revolution” in wheat that is in large
part attributed to the introduction of dwarfing genes verifies Donald's conclusion
(Borlaug, 1983). Donald's ideotype also had erect leaves to allow for maximum
11
illumination of the photosynthetic surfaces. The plant would have a few small leaves.
Small leaves were deemed better than large leaves, because they are more likely to be
erect. He also noted that theoretical work suggested that an arrangement of several
uniformly dispersed small leaves was better at capturing light than a few large leaves.
Donald proposed that the ear should be large to assure an adequate sink for
photosynthates and to allow for high levels of grain production. The ear should also
be erect and have awns to allow for maximum capture of light. A high proportion of
seminal roots were deemed desirable for gathering soil nutrients. Donald also stated
that the degree to which tillering is desirable is a product of production practices. He
suggested that it would be best to grow an ideotype that only had a single culm,
because the production of tillers is wastes a plant's resources when it results in
overproduction of tillers that are subsequently aborted. Changing the production
practices, namely seeding rate, would then be needed to best fit this plant type.
The main characteristic of Donald's ideotype, as proposed by him, is that the
variety should be a poor competitor. Donald argued that varieties that are strong
competitors use their resources to out compete their neighbors. In a well managed
production setting, those neighbors are primarily plants of the same variety. Thus,
there is no benefit to the yield of the crop when the plants spend resources trying to
out-compete other plants that also contribute to the overall yield. Therefore, it would
be better to have a variety that more efficiently use those resources for the production
of yield. This trait of Donald’s ideotype is probably the most important for breeders,
12
because early generation selections are likely to favor strong competitors. For
example, F3 generation plants are typically grown in bulk plots in breeding projects.
Under these conditions each plant will have neighbors of different genetic composition
next to them. The strongest competitors will out compete their weak neighbors. Thus,
the stronger competitors will yield more and appear to have a more favorable
phenotype than the weak competitors. This will likely bias the breeder towards
selecting for the strong competitors. It may therefore be more prudent for breeders to
select plants that don't necessarily look the best at these generations. These plants
may actually produce higher yielding varieties according to Donald.
2.3. Genetic Markers
Understanding the genetic control of yield components and other traits related
to yield requires the use of genetic markers. Genetic markers are powerful tools for
exploring the genetic composition of organisms. They can be used to construct
genetic maps that can subsequently be used to locate genes controlling various traits
within an organism. They can also be used by breeders for identifying favorable
genotypes in MAS.
For genetic markers to be useful, they have to be able to distinguish between
different alleles at a locus. When this is the case, a genetic marker is said to be
13
polymorphic. The ability of a marker to detect polymorphisms can be classified as
either dominate or co-dominate. Dominate markers have a binary response; they are
either expressed or not expressed. These types of markers are only able to detect the
presence of one particular allele at a locus, and they cannot definitely distinguish
between homozygous and heterozygous individuals. Negative results (lack of
expression) are not very informative with these markers, because there are potentially
multiple genotypes that cause this result. Thus, two individuals that share a positive
result can be said to share at least one allele at a locus. Two individuals that share a
negative result can’t be considered a match, unless it is known that there are only two
alleles present at that locus. Co-dominate markers are more informative than
dominate markers. These types of markers are able to distinguish between multiple
alleles at a locus. This gives these markers the ability to detect heterozygotes and they
are more frequently polymorphic than dominate markers (Paux and Sourdille, 2009).
The majority of genetic markers use at least one of two important
biotechnological tools (Paux and Sourdille, 2009). These tools are restriction
endonucleases (or restriction enzymes) and polymerase chain reaction (PCR).
Restriction enzymes cleave double stranded DNA at specific recognition sites. These
recognition sites are short sequences of DNA that are interspersed in plant and animal
DNA (Voet and Voet, 2004). Markers that utilize restriction enzymes detect
polymorphisms based on differences in the locations of these recognition sites (Paux
and Sourdille, 2009). PCR is a technique that amplifies DNA using a set of genetic
14
probes called primers and a heat stable polymerase. The polymerase is used to
exponentially increase the number of copies of DNA in regions that are flanked by
primers. Correct primer selection allows polymorphic regions of DNA to be amplified
and used as genetic markers (Voet and Voet, 2004). Primers can also be chosen to
amplify fragments created using restriction enzymes. This can be accomplished by
ligating a piece of DNA, called an adapter, to the end of these fragments that contains
a sequence of DNA that the primers bind to (Voet and Voet, 2004).
The most commonly used genetic markers for constructing wheat genetic
linkage maps are Amplification Fragment Length Polymorphisms (AFLPs), Simple
Sequence Repeats (SSRs), and the proprietary Diversity Array Technology (DArT)
markers (Paux and Sourdille, 2009). AFLPs are dominate markers that are generated
by cutting DNA with two restriction enzymes, ligating DNA to the end of the
fragments, and amplifying fragments with PCR. The main benefit of these markers is
that they don’t require any development. SSRs are codominate makers that amplify
regions of the genome that contain short, repeated sequences of DNA. These markers
are highly polymorphic and fairly high throughput, but they are very expensive to
develop. DArTs are dominate markers that are used in large capacity microarrays to
screen fragments of DNA for hybridization with probes. These markers are ultrahigh
throughput. On a per marker basis they are inexpensive, but thousands of markers are
screened at once making them expensive to use.
15
2.4. Linkage Analysis
To improve the usefulness of genetic markers, it is best to know where those
markers are located in the genome. The method for determining this is linkage
analysis. Linkage refers to the tendency of genes to be inherited together. This occurs
when they are located on the same chromosome and in relatively close proximity to
each other. Linkage can also be observed between marker loci. The examination of
linkage through linkage analysis is the foundation for constructing genetic maps (Wu
et al., 2007). Traits that are inherited qualitatively and by only one gene can be
directly incorporated into those genetic maps using linkage analysis (Kaeppler et al.,
1993). Wheat is an allohexaploid, so many of its qualitative traits are controlled by
three genes. Thus, if more than one of those genes are segregating in a population this
method cannot be used for mapping the trait. The real power in genetic maps is the
ability to place quantitative traits on those maps using QTL analysis (Wu et al., 2007).
Genetic maps that contain the locations of important traits can then be used by
breeders for identifying markers for MAS.
The first step in linkage analysis is obtaining a population of individuals that
share the same parents. Either F2, backcross, doubled haploid (DH), or recombinant
inbred line (RIL) populations are typically used. The DH and RIL populations are the
most common, because subsequent generations can be obtained through self-crossing
16
that will not differ genetically. This makes it easy to repeat experiments using these
populations in additional years. All of these populations allow for the measurement of
linkage between sets of markers that are polymorphic in the parents (Wu el al., 2007).
If the loci are unlinked, equal proportions of the parental, non-recombinant, and nonparental, recombinant, pairings of marker alleles would be expected. The lower the
proportion of recombinants between two loci, the higher the linkage between them is
likely to be. The greater the linkage between loci the closer they are in terms of
genetic distance. It is thus a matter of ordering loci according to their linkage to find
where they fit on a genetic map.
In theory, finding the optimum order of loci can be determined simultaneously
by maximizing the likelihood function for all possible orders. In practice there are
x!/2 orders for x loci, so the number of calculations that are required with an actual set
of markers would be impractical for current computing technology (Wu et al., 2007).
To overcome this limitation, the order of loci is determined using sets of only two or
three loci at a time (Wu et al., 2007). After considering many of these sets it is
possible to determine an overall order for all loci. The distance between these loci is a
function of the recombination rates between them. Recombination rates are not
additive, so for the purpose of creating a linkage map the recombination rates are
transformed to additive measures of genetic distance called Morgans. This
transformation is performed using one of several available mapping functions. The
most popular mapping functions are Haldane and Kosambi (Wu et al., 2007). The
17
difference between these functions is how they handle the occurrence of crossover
events between two markers. Haldane’s original formula called for a parameter that
controls the occurrence of crossovers called coincidence (1919). The mapping
function known as Haldane’s mapping function sets this parameter to one (Crow and
Dove, 1990). This means the function assumes that crossovers occur randomly and
independently of each other, so they can be modeled using a Poisson distribution (Wu
et al., 2007). Kosambi (1944) introduced a function that assumes the coefficient of
coincidence is proportional to the recombination rate between the markers. This
results in interference at short distances that diminishes as the distance increases
(Crow and Dove, 1990).
2.5. QTL Mapping
After constructing genetic maps, the next step for locating areas of the genome
that control a trait of interest is QTL mapping. QTL mapping is a statistical method
that aims to find associations between traits and regions of DNA. The regions that
associate with the trait are known as QTLs. Markers that are closely linked to these
QTLs are commonly used in MAS to select for the desired allele of the target QTL.
Many methods for conducting QTL mapping have been proposed.
18
The simplest technique for QTL mapping is Single Marker Analysis (SM).
This technique tests if different marker alleles at a locus are associated with different
values of the phenotype. This technique doesn't require a genetic map and there can
be multiple marker alleles tested at once (Wu et al., 2007). Analysis of variance
(ANOVA) is used to test the significance of the different marker alleles. This
approach has severe limitations. First, estimation of the QTL’s effect is confounded
with the genetic distance between the marker and the QTL (Lander and Botstein,
1989). That is, the less linkage the marker has with the QTL, the more the effect of
the QTL will be underestimated. Multiple comparisons are also problematic for single
marker analysis. Due to the large number of markers that are commonly tested, a
large number of type I errors are expected (Jansen, 1994). Multiple comparison
techniques can be used to reduce the incidence of type I error, but they will also
reduce the power of the test. Performing SM in a genetic mapping population can
offer some graphical tools that can help interpreting the results. In a genetic mapping
population there are only two alleles possible at each locus, so a t-test is used to test
the significance of different alleles. The genetic map also allows for the t-values to be
plotted according to their map position. Then multiple markers that test significant for
a QTL and are located next to each other on the genetic map can be assumed to be
measuring a single QTL. This makes using a genetic map with SM a better way of
interpreting the results when it comes to estimating how many QTLs are present and
their locations.
19
Another way of using markers to find QTLs would be to use them as predictors
in a multiple regression model. However, simultaneously estimating the partial
regression coefficients is statistically challenging, because the number of markers may
be greater than the number of individuals and many of the coefficients are close to
zero (Wu et al., 2007). Xu (2003) proposed using a Bayesian approach to estimating
the coefficients when marker density is high to cope with this simultaneous estimation
problem. His approach is referred to as whole-genome marker analysis. Compared to
SM this approach gives fewer significant peaks, and better estimates the effects and
locations of the QTLs (Xu, 2003).
Lander and Botstein (1989) formulated a QTL mapping approach that is
referred to as Interval Mapping (IM). This approach utilizes a genetic map in
performing the calculations for detecting QTLs. The genetic map allows IM to test for
the presence of a QTL in the intervals between markers. This is accomplished by
maximizing the likelihood function for QTL effect with given values of the flanking
markers, considered random variables, and constant values for recombination
frequencies between the QTL and each marker. A likelihood ratio (LR) is then
constructed comparing this model to the model that assumes there is no QTL present.
Traditionally the expectation-maximization algorithm developed by Dempster et al. in
1977 is used, but a regression approximation to the LR statistic that is computationally
simpler can be used as an alternative (Haley and Knott, 1992). The procedure is
repeated at multiple locations and the resulting LR statistics are plotted against their
20
location on the genetic map. With either method, significance of the LR statistics can
be tested using a permutation test that calculates the experiment-wise error (Churchill
and Doerge, 1994). Peaks on the map that surpass the significance threshold are used
as estimates for QTL locations and effects. A permutation test for significance that
sets levels differently for major and minor QTLs has been proposed, but has not been
widely adopted (Doerge and Churchill, 1996). IM only tests for the presence of one
QTL at a time, so the presence of multiple QTLs can impact the accuracy of the
estimates for location and effect of a QTL (Zeng, 1993). Like SM, IM is prone to type
I and type II error (Jansen, 1994).
Zeng (1994) along with Jansen and Stam (1994) independently proposed a
QTL mapping technique that is called Composite Interval Mapping (CIM). This
approach expands IM to include covariates chosen from markers outside the test
interval. The covariates work to offset the impact that QTLs outside of the test
interval have on the performing the likelihood test. There are several methods for
choosing markers as covariates available. Zeng (1994) originally tested using all
unlinked markers or all markers. Using all unlinked markers gave greater power and
using all markers gave better estimates of QTL effects. An additional method using
multiple regression modeling techniques of forward, reverse, or forward-reverse
selection to choose markers has become popular. These techniques chose from
markers not used in the test site and outside of a certain distance from the test site.
They are chosen at a distance from the test site to prevent from eliminating the QTL
21
signal. Viewing results and determining significance in CIM is conducted in a similar
fashion as with IM.
Jiang and Zeng (1994) proposed an extension of CIM to handle more than one
trait that is called Multiple-Trait Composite Interval Mapping (MT-CIM). This test
only tests one site at a time and uses covariates like CIM, but it looks at more than one
trait at a time. This approach offers greater power to detect QTLs when traits are
inversely correlated. It also provides a mechanism to test a site that has effects on
more than one trait for pleiotropy or close linkage. This can be a valuable tool to
breeders, because it allows them to assess if it is possible to break unfavorable
combinations of traits. Permutation tests can be used to set significance thresholds as
in CIM, but there is some debate as to what is an appropriate null model. A null
model that assumes evenly spaced QTLs for the traits might be more favorable than a
model that assumes there are no QTLs. In practice, MT-CIM is severely limited by
the computational difficulty of handling many traits at one time.
Multiple Interval Mapping (MIM) provides a mechanism for testing multiple
QTLs simultaneously (Kao et al., 1999). This allows for interactions between QTLs,
epistasis, to be examined. However, the level of computation required for this
analysis is greater than those previously mentioned. Thus, putative QTLs detected
using another method, such as CIM, are best used in the first steps of model building
in MIM (Hackett et al., 2000). The models can then be further refined by optimizing
the positions and effects of the QTLs and testing for interactions (Kao et al., 1999). A
22
permutation test for determining significance levels doesn't exist for this method.
Thus, minimizing a model selection criterion is the preferred method for fitting models
(Hackett et al., 2000). However, this method is not able to fully protect against over
fitting a model, so other factors such as not exceeding the heritability of a trait are
often considered. MIM can be expanded to consider multiple traits.
2.6. Factor Analysis
There are two ways to deal with multiple correlated traits in QTL mapping:
map each trait separately or use a multiple-trait mapping procedure. Both approaches
have disadvantages. Mapping single traits inflates type I error, because multiple
analyzes are being performed. Multiple-trait analyzes have greater power to detect
QTLs than single trait analyzes, but they are computationally prohibitive and there is
no universally accepted model for determining significance of a QTL. Combining
factor analysis, a multivariate dimension reduction analysis, with QTL mapping may
be a beneficial alternative to those two methods. Factor analysis would reduce the
number of variables to be studied and those variables would be uncorrelated.
The beginnings of factor analysis date back to attempts by Karl Pearson,
Charles Spearman, and others in the early 20th century to quantify intelligence
(Johnson and Wichern, 2007). The technique has been used by cereal breeders and
23
agronomists as a way of examining correlated traits such as yield components (e.g.
Walton, 1971). The intent of factor analysis is to explain inter-correlations between a
set of variables as coming from a few underlying latent variables called common
factors. The analysis uses the covariance matrix resulting from this set of variables or
its standardized version the correlation matrix. The factors are then extracted by one
of three methods: principle component, common factor, or maximum likelihood. The
number of factors to be considered is determined and then their coefficients are either
examined directly, or after an orthogonal or oblique factor rotation. Factor scoring
coefficients can also be calculated. These can be used to determine the factor values
of individuals for use in subsequent analyzes.
The principle component method for factor extraction is the most popular
(Johnson and Wichern, 2007). The first factor is extracted by finding a linear
combination of the variables that explains as much of the total variance as possible.
The second factor is then determined by finding a linear combination of the variable
that is orthogonal to the first and explains as much of the remaining variance as
possible. This is process is then iterated until the desired number of factors are
obtained, to a maximum of a number of factors equal to the number of variables. The
researcher can use multiple methods for determining how many factors to retain. The
eigenvalue greater than unity and the scree test are the most popular, but are not
necessarily the most accurate (Tanguma, 2000). Parallel analysis is widely considered
the most accurate method for selecting the number of factors to retain (Hayton et al.,
24
2004). This method compares the eigenvalues of the observed factors to those of
randomly generated data to determine if a factor should be retained. Since the
principal component method uses all of the variance and not just the shared variance
to extract factors it is not considered a true method of factor analysis (Costello and
Osborne, 2005). That is, it is really a method of data reduction and not a method for
identifying latent variables.
A true factor analysis can be performed with the common factor method. The
common factor method is performed on a reduced correlation matrix (Johnson and
Wichern, 2007). This reduced correlation matrix has the variance that is specific to a
variable removed from the diagonal of the matrix. Thus, the matrix only represents
the variance shared between variables. There are multiple methods for determining
what values should be used for the diagonals. After one of these methods is chosen,
the factor solution is then obtained iteratively. However, using the reduced matrix
now makes it possible to obtain negative eigenvalues for some of the factors since the
reduced matrix doesn't represent all of the variance. The common factor method more
accurately reproduces population patterns than the principal component method, but
these differences converge when the number of variables is large and the loading
patterns increase (Snook and Gorsuch, 1989).
The maximum likelihood method is another form of a true factor analysis that
assumes the common and specific factors are normally distributed (Johnson and
Wichern, 2007). This method can give fairly different results from the common factor
25
method. It is also possible to perform hypothesis tests for the adequacy of the number
of factors retained when the assumptions of the maximum likelihood method are met.
This gives an additional tool to researcher for deciding how many factors to retain.
However, the normal distribution restriction can make the maximum likelihood
method unusable for some sets of data.
Factor rotation is often applied to aid in the interpretation of factors. The
mathematical justification for factor rotation is that orthogonally rotated factors will
explain the covariance just as well as the original factors, so there is no unique
solution for finding factors (Johnson and Wichern, 2007). The varimax rotation, the
most common orthogonal rotation, seeks a rotation that makes as many coefficients
close to zero as possible (Kaiser, 1957). The resulting rotated factors can then be
easier to interpret than the originals, because individual factors have fewer variables
making large contributions to them. In cases where the underlying factors are not
assumed to be independent, oblique rotations may be more appropriate (Johnson and
Wichern, 2007). These types of rotations seek to explain variables with as few factors
as possible. Even after performing one of these types of rotations there can still be a
good deal of ambiguity in the interpretation of the factors. The researcher has to rely
on his experience to make sense of the factors.
The lack of uniqueness in the factor solution that allows for rotation is also a
reason that many researchers are opposed to using factor analysis (Kaiser, 1957). The
large number of subjective decisions that a researcher has to make can bias the results
26
of the analysis. These biases can have the greatest impact when it comes to selecting
the number of factors to retain and interpreting the resulting factor coefficients
(Costello and Osborne, 2005). Overall, factor analysis can appear to be more of an art
than a science and great care should be taken when interpreting the results (Johnson
and Wichern, 2007).
2.7. Crop Models
Simply finding QTLs for a trait is of no use unless it is known what the ideal
value for that trait is. A method for determining the optimum values for yield
components and other traits related to yield is using crop models. Crop models are
statistical models that are typically used to predict the yield of a crop. They use
predictors based on weather patterns, production practices, nutrient levels, and
components known as genetic coefficients (Loomis et al., 1979). Crop models can
potentially be used to assist in the design of crop ideotypes by maximizing the
contribution from genetic coefficients (Haverkort and Kooman, 1997). The main
benefit of this approach is that it can be used to manipulate the genetic coefficients to
test plant types not currently available to the breeder (Loomis et al., 1979). Some
estimates of genetic coefficients are actually based on observed phenotypic values, so
they may really be measuring genotype by environment (GxE) interactions and not
27
actual genetics (Baenziger et al., 2004). For models to have a useful role in
structuring a crop ideotype, they have to have input parameters that are firmly based
on real genetics. If the parameters are not based on genetics, breeding will not easily
be able to produce plants with the desired phenotypic values of those parameters,
because it is based on the manipulation of genetics (Baenziger et al., 2004).
A recent approach to crop modeling called “QTL-based ecophysiological
modeling” has replaced genetic coefficients with QTLs in crop models to incorporate
actual genetics into them (Yin et al., 2003). QTLs are regions of DNA that are
associated with one or more traits. A QTL is often thought of as a gene, but that is not
necessarily the case. A QTL can be a gene, a group of genes, or some other part of
DNA that influences the trait. The effects of individual QTLs can be highly dependent
on GxE interactions (Jansen et al., 1995). This is particularly the case for complex
traits such as yield (Campbell et al., 2003). To overcome this, QTL-based
ecophysiological modeling uses QTLs for the component traits of yield which are
likely to be less impacted by GxE interactions (Yin et al., 2000). The effects of those
QTLs on yield and how the environment influences yield are then considered in the
framework of a crop model to produce an estimate for yield. The first attempt at this
approach was made in barley with poor results, because the model lacked the power to
distinguish small changes in yield (Yin et al., 2000). However, success using the
same modeling approach for less complex traits suggests that this approach may be
successful for yield in the future (Buck-Sorlin, 2002; Yin et al., 2005).
28
3. MATERIALS AND METHODS
3.1. Plant Populations
The populations used for linkage mapping and QTL analysis were obtained by
randomly selecting heads from two F2 populations in OSU's wheat breeding program.
One population consisted of a cross between Tubbs, an awned, soft white winter wheat
variety widely grown in Oregon, and NSA 98-0995, an awnless experimental hard red
winter wheat line from the Nickerson breeding program. 280 F5 derived RILs were
generated from this population. The other population was a cross of Einstein, an
awnless, hard red winter wheat variety from Nickerson that is widely grown in
Western Europe, and Tubbs. 276 F5 derived RILs were produced from this population.
OSU's wheat breeding program produced these populations by combining two
separate crosses. Each individual cross is typically only pollinated with one spike, but
a second spike may be used when low levels of pollen are obtained. No record is kept
of the precise individuals used in these crosses; just what lines they came from. Thus,
it is possible for there to actually be up to four true populations within a nominal
population and up to six separate parents. However, only one female plant and one or
two males are typically used to create a population. To account for this, the
segregation of multi-allelic loci was examined in these populations in an attempt to
29
determine the number of true populations present in each nominal population. This
information was used to divide the RILs into “true” populations if necessary and
possible. Lines that could not be assigned to a population with confidence were
discarded.
3.2. Data Collection and Summary Statistics
The ExT and TxN RILs were grown in separate experiments with local
varieties to serve as agronomic checks and their parent lines. Since actual parents
weren't known, the parental varieties were used in their place. Each experiment was
grown at two locations with two replications in a randomized complete block design
planted in the fall of 2008. Each replication was planted in a plot that was 1.52 m
wide by 4.27 m long using about 90 grams of seed. The two locations were Corvallis,
OR and Pendleton, OR. The Corvallis plots contained six evenly spaced rows and the
Pendleton plots had seven rows. These locations represent two of the highest yielding
locations in Oregon. Each location was managed according to standard fertility and
weed control practices in their respective locations. All seed was treated with a
fungicidal seed treatment prior to planting. No subsequent fungicides or insecticides
were used.
30
Fourteen traits were measured on each plot. A list of those traits and the
methods used to measure them are presented in table 1. Each trait was analyzed
separately in each population using the general linear model procedure in the SAS
software package (The SAS Institute, Cary, NC). Environment was considered a fixed
effect and genotype was considered a random effect. Type III sum of squares were
used to test significance of the terms in the model. Least square means of RILs were
calculated on a location basis for subsequent use in determining Pearson productmoment correlation coefficients and for QTL mapping. Heritabilities for each trait
were calculated across locations on a plot mean basis using the sum of squares from
the ANOVA output as outlined by Bernardo (2002).
3.3. Linkage and QTL Mapping
Each population was screened with one STS (orw5), 123 SSR, and 6,500
DArT markers. Markers that were polymorphic were used with the phenotypic marker
for the awn repressor gene, B1, to create genetic linkage maps. The regression
mapping algorithm in JoinMap 4.0 was used to generate linkage maps (Van Ooijen
and Voorrips, 2006). Large clusters of markers were replaced with “delegate”
markers to keep from exceeding resolution power and causing unreliable mapping
(Korol et al., 2009).
31
Table 1 Measured traits.
Grain Fill Duration
Height
Ripening minus Flowering
Distance from ground to top of an average
spike in the plot. Measured no earlier than
two weeks before physiological maturity.
Yield
Weight of grain harvest in a plot divided by
area of a plot.
Test Weight
Sample of grain from a plot analyzed with an
automated test weight machine (model GAC
2100b, Dickey-John, St. Louis, MO)
Grain Protein
Sample of grain from a plot analyzed using
NIR-spectroscopy (model InfratecTM 1241
Gain Analyzer, Foss, Laurel, MD)
Spikes / m2
Measured from a 0.5 meter section of row at
Corvallis and two 0.5 meter sections at
Pendleton.
Fertile Spikelets / Spike
The average of 10 randomly chosen spikes in
a plot.
Sterile Spikelets / Spike
The average of the same 10 randomly chosen
spikes in a plot.
The 10 randomly chosen spikes were
threshed, the seed counted, and divided by
the total number of fertile spikelets on those
spikes.
Seeds / Fertile Spikelet
Seeds / Spike
The 10 randomly chosen spikes were
threshed, the seed counted, and divided by
10.
Seed Weight
The seed from the 10 randomly chosen spikes
was weighed and divided by the number of
seeds counted.
"GDD" stands for Celsius growing degree days.
[----Spike Fertility----]
GDD from planting to 50% of flag leaves
have senesced (physiological maturity).
[-----------Spike Morphology Traits-----------]
Ripening
Trait Categories
[Maturity Traits]
Method of Measurement
GDD from planting to 50% of spikes have
extruded anthers.
[---------------------Yield Components---------------------]
Trait
Flowering
32
Linkage groups were assigned to their appropriate chromosome based on
marker locations documented at the public source website Grain Genes 2.0
(http://wheat.pw.usda.gov/GG2). Multiple linkage groups that were attributed to the
same chromosome were joined on a single linkage group by relaxing criterion for
joining if possible.
QTL mapping was conducted in Windows QTL Cartographer using the
composite interval mapping procedure (Wang et al., 2010). Each location was
analyzed separately using the least square means of the RILs. Covariates were
selected using forward-reverse regression with a probability of 0.1 for adding to and
removing from the model. Covariates were excluded if they were within 10 cM of the
test site to prevent elimination of the QTL signal at that site. The significance
threshold for declaring a QTL was set to 0.05 using 1000 permutations. QTLs were
scanned for automatically with the requirement that they are at least 5 cM apart and
there is at least a 1 LOD drop between peak and trough. QTLs that appeared to be
detected due to incorrectly ordered loci in the genetic map were removed.
3.4. Factor Analysis
Factor analysis was performed on the spike morphology traits (Table 1) in the
TxN population using the factor procedure in the SAS software package (The SAS
33
Institute, Cary, NC). Each location was analyzed separately. The principal
component method was used to extract factors from the correlation matrix. The
number of factors retained was determined by looking at correlations between factors
and grain yield, test weight, and grain protein. The smallest number of factors
possible was chosen such that a factor with a significant correlation with grain yield,
test weight, or grain protein was not discarded. Extracted factors were rotated with the
varimax function to aid in interpretation. Scoring coefficients were calculated for each
factor and used to determine factor scores for the RILs. These factor scores were used
for QTL mapping. QTL mapping of factor scores was performed using the same
methods as QTL mapping of the original traits.
34
4. RESULTS AND DISCUSSION
The TxN RILs showed no marker loci with more than two alleles, so all 280
lines in the population were considered to have come from a single true population
(results not shown). Three alleles were present at some marker loci in the ExT RILs.
The segregation pattern of those alleles suggested there were two populations with a
common Einstein parent and different Tubbs parents in this cross (results not shown).
RILs from this cross were divided into two populations, ExT1 and ExT2. ExT1
included 153 RILs and ExT2 included 87 RILs. The remaining 36 lines were not
analyzed due to lack of sufficient marker data to assign them to a population, or they
had alleles that appeared to come from two different Tubbs parents at a locus. The
latter occurrence could be an indication of out-crossing in this population. Lines
lacking sufficient seed for the field trial were included in linkage mapping and not
QTL mapping. This brought the total lines for QTL mapping to 271, 142, and 81 in
the TxN, ExT1, and ExT2 populations respectively.
The accuracy of population assignments is critical to the resulting linkage and
QTL mapping procedures. Individuals incorrectly assigned to populations, could
generate outliers that bias linkage and QTL maps. For this reason, all results obtained
from these populations must be interpreted with caution. This is particularly the case
for those from the ExT populations, because a high number of lines couldn't be
35
assigned to a population. Further examination of codominate markers in these lines is
needed to split them into their appropriate populations. This would increase the sizes
of these populations and add statistical power to the subsequent linkage and QTL
mappings.
The range of phenotypic values for RILs in each population was broad and
transgressive segregates were present for all traits (Tables 2, 3, and 4). The broad
range of these values should give good statistical power to detect QTLs, if the
complexity of the trait is low. The presence of transgressive segregates indicates that
QTLs with positive contributions to these traits are present in both parents or epistatic
interactions between QTLs exist.
The majority of traits showed significant GxE interactions (Table 5). The
genotype term was significant with a probability less than 0.001 for all traits and in all
populations except for grain fill duration in the ExT2 population which has a p-value
of 0.019 (data not shown). A large number of degrees of freedom were used to test
these terms, so significant results don’t necessarily indicate the presence of a
meaningful effect. Correlations between locations and estimates of heritability may
give better indications of meaningful GxE interactions (Table 5). For example, plant
height showed significant or suggestive GxE interactions in all populations, but the
correlations between locations (0.82-0.96) and heritability estimates (0.89-0.97) were
high in those populations. This indicates that the GxE interactions are probably not
very meaningful compared to the simple effect of genotype.
36
Table 2 Least square means for parents and recombinate inbred lines
(RILs) in the Tubbs x NSA 98-01995 population (TxN).
Trait
Parents
NSA 98-0995 Tubbs
Corvallis
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
1610
2200
590
94
74.6
10.4
8.3
Spikes / m2
289
22.4
1.2
2.9
43
66
Fertile Spikelets / Spike
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
Seed Weight (mg)
Seeds / Spike
Range
RILs
Mean
Std. Err.
1589
2217
628
113
74.1
9.9
8.3
1541-1681
2173-2347
554-701
75-140
67.5-78.3
8.5-14.7
4.4-9.6
1600
2215
615
100
73.7
10.5
7.8
8
20
20
3
0.4
0.6
0.3
339
19.1
2.7
2.9
49
56
217-449
17.1-25.2
0.4-3.6
2.2-3.6
33-57
42-77
321
20.9
1.6
2.9
46
61
35
0.5
0.2
0.1
2
3
1403-1538
2017-2249
556-739
63-108
67.1-77.5
7.6-12.7
3.2-6.9
1458
2126
669
83
73.2
9.6
5.3
12
22
20
3
0.7
0.6
0.4
Pendleton
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
Spikes / m2
1469
2127
659
77
73.1
9.4
5.5
1448
2099
651
94
74.6
8.9
6.0
369
415
276-526
397
44
Fertile Spikelets / Spike
21.4
16.8
15.1-21.8 18.9
0.5
Sterile Spikelets / Spike
1.3
3.6
0.6-4.1
2.0
0.3
Seeds / Fertile Spikelet
2.6
2.8
2.1-3.2
2.7
0.2
Seed Weight (mg)
37
43
32-52
40
1
Seeds / Spike
55
47
38-64
50
4
"Std. Err." refers to the standard error of a RIL least square mean assuming no missing data.
"GDD" refers to Celsius growing degree days.
37
Table 3 Least square means for parents and recombinate inbred lines
(RILs) in the larger Einstein x Tubbs population (ExT1).
Trait
Parents
Einstein Tubbs
Corvallis
Range
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
1580
2219
639
86
74.3
10.3
9.0
1591
2227
637
112
75.2
10.1
8.7
1541-1690
2173-2457
575-776
58-160
70.7-80.1
9.3-15.1
3.3-9.8
1602
2247
645
100
75.9
11.1
7.8
8
16
17
4
0.4
0.5
0.3
Spikes / m2
322
21.2
1.9
3.1
47
66
315
19.6
2.5
3.0
48
59
228-472
17.3-24.5
0.9-4
1.4-3.5
35-58
31-75
329
20.8
2.4
2.7
46
57
36
0.6
0.3
0.2
1
4
1412-1582
2004-2317
584-795
48-123
69.4-78.3
7.5-13.7
1.6-6.6
1475
2163
688
81
74.5
9.2
4.9
14
30
29
3
0.4
0.5
0.5
Fertile Spikelets / Spike
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
Seed Weight (mg)
Seeds / Spike
ExT1 RILs
Mean Std. Err.
Pendleton
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
Spikes / m2
1484
2140
656
68
73.3
8.4
5.2
1455
2117
662
91
75.1
8.2
5.5
351
356
223-462
330
42
Fertile Spikelets / Spike
20.4
16.9
15.6-22.3 18.8
0.6
Sterile Spikelets / Spike
2.5
3.6
1.3-4.8
2.8
0.4
Seeds / Fertile Spikelet
2.5
2.8
0.9-3.2
2.5
0.2
Seed Weight (mg)
39
43
32-46
39
1
Seeds / Spike
51
47
18-63
48
3
"Std. Err." refers to the standard error of a RIL least square mean assuming no
missing data. "GDD" refers to Celsius growing degree days.
38
Table 4 Least square means for parents and recombinate inbred lines
(RILs) in the smaller Einstein x Tubbs population (ExT2).
Trait
Parents
Einstein Tubbs
Corvallis
Range
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
1580
2219
639
86
74.3
10.3
9.0
1591
2227
637
112
75.2
10.1
8.7
1561-1713
2173-2362
534-711
65-138
70.3-78.9
8.7-11.9
5.6-9.9
1605
2239
634
103
75.0
10.5
8.4
8
16
17
3
0.5
0.6
0.4
Spikes / m2
322
21.2
1.9
3.1
47
66
315
19.6
2.5
3.0
48
59
220-492
16.5-23.3
0.9-4
2.3-4
38-58
44-83
347
20.2
2.5
2.9
47
58
36
0.7
0.4
0.2
2
4
1431-1654
2012-2317
551-725
58-118
70.3-78
7.3-11.1
3.3-6.5
1490
2164
675
85
73.9
8.6
5.1
18
36
37
3
0.6
0.5
0.5
Fertile Spikelets / Spike
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
Seed Weight (mg)
Seeds / Spike
ExT2 RILs
Mean Std. Err.
Pendleton
Flowering (GDD)
Ripening (GDD)
Grain Fill Duration (GDD)
Plant Height (cm)
Test Weight (kg/hL)
Grain Protein (% on dry basis)
Yield (metric tons/ha)
Spikes / m2
1484
2140
656
68
73.3
8.4
5.2
1455
2117
662
91
75.1
8.2
5.5
351
356
225-434
331
47
Fertile Spikelets / Spike
20.4
16.9
15.8-21.4 18.8
0.5
Sterile Spikelets / Spike
2.5
3.6
1.2-5.1
3.1
0.3
Seeds / Fertile Spikelet
2.5
2.8
1.7-3.4
2.7
0.1
Seed Weight (mg)
39
43
32-47
40
1
Seeds / Spike
51
47
32-70
50
3
"Std. Err." refers to the standard error of a RIL least square mean assuming no
missing data. "GDD" refers to Celsius growing degree days.
***
*
*
ns
***
**
***
Flowering
Ripening
Grain Fill Duration
Plant Height
Test Weight
Grain Protein
Yield
0.72***
0.60***
0.38***
0.82***
0.74***
0.29***
0.45***
0.98
0.89
0.80
0.93
0.92
0.71
0.93
Correlation R2
2
0.84
0.75
0.54
0.89
0.85
0.44
0.60
h
***
***
**
***
***
***
***
GxE
0.72***
0.61***
0.31***
0.95***
0.83***
0.55***
0.58***
0.98
0.88
0.76
0.97
0.96
0.88
0.95
Correlation R2
0.82
0.73
0.45
0.96
0.90
0.71
0.72
2
h
Larger Einstein x Tubbs Population
(ExT1)
**
ns
ns
*
***
ns
**
GxE
0.64***
0.58***
0.12
0.96***
0.77***
0.26*
0.54***
0.96
0.80
0.65
0.97
0.93
0.83
0.95
Correlation R2
0.78
0.74
0.28
0.97
0.87
0.52
0.69
2
h
Smaller Einstein x Tubbs Population
(ExT2)
Spikes / m
ns
0.28***
0.69
0.44 ns
0.32***
0.65
0.45 ns
0.24*
0.67
0.47
Fertile Spikelets / Spike
***
0.69***
0.90
0.80 **
0.76***
0.90
0.85 ns
0.76***
0.87
0.87
Sterile Spikelets / Spike
**
0.65***
0.81
0.79 **
0.60***
0.81
0.75 **
0.65***
0.84
0.78
Seeds / Fertile Spikelet
***
0.43***
0.78
0.60 ***
0.48***
0.85
0.68 ns
0.58***
0.80
0.74
Seed Weight
***
0.67***
0.91
0.78 ***
0.73***
0.94
0.82 ns
0.60***
0.92
0.70
Seeds / Spike
***
0.51***
0.84
0.67 ***
0.45***
0.88
0.63 ***
0.53***
0.82
0.78
"GxE" is the probability of a genotype by environment interaction from the ANOVA. *, **, and *** represent 0.05, 0.01, and 0.001 significance
levels, respectively. "ns" stands for not significant. "Correlation" is the Pearson product-moment correlations between RIL means at Pendleton and
Corvallis. "R2" is the coefficient of determination from the ANOVA. "h2" is the across location narrow sense heritability estimates for the RILs on a
plot mean basis.
2
GxE
Trait
Tubbs x NSA 98-0995 Population
(TxN)
Table 5 Summary of across location analyzes for the three populations of recombinate inbred lines (RILs).
39
40
Under these conditions, QTLs detected at one location should have a high
degree of correspondence with QTLs detected at the other location. Low correlations
and heritability estimates with a lack of a significant GxE interaction indicates that a
trait was measured with low precision. A low coefficient of determination also
indicates this (Table 5). Spikes per m2 is an examples of such a trait. QTL mapping
for these traits is unreliable due to a lack of statistical power. The relatively low
heritabilties and coefficients of determination for grain protein are likely due to
environmental variability within each test location.
Correlations between traits are presented in tables 6, 7, and 8 for the TxN,
ExT1, and ExT2 populations respectively. The relationships between these traits may
not be strictly linear, so the correlation coefficients won’t accurately represent the
relatedness between these traits. That is because correlation coefficients make the
assumption that traits are linearly related. Even if the relationship is linear, the simple
correlations between traits may not fully represent relatedness between the traits
(Fonseca and Patterson, 1968).
The relationship between height and yield in the ExT populations does not
appear to be linear. Biplots of the traits suggest an increase in yield as height increase
until the highest values of height (results not shown). At these higher values, there
appears to be a decrease in yield. Thus, the height to yield correlations in the ExT
populations can be misleading.
-0.23***
0.12*
-0.24***
Height
Test Weight
Grain Protein
Yield
-0.17**
0.01
-0.22***
0.11
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
Seed Weight
Seeds / Spike
0.18**
-0.24***
0.11
-0.16**
0.18**
-0.09
-0.13*
0.21***
-0.10
0.11
0.72***
0.67***
0.13*
-0.08
0.12*
-0.04
0.06
-0.00
0.04
0.15*
0.07
0.19**
0.68***
-0.09
-0.09
0.23***
-0.10
0.22***
-0.00
0.01
0.17**
0.03
0.06
0.24***
0.10
-0.11
Height
*, **, and *** represent 0.05, 0.01, and 0.001 significance levels, respectively
0.19**
Spikes / m
Fertile Spikelets / Spike
-0.16*
-0.09
Grain Fill Duration
2
0.58***
-0.13*
Ripening
Flowering
Flowering Ripening
Grain Fill
Duration
-0.04
0.07
0.06
0.25***
-0.15*
-0.13*
-0.14*
0.30***
0.29***
0.13*
-0.16**
-0.35***
Test
Weight
-0.10
-0.01
-0.00
0.03
-0.18**
-0.03
-0.20**
0.33***
0.03
0.30***
0.16**
-0.09
Grain
Protein
0.25***
0.16**
0.21***
0.04
0.15*
0.13*
-0.19**
0.02
0.35***
0.09
-0.06
-0.17**
Yield
-0.20***
-0.22***
-0.17**
-0.13*
-0.13*
0.05
0.15*
0.05
-0.08
0.11
0.06
-0.03
m
2
Spikes /
0.60***
-0.09
0.08
-0.40***
-0.11
0.09
-0.12
-0.22***
0.01
0.09
0.35***
0.39***
-0.39***
0.40***
-0.22***
-0.15*
-0.07
-0.14*
-0.14*
0.14*
0.11
-0.10
0.03
0.14*
0.85***
-0.19**
-0.29***
-0.03
-0.22***
0.23***
-0.24***
-0.10
0.02
-0.09
-0.24***
-0.24***
Fertile
Sterile
Seeds /
Spikelets / Spikelets / Fertile
Spike
Spike
Spikelet
-0.20***
-0.13*
0.10
-0.33***
-0.20***
0.18**
0.18**
0.34***
0.19**
-0.04
-0.25***
-0.30***
Seed
Weight
-0.30***
0.80***
-0.33***
0.57***
-0.24***
0.24***
-0.27***
-0.22***
0.01
-0.02
0.00
0.02
Seeds /
Spike
Table 6 Pearson product-moment correlations for recombinate inbred lines in the Tubbs x NSA 98-0995 population (TxN).
Corvallis correlations are above the diagonal and Pendleton correlations are below the diagonal.
41
-0.11
-0.35***
Test Weight
Yield
-0.22**
Seeds / Spike
-0.19*
-0.04
-0.24**
-0.20*
0.06
0.42***
-0.02
-0.37***
-0.25**
-0.43***
0.83***
0.70***
-0.09
-0.07
-0.12
-0.10
0.03
0.32***
0.01
-0.24**
-0.23**
-0.31***
0.63***
-0.11
0.03
0.28***
0.18*
0.30***
-0.25**
-0.26**
-0.16
0.37***
0.41***
-0.28***
-0.38***
-0.23**
Height
*, **, and *** represent 0.05, 0.01, and 0.001 significance levels, respectively
-0.25**
0.01
Seed Weight
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
0.07
-0.23**
Fertile Spikelets / Spike
0.33***
Grain Protein
-0.05
-0.34***
Height
2
0.11
Grain Fill Duration
Spikes / m
0.64***
Ripening
Flowering
Flowering Ripening
Grain Fill
Duration
-0.32***
0.16
-0.21*
0.15
-0.20*
0.24**
-0.19*
-0.16
0.58***
-0.09
-0.18*
-0.15
Test
Weight
0.49***
0.05
0.53***
0.22**
-0.06
-0.55***
0.29***
0.16
0.41**
-0.09
-0.38***
-0.41***
Yield
0.04
-0.31***
0.07
-0.03
-0.06
-0.19*
0.09
0.03
-0.08
0.13
0.01
-0.11
m
2
Spikes /
-0.48***
0.04
-0.50***
-0.11
0.03
-0.04
-0.49***
0.26**
-0.07
0.38***
0.41***
0.18*
Grain
Protein
0.24**
-0.09
-0.26**
-0.28***
-0.03
-0.20*
0.04
-0.24**
-0.08
0.11
0.09
0.01
-0.27**
0.41***
-0.13
-0.25**
-0.12
0.06
0.09
0.36***
0.32***
-0.08
-0.02
0.04
0.86***
-0.25**
-0.18*
-0.14
-0.48***
-0.08
0.58***
-0.10
0.14
-0.20*
-0.46***
-0.40***
Fertile
Sterile
Seeds /
Spikelets / Spikelets / Fertile
Spike
Spike
Spikelet
-0.30***
-0.14
0.22**
-0.11
0.07
-0.34***
0.01
0.27***
0.37***
-0.25**
-0.09
0.11
Seed
Weight
-0.17*
0.84***
-0.29***
0.42***
-0.48***
-0.19*
0.56***
-0.22**
0.09
-0.14
-0.38***
-0.37***
Seeds /
Spike
Table 7 Pearson product-moment correlations for recombinate inbred lines in the larger Einstein x Tubbs population (ExT1).
Corvallis correlations are above the diagonal and Pendleton correlations are below the diagonal.
42
-0.37***
-0.15
-0.40***
Height
Test Weight
Yield
0.14
0.00
-0.08
0.27*
0.19
-0.15
-0.07
-0.15
0.05
Grain Protein
Fertile Spikelets / Spike
Sterile Spikelets / Spike
Seeds / Fertile Spikelet
Seed Weight
Seeds / Spike
-0.16
0.09
0.08
0.08
0.02
0.05
-0.10
-0.20
-0.07
-0.10
0.06
0.64***
-0.20
0.21
-0.09
0.30**
-0.17
0.12
-0.12
0.22
0.46***
-0.04
-0.26*
-0.29**
Height
*, **, and *** represent 0.05, 0.01, and 0.001 significance levels, respectively
0.12
-0.11
0.22
-0.20
-0.39***
-0.21
-0.25*
0.64***
0.66***
-0.05
2
-0.20
Grain Fill Duration
Spikes / m
0.61***
Ripening
Flowering
Flowering Ripening
Grain Fill
Duration
-0.25*
0.26*
-0.22
0.27*
-0.12
0.48***
-0.13
-0.02
0.64***
0.05
-0.22
-0.33**
Test
Weight
0.25*
0.02
0.34**
0.05
-0.06
-0.28*
0.36***
0.07
0.09
0.28*
-0.00
-0.28*
Yield
2
-0.10
-0.04
0.01
0.01
-0.21
-0.13
0.19
-0.10
-0.06
0.19
0.11
-0.04
m
Spikes /
-0.26*
0.07
-0.34**
0.06
0.03
-0.01
-0.08
0.37***
0.26*
0.37***
0.24*
-0.05
Grain
Protein
0.51***
-0.40***
-0.03
-0.45***
0.04
-0.11
-0.17
-0.16
-0.09
0.18
0.18
0.06
-0.45***
0.24*
-0.25*
-0.32**
0.26*
0.10
0.06
0.37***
0.28*
0.08
0.02
-0.05
0.84***
-0.07
-0.12
-0.29**
-0.24*
-0.13
0.35**
-0.23*
-0.05
-0.16
-0.22*
-0.13
Fertile
Sterile
Seeds /
Spikelets / Spikelets / Fertile
Spike
Spike
Spikelet
-0.28*
0.08
0.06
-0.24*
-0.14
-0.34**
0.22
0.26*
0.45***
-0.11
-0.13
-0.06
Seed
Weight
Table 8 Pearson product-moment correlations for recombinate inbred lines in the smaller Einstein x Tubbs population
(ExT2). Corvallis correlations are above the diagonal and Pendleton correlations are below the diagonal.
-0.09
0.74***
-0.33**
0.41***
-0.20
-0.22
0.21
-0.33**
-0.12
-0.02
-0.08
-0.09
Seeds /
Spike
43
44
All other traits appear to be at least close to linearly related over the ranges
measured (not shown). However, it should not be assumed that these relationships
will occur beyond the ranges measured.
The correlations between the spike fertility traits, seeds per fertile spikelet and
seeds per spike, were consistently the strongest across all locations and populations.
Mathematically, fertile spikelets per spike multiplied by seeds per fertile spikelet
equals seeds per spike. Thus, it is expected that the spike fertility traits are under vary
similar genetic control. The correlations between other traits depended to a large
degree on the population and the location where they were measured. Thus, there may
be different genes segregating in the different populations and/or the presence of gene
by environment interactions. There is a lack of consistently strong correlations
between yield and any of the other traits in all populations at every location (Tables 6,
7, and 8). Generally, yield is correlated with the spike fertility traits and inversely
correlated with flowering. In the TxN population no trait has a correlation with yield
stronger than 0.35. However, correlations greater than 0.5 for the spike fertility traits
with yield are present in the ExT1 population. This lack of consistently strong
correlations makes identifying traits to manipulate for increased yield difficult.
Ultimately, selection of QTLs for these traits will require considerable knowledge of
the genetic backgrounds that they are in.
45
4.1. Genetic Linkage Maps
The TxN population mapped 373 markers on 21 linkage groups. This map
covered 1387.6 cM, 1245.5 cM and 1163 cM for the A, B, and D genomes
respectively. The ExT1 population mapped 353 markers on 22 linkage groups
covering 1318.6 cM, 1552.5 cM, and 1444.7 cM for the A, B, and D genomes. The
ExT2 population mapped 358 markers on 22 linkage groups covering 1548.9 cM,
1423.9 cM, and 955.2 cM for the three genomes. All chromosomes were represented
in each population. Relaxing criterion for merging linkage groups from the same
chromosome resulted in only two gaps in all three maps. However, there were 57
marker intervals that were greater than 50 cM. Seven of those intervals were over 100
cM. Any QTLs detected in these intervals need to be interpreted with care, because
the markers are nearly unlinked at these distances.
4.1. QTL Mapping
Multiple QTLs were detected for all traits for a total of 146 QTLs (Tables 9,
10, 11, 12, and 13). Grain protein (Table 9), seed weight (Table 13), and spikes per
m2 (Table 9) had at least one population that failed to identify QTLs for that trait.
46
Table 9 QTLs for yield, grain protein, test weight, and spikes per m2
Trait
Yield
Grain Protein
Test Weight
2
Spikes / m
Pop. Chrom. Loc.
ExT1 4D
Corv
7A
Corv
ExT2 2B
Pend
TxN 2B
Pend
ExT2 3D
Pend
6B
Pend
TxN 4A
Corv
ExT1 4A
Corv
4A
Pend
4B
Corv
5A-2
Corv
5B
Corv
6B
Corv
ExT2 4D
Corv
5A
Pend
TxN 5A
Corv
5A
Pend
5A
Pend
6B
Pend
7D
Pend
Nearest Marker
wPt-0472cluster
wPt-0031
wPt-2110
wPt-2327cluster
wPt-6066
381509cluster
wmc617
wPt-4544
wPt-1007cluster
barc90cluster
wPt-7061
348314cluster
wPt-6441
wPt-3058
gwm291
B1
barc330
B1
barc79cluster
cfd14
Position LOD
(cM)
Score Add. E.
20.8
3.296 0.31682
20.5
4.84 -0.3644
184.1
3.387 0.2889
90.9
4.342 0.17337
114
5.012 0.398
48.7
3.179 -0.2728
14.9
3.188 -0.2622
93.5
3.877 1.14463
151.7
6.197 0.72838
138.9
4.054 -0.58725
4.2
3.64 0.57225
74.6
3.588 0.51
216
3.623 1.13775
0
4.686 0.815
155.3
3.669 1.24788
136.4
3.356 0.41863
42.2
3.622 0.47363
134.4
7.617 0.63588
80.1
4.932 0.49213
170.3
3.882 -0.631
R2
0.1
0.13
0.16
0.07
0.23
0.12
0.08
0.28
0.14
0.08
0.08
0.07
0.31
0.15
0.18
0.05
0.06
0.11
0.06
0.1
ExT2 3A
Corv wPt-0714
184.8
3.743 23.7785 0.15
3D
Pend gdm72
95.2
4.952 25.5167 0.24
4B
Corv barc90cluster
150.4
3.315 -21.45
0.12
5D-1
Corv wPt-1197
89.5
3.749 34.8061 0.34
7A
Pend wPt-3425
251.1
4.144 -28.7746 0.18
TxN 1A
Corv wmc278
96.1
5.102 13.3076 0.09
1A
Pend wPt-4765cluster 116
6.092 24.3792 0.28
"Pop." refers to mapping population. "TxN" refers to Tubbs x NSA 98-0995 population.
"ExT1" and "ExT2" refer to the large and small Einstein by Tubbs populations respectively.
"Chom." is chromosome. "Loc." is location. "Add. E." is the additive effect for replacing an
allele from a Nickerson parent with an allele from Tubbs. "R2" is the coefficient of
determination for how the QTL explains the residuals after accounting for the covariates.
47
Table 10 QTLs for flowering, ripening, and grain fill duration.
Position LOD
Score Add. E. R2
Pop. Chrom. Loc. Nearest Marker (cM)
ExT1 1B
Corv wPt-6690
104.9
3.48 8.6361 0.07
3A
Corv tPt-1143
166.8
4.897 11.9123 0.12
3A
Pend tPt-1143
164.8
7.312 12.9231 0.18
4D
Corv wPt-0472cluster 20.8
5.12 -11.3638 0.13
5B
Corv barc4cluster
56.7
5.989 -12.2117 0.13
ExT2 5A
Pend wPt-4419
56.6
2.799 -11.0962 0.11
TxN 5B
Corv wPt-1895cluster 122.7
5.396 -7.4737 0.08
5B
Pend wPt-1895cluster 122.7
7.742 -11.4246 0.11
7A
Corv wPt-1928
99.1
7.018 -8.4414 0.1
Ripening
ExT1 1B
Corv wPt-6690
104.3
4.751 12.7814 0.09
3A
Corv gwm666
168.8
3.957 13.5863 0.1
3B
Corv 381874cluster
154.6
4.153 11.5948 0.08
ExT2 4B
Corv tPt-0602
154.8
3.905 -14.7623 0.12
4B
Pend gwm251
158.8
4.106 -19.7188 0.18
4D
Corv wPt-3058
0
4.628 -15.0344 0.14
6B
Corv barc101cluster
55.8
3.233 11.748 0.09
TxN 4A
Corv wPt-7924
66.9
3.473 13.4196 0.14
4A
Pend wPt-7924
62.9
3.853 15.0062 0.16
5B
Corv 303931cluster
119.2
4.501 -12.0272 0.08
5B
Corv wPt-9467
77.8
4.581 14.0095 0.07
6A
Pend wPt-7063
74.7
3.857 9.3901 0.06
Grain Fill Duration ExT1 1A
Pend wPt-3904
31
5.129 16.0089 0.14
ExT2 1B
Pend barc8cluster
35.1
3.459 15.7915 0.15
4B
Pend gwm251
161.2
3.746 -15.1368 0.16
5B
Corv wPt-1881
100.1
4.277 -11.3228 0.15
TxN 2A
Corv gwm312
145.5
4.635 7.2015 0.07
2B
Corv wPt-8760
211.1
3.711 6.1693 0.05
3B
Pend wPt-3107
110.3
5.293 9.4677 0.1
5B
Corv 311648cluster
248.8
3.167 5.776
0.04
7D
Pend cfd14
166
4.17 -9.2513 0.09
"Pop." refers to mapping population. "TxN" refers to Tubbs x NSA 98-0995 population.
"ExT1" and "ExT2" refer to the large and small Einstein by Tubbs populations respectively.
"Chom." is chromosome. "Loc." is location. "Add. E." is the additive effect for replacing an
allele from a Nickerson parent with an allele from Tubbs. "R2" is the coefficient of determination
for how the QTL explains the residuals after accounting for the covariates.
Trait
Flowering
48
Table 11 QTLs for plant height and seeds per spike.
Position LOD
Score Add. E. R2
Pop. Chrom. Loc. Nearest Marker (cM)
ExT1 1D
Pend barc229cluster
66
3.42 3.5724 0.05
2A
Pend wPt-7024
212.8
3.182 3.4297 0.04
4B
Corv barc20cluster
143
15.44 -11.6664 0.27
4B
Pend barc20cluster
143.6
17
-9.6777 0.31
4D
Corv wPt-0472cluster 20.8
24.08 16.2917 0.52
4D
Pend wPt-0472cluster 18.8
26.89 13.904 0.64
5A-1
Corv wPt-4419
102.4
3.483 4.4523 0.05
ExT2 2B
Corv wmc175
234
5.241 6.8583 0.13
2B
Pend wmc175
234
5.437 5.6995 0.13
4B
Corv barc20cluster
146.2
8.395 -9.3857 0.2
4B
Pend barc20cluster
146.2
7.61 -7.1513 0.18
4D
Corv wPt-0472cluster 14.7
9.64 11.3276 0.3
4D
Pend wPt-0472cluster 14.7
8.467 9.0307 0.29
7A
Corv wPt-0961cluster 243.8
5.826 6.9774 0.15
7A
Pend wPt-0961cluster 245.2
6.142 5.4881 0.14
7B
Corv tPt-8569cluster 26.2
4.315 5.7242 0.1
TxN 2A
Pend gwm515
98.2
3.759 2.0655 0.09
4B
Corv gwm513
36.2
3.868 -2.0761 0.06
4B
Pend gwm513
36.2
6.203 -2.1745 0.1
5A
Corv barc330
40.2
5.992 2.6729 0.1
5A
Pend barc330
38.2
4.531 1.9373 0.08
5B
Pend 305092cluster
124.2
4.159 1.6379 0.06
Seeds per Spike
ExT1 2A
Corv wPt-1142cluster 171.8
6.066 -3.2037 0.15
3B
Corv wPt-3041
7.7
5.428 3.4932 0.18
5A-1
Corv barc186
56.2
3.298 2.498
0.09
ExT2 3A
Pend wPt-3697cluster 73.4
3.015 2.4179 0.11
4A
Pend wPt-2301
199.5
3.142 -2.2382 0.1
4B
Pend wPt-0391
179.6
6.654 4.3063 0.39
TxN 2A
Corv gwm372cluster 110.7
12.37 -2.8105 0.18
2A
Pend wPt-1142cluster 115.8
11.01 -2.1352 0.17
4A
Corv rPt-2478cluster 116
4.39 1.5737 0.05
5B
Corv 310624
71.3
3.078 -1.298
0.04
"Pop." refers to mapping population. "TxN" refers to Tubbs x NSA 98-0995 population.
"ExT1" and "ExT2" refer to the large and small Einstein by Tubbs populations respectively.
"Chom." is chromosome. "Loc." is location. "Add. E." is the additive effect for replacing an
allele from a Nickerson parent with an allele from Tubbs. "R2" is the coefficient of
determination for how the QTL explains the residuals after accounting for the covariates.
Trait
Plant Height
49
Table 12 QTLs for fertile spikelets per spike and sterile spikelets per spike.
Position LOD
Score Add. E. R2
Pop. Chrom. Loc. Nearest Marker (cM)
ExT1 2A
Corv gwm372cluster 175.2
6.7
-0.5348 0.11
2A
Pend wPt-1142cluster 173.8
6.864 -0.4372 0.12
4A
Corv wPt-6757
55.5
5.307 -0.473
0.08
4B
Pend gwm149
5.8
3.54 -0.413
0.09
5A-2
Pend wPt-1200
0
3.591 -0.3091 0.06
7A
Corv wPt-5987
190.5
16.26 0.8717 0.29
7A
Pend wPt-5987
191.5
14.12 0.6934 0.29
7B
Corv gwm111
136.8
6.445 -0.5216 0.1
ExT2 5A
Corv gwm666
16.3
5.252 -0.7029 0.2
7A
Corv wPt-5987
225.8
13.3 1.01
0.4
7B
Corv gwm111
83.2
4.237 -0.5263 0.11
TxN 2A
Corv 348413cluster
117.8
9.019 -0.4952 0.13
2A
Pend gwm95
108.4
8.312 -0.363
0.12
3B
Corv wPt-0280cluster 179.8
3.198 0.2744 0.04
4B
Pend gwm251
41.5
3.925 -0.2389 0.05
5A
Pend B1
136.4
4.573 -0.2656 0.06
5B
Corv wPt-2273cluster 93
4.018 -0.3364 0.06
5B
Pend wPt-2273cluster 90.1
4.184 -0.2404 0.05
Sterile Spikelets
ExT1 2A
Pend wPt-0094cluster 205.5
4.108 0.2084 0.09
per Spike
5A-2
Corv wPt-7061
4.2
5.097 0.2326 0.14
5A-2
Pend wPt-7061
4.2
7.46 0.3165 0.2
ExT2 2D
Pend wPt-9580
5.9
5.248 0.3206 0.14
3A
Pend wmc11
214.4
3.959 -0.2741 0.11
4B
Corv rPt-6847
167.6
5.42 -0.2985 0.21
4B
Pend rPt-6847
171.6
5.514 -0.4274 0.25
5A
Corv B1
156.7
6.631 0.2995 0.23
5A
Pend B1
156.7
6.206 0.3669 0.19
TxN 2D
Corv cfd51cluster
8.2
5.667 -0.1537 0.08
2D
Pend gwm296
2.2
3.797 -0.1414 0.05
4D
Corv wmc473cluster
11
7.313 0.1861 0.11
4D
Pend wmc473cluster
13
10.78 0.2503 0.17
5A
Corv B1
134.4
11.01 0.2221 0.17
5A
Pend B1
134.4
9.663 0.2097 0.12
"Pop." refers to mapping population. "TxN" refers to Tubbs x NSA 98-0995 population.
"ExT1" and "ExT2" refer to the large and small Einstein by Tubbs populations respectively.
"Chom." is chromosome. "Loc." is location. "Add. E." is the additive effect for replacing an
allele from a Nickerson parent with an allele from Tubbs. "R2" is the coefficient of
determination for how the QTL explains the residuals after accounting for the covariates.
Trait
Fertile Spikelets
per Spike
50
Table 13 QTLs for seeds per fertile spikelet and seed weight.
Position LOD
Score Add. E. R2
Pop. Chrom. Loc. Nearest Marker (cM)
ExT1 3B
Corv wPt-3041
11.7
6.021 0.1531 0.17
4A
Pend wPt-1007cluster 153.7
4.725 -0.1164 0.12
5A-1
Corv barc186
56.2
3.212 0.1082 0.09
ExT2 4A
Pend wPt-6900
198.3
5.252 -0.129
0.17
4B
Pend wPt-0391
179.6
3.925 0.1443 0.21
6A
Corv tPt-6278
135.3
3.269 -0.1058 0.12
6A
Pend tPt-6278
135.3
3.254 -0.1005 0.1
TxN 2A
Corv gwm372cluster 110.7
3.268 -0.0609 0.06
2A
Pend wPt-1142cluster 112.7
3.823 -0.053
0.06
3A
Pend gwm2
161.9
3.864 -0.0604 0.07
5A
Pend wPt-4203cluster 138.4
4.213 0.0553 0.06
Seed Weight
ExT1 1A
Pend barc83cluster
58.8
3.451 -0.8668 0.08
5A-2
Pend wPt-7061
2.2
3.416 0.822
0.07
7A
Corv gwm332
202.4
5.306 -1.7144 0.17
7B
Corv wPt-1266
132.6
4.423 1.3819 0.11
7B
Pend wPt-4258cluster 128.2
6.272 1.1377 0.14
TxN 2B
Corv wPt-4417cluster 115.7
3.356 0.8225 0.05
4A
Corv 311358cluster
98.7
7.414 -1.2428 0.1
4D
Corv wmc331
15
7.699 1.5327 0.16
4D
Pend wmc331
15
5.61 0.8952 0.1
5A
Pend gwm291
130.4
3.487 0.5972 0.04
5B
Corv wPt-3457cluster 137.2
6.937 1.5656 0.16
5B
Corv barc4cluster
67.5
4.122 0.8847 0.05
5B
Pend wPt-3329
149.2
5.66 1.2411 0.19
"Pop." refers to mapping population. "TxN" refers to Tubbs x NSA 98-0995 population.
"ExT1" and "ExT2" refer to the large and small Einstein by Tubbs populations respectively.
"Chom." is chromosome. "Loc." is location. "Add. E." is the additive effect for replacing an
allele from a Nickerson parent with an allele from Tubbs. "R2" is the coefficient of
determination for how the QTL explains the residuals after accounting for the covariates.
Trait
Seeds per
Fertile Spikelet
51
Significant differences existed between RILs for grain protein, seed weight and
spikes per m2. Thus, the failure to detect QTLs for these traits is likely the result of a
lack of statistical power. For spikes per m2, the lack of statistical power is probably
the result of a lack of precision in the measurement of that trait. The relatively few
(81) RILs mapped in the ExT2 population could explain the failure to detect QTLs for
seed weight in that population. A smaller number of lines used in QTL mapping
results in less statistical power (Zeng, 1994). It is not immediately clear why QTLs
weren’t mapped for grain protein in the ExT1 population. Grain protein had higher
heritability in the ExT1 population than it did in the other two populations (Table 5).
Since power to detect QTLs is inversely proportional to heritability, this should have
been the population that had the greatest power to detect QTLs for grain protein
(Beavis, 1994). The best explanation for the failure to detect these QTLs is that
genetic control of grain protein is complex or that it is more heavily influenced by
environment than genetics.
QTLs for a trait that occur at nearly the same position on the genetic map for
both locations likely represent a single QTL that is effective across those locations
(Jiang and Zeng, 1995). These QTLs are useful to breeders, because they can be
selected for with minimal consideration of GxE interactions. QTLs of this type were
detected for all traits except for grain fill duration, grain protein, and yield. The lack
of across location QTLs detected for grain fill duration may be attributed to the high
standard error of the RIL means relative to their spread (Tables 2, 3, and 4). This high
52
relative standard error indicates that the measured values for this trait are too
inaccurate for consistent QTL mapping. Grain protein is highly influenced by
environmental factors (Gonzalez-Hernandez et al., 2004). Failure to account for in
field variability of those environmental factors could be limiting the ability to detect
QTLs for grain protein. For yield, the lack of across environment QTLs is probably
attributable to GxE interactions and complex genetic control of the trait. The genetic
control of this trait may poorly fit the additive genetic model that is assumed by QTL
analysis using CIM (Zeng, 1994). Techniques that allow for epistatic interactions,
such as MIM, could have greater power to detect QTLs for yield (Kao et al., 1999).
However, an overall lack of statistical power to detect sufficient quantities of QTLs to
explain a meaningful percent of variation for yield will probably still exist.
There are several instances of QTLs occurring on the same chromosome for a
particular trait in more than one population. This may indicate that a single QTL
occurs in more than one population. If this is the case, it would provide some measure
of conformation for that QTL. Such an assumption should be made with great care,
because of the large differences between linkage maps of different populations and the
sparseness of those maps. However, these finding at least show that one chromosome
is important to a trait in more than one population. Examples of this include 7A for
fertile spikelets per spike in ExT1 and ExT2 populations and 2D for sterile spikelets
per spike in the ExT2 and TxN populations (Table 12). Finer mapping is needed to
determine if only one QTL is actually present in those populations.
53
There are four loci that show close linkage or pleiotropy with 36% of the
detected QTLs. These loci are Rht-B1, Rht-D1, B1, and Xgwm372 on chromosomes
4B, 4D, 5A, and 2A respectively. The large numbers of QTLs found at these loci
indicates that they are strong determinants of the variability present in these
populations and further examination is warranted.
Rht-B1 and Rht-D1 are the loci for the two major dwarfing genes in these
populations. Since the ExT populations are segregating for dwarfing alleles at Rht-B1
and Rht-D1, the strong QTLs for height on 4B and 4D can be explained by the
dwarfing genes (Table 11). The TxN population is fixed at these loci, so the QTL for
height on 4B in that population represents a minor QTL for height. This QTL appears
to be closed linked to Rht-B1, because its nearest marker, gwm513, maps closely to the
QTLs for Rht-B1 in both ExT populations. The height QTLs account for 10 of the 28
QTLs detected in close proximity to these loci. Pleiotropy may explain the occurrence
of many of the other QTLs. If the other QTLs are not due to pleiotropy, it may be
possible to break unfavorable linkages between QTLs. Finer mapping of these
chromosomes combined with multiple-trait QTL mapping would be able to better
position the QTLs on these chromosomes and be able to distinguish between
pleiotropy and close linkage (Jiang and Zeng, 1995). This is needed to determine how
useful the QTLs detected at these loci are to breeders. Pleiotropic effects are probably
not very useful, unless they indicate that one dwarfing gene is more favorable than the
other.
54
The B1 locus is the location of the awn inhibitor gene that is segregating in all
of the populations. A total of 15 QTLs showed pleiotropy or close linkage with this
gene. The awned allele is associated with more sterile spikelets and fewer fertile
spikelets. For sterile spikelets per spike there are QTLs at both locations for all three
populations at the B1 locus (Table 12). QTLs for fertile spikelets per spike are present
at Corvallis for the TxN and ExT1 populations and Pendleton for the ExT2 population
near the B1 locus (Table 12). Based on QTLs detected for seed weight (Table 13) and
test weight (Table 9) close to B1, it appears that awned varieties have higher seed
weights and greater test weights. Laperche et al. (2007) also detected QTLs for seed
weight at the B1 locus. However, they found greater seed weight for awned varieties
at low soil nitrogen levels and greater seed weight for awnless varieties at high soil
nitrogen levels. Thus, the effects of these QTLs appear to depend on production and
environmental practices.
As with the Rht-B1 and Rht-D1 loci, finer linkage mapping and multiple-trait
QTL mapping can be used to find more precise positions of these QTLs and test for
pleiotropic effects (Jiang and Zeng, 1995). If the QTLs prove to be pleiotropic, then it
appears that there is a trade-off of number of fertile spikelets for greater seed weight
when choosing awned varieties under the conditions studied. These findings indicate
that selection of an awned or awnless variety is not trivial, but it is not obvious if one
type is universally preferred over the other. A preference for awned wheat has been
shown in studies in the Great Plains, attributed to awns the available photosynthetic
55
area of a variety (Donald, 1968). However, it has been suggested that the awnless trait
makes varieties less prone to pre-harvest sprouting and thus preferable in regions
prone to pre-harvest sprouting (King and Richards, 1984). Further examination is
needed to identify which plant type is preferable for Oregon.
Another 8 QTLs were detected on 5A in the general vicinity of Xbarc330.
When combined with the QTLs detected at B1, at least one QTL was mapped on
chromosome 5A for sterile spikelets per spike, fertile spikelets per spike, flowering,
height, seed weight, seeds per fertile spikelet, seeds per spike, and test weight. Other
researchers have also observed QTLs for yield and yield components on chromosome
5A (Kato et al., 1998). This indicates that 5A is a good candidate for further research.
Important genes that have been mapped to this chromosome include Vrn-A1, a gene
that controls vernalization requirement; and Rht12, a dwarfing gene that is thought to
be tightly linked with B1 (Worland et al., 2006). Nothing is known about the allelic
variation of these loci in the populations.
Thirteen QTLs were detected near the Xgwm372 locus in the TxN and ExT1
populations. It is not immediately apparent what gene or genes are responsible for
these QTLs. QTLs found near this locus include those for seeds per fertile spikelet
(Table 13), seeds per spike (Table 11), and fertile spikelets per spike (Table 12).
These spike fertility traits are the yield components that are most consistently
correlated with yield (Tables 6, 7, and 8), so these QTLs might be useful for
increasing yield. It appears that all of the favorable alleles are coming from the
56
Nickerson parent in their respective populations. Thus, it is only of minor
consequence if the QTLs are pleiotropic or closely linked. Selecting for the favorable
allele at the Xgwm372 locus could provide a bump to yield. However, no QTLs for
yield were detected at this locus. Thus, a better understanding of the interactions with
this locus and others is probably needed to maximize any potential gain.
4.2. Factor Analysis
Four factors were retained for the Corvallis spike morphology traits and three
factors were retained for the Pendleton traits based on significant correlations with
yield, test weight, or grain protein (Table 14). This criterion for selecting the number
of factors was used to guard against throwing out potentially informative factors.
Selection of factors based on the proportion of variation explained can result in
disregarding a factor that only explains a small portion of the total variation, but is
important for explaining the economically important traits: yield, test weight, and
protein. The significant correlation criterion is quite conservative, because the many
degrees of freedom used to test correlations can identify correlations as significant that
may not be very meaningful.
57
Table 14 Pearson product-moment
correlations between initial factors and yield,
grain protein, and test weight.
Factor
Yield
Corvallis
Test weight Protein
F1
F2
F3
F4
F5
-0.22***
0.04
-0.30***
-0.01
0.06
0.19**
0.20**
-0.04
0.22***
-0.03
-0.28***
0.19**
0.01
0.15*
0.09
Pendleton
F1
0.16**
-0.09
-0.12
F2
0.27*** -0.00
0.20***
F3
0.12*
-0.15*
-0.12
F4
0.01
-0.08
0.12
F5
-0.05
0.03
-0.01
*, **, and *** represent 0.05, 0.01, and
0.001 significance levels, respectively
58
After rotation, Corvallis factors three and four are essentially factors for single
traits (Table 15). These traits are seed weight for factor three and sterile spikelets for
factor four. Since QTL mapping was already carried out for these traits individually,
these factors were not considered further. The remaining three spike morphology
traits primarily loaded on the first two factors (Table 15). Factor one can be thought
of as a factor for the seeds per fertile spikelet contribution to seeds per spike and factor
two can be thought of as a factor for the fertile spikelets per spike contribution to
seeds per spike. Using this reasoning, the QTLs detected for factor one would be
expected to be the same as those for seeds per fertile spikelet. The QTLs detected for
factor two would be expected to be the same as those for fertile spikelets per spike.
One QTL was detected for factor one at Corvallis (Table 16) and that QTL was
found at the same site as the only QTL detected for seeds per fertile spikelet at that
location (Table 13). Three QTLs were detected for factor two (Table 16) and three
QTLs were detected for fertile spikelets per spike (Table 12). Each had matching
QTLs on 2A and 5B, but the third QTL differed in the two traits. The LOD threshold
for declaring a QTL significant was 3.1 for both of these traits. On 3B, the LOD score
for seeds per fertile spikelet exceeded this threshold by a score less than 0.1 LOD over
and factor two was below the threshold by a score less than 0.1 LOD under (results not
shown). Thus, QTL mapping of these traits at that locus was essentially the same. On
chromosome 3A a significant QTL was found for factor two (Table 16), but none was
found for fertile spikelets per spike (Table 12).
59
Table 15 Factor loadings for rotated spike morphology traits.
fertile
sterile
seeds /
spikelets spikelets fertile
seed
seeds / variance
Factor
/ spike
/ spike
spikelet weight spike
explained
Corvallis
1
2
3
4
Communuality
0.06
0.98
-0.16
-0.07
1.00
-0.16
-0.07
0.04
0.98
1.00
0.98
-0.11
-0.05
-0.14
1.00
-0.09
-0.16
0.98
0.04
1.00
0.84
0.50
-0.14
-0.15
1.00
34.40%
25.05%
20.26%
20.23%
100%
Pendleton
1
0.14
-0.11
0.99
-0.11 0.87
35.42%
2
0.95
-0.52
-0.03
0.06
0.48
28.31%
3
-0.03
0.66
-0.12
0.92
-0.12
26.37%
Communuality 0.93
0.73
0.99
0.86
0.99
90.11%
"Communuality" represents proportion of the variation for that variable
explained by the factors.
60
Table 16 QTLs for spike morphology factors.
Position LOD
2
Score Add. E. R
Factor Chrom. Nearest Marker (cM)
Corvallis
1
2
2A
2A
3A
5B
gwm372cluster
348413cluster
barc346
wPt-2273cluster
110.7
117.8
145.5
93
5.391 -0.296 0.086
8.753 -0.365 0.132
3.482 0.3069 0.094
3.449 -0.23 0.051
Pendleton
1
2A
wPt-1142cluster
113.8 4.616 -0.255 0.063
3A
gwm2
161.9 3.27 -0.248 0.061
5A
wPt-4203cluster
138.4 6.068 0.3005 0.088
2
2A
gwm95
108.4 6.139 -0.296 0.087
5A
B1
136.4 8.067 -0.345 0.117
1B
rPt-9074
147.5 3.184 0.2203 0.047
4B
gwm513
36.2 3.374 -0.221 0.047
5B
378650cluster
88.1 3.005 -0.195 0.037
6B
wmc494cluster
59.7 3.071 -0.197 0.038
3
5A
B1
134.4 6.618 0.2897 0.083
5B
wPt-3329
149.2 5.099 0.4005 0.159
4D
wmc331
15 9.329 0.3989 0.155
"Factor" corresponds to factors in table 14. "Chom." is
chromosome. "Add. E." is the additive effect for replacing an allele
2
from a Nickerson parent with an allele from Tubbs. "R " is the
coefficient of determination for how the QTL explains the residuals
after accounting for the covariates.
61
Overall, the two Corvallis factors detected or nearly detected the same
significant QTLs that were found for seeds per fertile spikelet and fertile spikelets per
spike as expected. The factors also identified an additional QTL not found mapping
the individual traits. These results indicate that factor analysis may have increased
power to detect QTLs. However, there was a QTL for seeds per spike on 4A that
didn’t show up with the factors or the other traits (Table 11). This could indicate that
the factor analysis didn’t account for some of the genetic variation in seeds per spike.
On the other hand, it is possible that this QTL is false. Analyzing seeds per spike for
QTLs after already scanning the two traits that should account for all of its variation
inflates the probability of type I error (Jiang and Zeng, 1995). It is important to know
which of the QTLs detected were real and which are false positives in order to
adequately rate the benefits of using the factor analysis approach. The best way to do
this is to collect more years of data and compare the results of multiple years.
The rotated factors for Pendleton didn’t show single trait factors like Corvallis
(Table 15). Factor one can be considered a factor for the seeds per fertile spikelet
contribution to seeds per spike just like factor one at Corvallis. Factor two can be
considered the fertile spikelets per spike contribution to seeds per spike, but it also has
a meaningful negative loading of sterile spikelets per spike. Biologically this could
make sense. The number of total spikelets per spike is determined at an early growth
stage and when the plant experiences stress at a later growth stage it will abort some of
these spikelets (Acevedo et al., 2002). The abortion of spikelets results in sterile
62
spikelets. There is a higher proportion of sterile spikelets to fertile spikelets at
Pendleton than there is at Corvallis (Table 2). This is probably explained by greater
water and heat stress at Pendleton that caused the formation of sterile spikelets. It may
be that lines that produced more spikelets per spike were also less likely to abort those
spikelets. This reasoning is supported by the significant inverse correlation between
fertile spikelets per spike and sterile spikelets per spike that was observed at Pendleton
(Table 6). Factor three loaded seed weight and sterile spikelets per spike (Table 15).
The biological explanation for this factor could be that lines that aborted more
spikelets had relatively less seeds to act as a sink for photosynthates compared to their
production of photosynthates. Thus, these lines produced larger seeds on average.
For the sake of comparing the QTL mappings of the Pendleton factors to those
of the single traits, factor one QTLs are compared to QTLs for seeds per fertile
spikelet, factor two is compared to fertile spikelets per spike, and factor three is
compared to seed weight. Factor one mapped QTLs (Table 16) that were essentially
the same as those for seeds per fertile spikelet (Table 13). Factor two mapped QTLs
(Table 16) that included essentially the same QTLs that were mapped for fertile
spikelets per spike (Table 12). Factor two also identified QTLs on 4B and 6B that
were not identified by fertile spikelets per spike. Factor three found QTLs (Table 16)
that were essentially the same as those for seed weight (Table 13).
Pendleton factors one and three behaved the same as their equivalent single
traits. Factor two appeared to have greater power to detect QTLs than its equivalent
63
single trait. The increased power may be attributable to considering multiple traits at
once with the factor scores. In particular, the inverse correlation of sterile spikelets
per spike with fertile spikelets per spike may be increasing detection power in a
similar way to the power increase observed in multiple-trait QTL analysis (Jiang and
Zeng, 1994).
There is a lack of conclusive evidence supporting using factor analysis as a
method for improved QTL mapping. However, the method did find QTLs that
weren’t previously detected. This provides some anecdotal evidence that the
procedure has merit. Without knowing the true composition of QTLs for these traits,
it cannot be determined if this is the result of increased power or simply error.
Ultimately, the best way of analyzing this method is through simulation studies were
the true distribution of QTLs is known. The results in this study indicate that such a
simulation study should be performed. If the simulation study does show increased
power and reduce error using factor analysis, the results from this method are probably
more useful than those gathered using the traits individually.
64
5. CONCLUSION
A large number of QTLs were detected for traits related to yield in three
genetic mapping populations representative of populations that produce elite
germplasm in OSU's wheat breeding program. These QTLs may eventually be useful
for developing MAS procedures for lines in OSU's program. For example, selection
for the favorable allele at the Xgwm372 locus can be implemented in relatively short
order. This should result in wheat that has greater spike fertility, and more fertile
spikelets. However, greater understanding of all of the detected QTLs and how they
are interrelated is needed to maximize the effectiveness of a MAS program.
This greater understanding includes finer genetic mapping to identify genetic
markers that are more closely linked to the identified QTLs and are inexpensive to use.
Additional years of data are also needed to expand the scope of inference for this
study. More complicated analyzes could also improve the usefulness of the data
collected. Multiple-trait QTL mapping can be used to distinguish between pleiotropy
and close linkage of QTLs (Zeng, 1994). This would allow a breeder to determine if it
is possible to break unfavorable linkages. The impact of epistasis on these traits also
needs to be examined and can be accomplished using MIM (Kao et al., 1999). These
two techniques would greatly enhance understanding of genetic control of the
examined traits in these populations. Since these populations have similar genetic
65
backgrounds to other populations used to generate varieties in OSU’s wheat breeding
program, the effects of the detected QTLs may be very similar in those other
populations.
Using factor analysis to handle correlated traits showed promising initial
results. Previously undetected QTLs were identified with this approach. However,
there is insufficient evidence to conclude that factor analysis increased power to detect
QTLs and decreased type I error. A simulation study needs to be conducted to
definitively answer these questions.
66
6. BIBLIOGRAPHY
Acevedo E, Silva P, Silva H (2002) Wheat growth and physiology. In Curtis BC,
Rajaram S, Gomez Macpherson H (eds) Bread wheat: improvement and
production. Food and Agriculture Organization of the United Nations, Rome
Addisu M, Snape JW, Simmonds JR, Gooding MJ (2009) Reduced height (Rht) and
photoperiod insensitivity (Ppd) allele associations with establishment and early
growth of wheat in contrasting production systems. Euphytica 166:249-267
Alexander WL, Smith EL, Dhanasobhan C (1984) A comparison of yield and yield
component selection in winter wheat. Euphytica 33:953-961
Babu R, Nair SK, Prasanna BM, Gupta HS (2004) Integrating marker-assisted
selection in crop breeding – Prospects and challenges. Current Science 87:607619
Baenziger PS, McMaster GS, Wilhelm WW, Weiss A, Hays CJ (2004) Putting genes
into genetic coefficients. Field Crops Research 20:133-143
Beavis WD (1994) The power and deceit of QTL experiments: lessons from
comparative QTL studies. Proceedings 49th Annual Corn and Sorghum
Industry Research Conference, Washington DC, ASTA
Bernardo R (2002) Breeding for quantitative traits in plants. Stemma Press,
Woodberry pp 121-126
Borlaug NE (1983) Contributions of conventional plant breeding to food production.
Science 219:689-693
Börner A, Schumann E, Fürste A, Coster H, Leithold B, Röder MS, Weber WE (2002)
Mapping of quantitative trait loci determining agronomic important characters
in hexaploid wheat (Triticum aestivum L). Theoretical Applied Genet 105:921936
Buck-Sorlin GH (2002) L-system model of the vegetative growth of winter barley
(Hordeum vulgare L.). In Polani D, Kim J, Martinetz T. (eds) Fifth German
workshop on artificial life, abstracting and synthesizing the principles of living
T
67
systems, March 2002, Lübeck, Germany. AKA Akademische
Verlagsgesellschaft, Berlin, pp 53-64
Buerstmayr H, Ban T, Anderson JA (2009) QTL mapping and marker-assisted
selection for Fusarium head blight resistance in wheat: a review. Plant
Breeding 128:1-26
Calderini DF, Dreccer MF, Slafer GA (1995) Genetic improvement in wheat yield and
associated traits. A re-examination of previous results and the latest trends.
Plant Breeding 114:108-112
Campbell BT, Baenziger PS, Gill KS, Eskridge KM, Budak H, Erayman M, Dweikat
I, Yen Y (2003) Identification of QTLs and environmental interactions
associated with agronomic traits on chromosome 3A of wheat. Crop Science
43:1493-1505
Churchill GA, Doerge RW (1994) Empirical threshold values for quantitative trait
mapping. Genetics 138:963-971
Costello AB, Osborne JW (2005) Best practices in exploratory factor analysis: four
recommendations for getting the most from your analysis. Practical
Assessment Research & Evaluation doi:10.1.1.110.9154
Crow JF, Dove WF (1990) Mapping functions. Genetics 125:669-671
Cuthbert JL, Somers DJ, Brûlé-Babel AL, Brown PD, Crow GH (2008) Molecular
mapping of quantitative trait loci for yield and yield components in spring
wheat (Triticum aestivum L.). Theoretical Applied Genetics 117:595-608
Dempster AP, Laird NM, Rubin DB (1977) Maximum likelihood from incomplete
data via the EM algorithm. Journal of the Royal Statistical Society 39:1-38
Doerge RW (2002) Mapping and analysis of quantitative trait loci in experimental
populations. Nature Genetics 3:43–52
Doerge RW, Churchill GA (1996) Permutation tests for multiple loci affecting a
quantitative character. Genetics 142:285-294
Donald CM (1968) The breeding of crop ideotypes. Euphytica 17:385-403
Donmez E, Sears RG, Shroyer JP, and Paulsen GM (2001) Genetic gain in yield
attributes of winter wheat in the Great Plains. Crop Science 41:1412-1419
68
Ellis MH, Spielmeyer W, Gale KR, Rebetzke GJ, Richards RA (2002) 'Perfect'
markers for the Rht-B1b and Rht-D1b dwarfing genes in wheat. Theoretical
and Applied Genetics 105:1038-1042
Engledow FL, Wadham SM (1923) Investigation on yield in the cereals, Part 1.
Journal of Agricultural Science 13:390-439
Fonseca S, Patterson FL (1968) Yield component heritabilities and interrelationships
in winter wheat (Triticum aestivum L.). Crop Science 8:614-617
Gonzalez-Hernandez JL, EM Elias, SF Kianian (2004) Mapping genes for grain
protein concentration and grain yield on chromosome 5B of Triticum turgidum
(L.) var. dicoccoides. Euphytica 139:217-225
Hackett CA, Meyer RC, Thomas WTB (2001) Multi-trait QTL mapping in barley
using multivariate regression. Genetic Research 77:95-106
Haldane JBS (1919) The combination of linkage values and the calculation of distance
between the loci of linked factors. Journal of Genetics 8:299-309
Haley CS, Knott SA (1992) A simple regression method for mapping quantitative trait
loci in line crosses using flanking markers. Heredity 69:315-324
Haverkort AJ, Kooman PL (1997) The use of systems analysis and modeling of
growth and development in potato ideotyping under conditions affecting
yields. Euphytica 94:191-200
Hayton JC, Allen DG, Scarpello V (2004) Factor retention decisions in exploratory
factor analysis: a tutorial on parallel analysis. Organizational Research
Methods 7:191-205 doi:10.1177/1094428104263675
Jansen RC (1994) Controlling the type I and type II errors in mapping quantitative
trait loci. Genetics 138:871-881
Jansen RC, Van Ooijen JW, Stam P, Lister C, Dean C (1995) Genotype-byenvironment interaction in genetic mapping of multiple quantitative trait loci.
Theoretical and Applied Genetics 91:33-37
Jansen RC, Stam P (1994) High resolution of quantitative traits into multiple loci via
interval mapping. Genetics 136:1447-1455
Jiang C, Zeng Z-B (1995) Multiple trait analysis of genetic mapping for quantitative
trait loci. Genetics 140:1111-1127
69
Johnson RA, Wichern DW (2007) Applied multivariate statistical analysis, sixth
edition. Pearson Education, Upper Saddle River
Kaiser HF (1958) The varimax criterion for analytic rotation in factor analysis.
Psychometrika 23:187-200
Kao C-H, Zeng Z-B, Teasdale RD (1999). Multiple interval mapping for quantitative
trait loci. Genetics 152:1203-1216
Kaeppler SM, Phillips RL, Kim TS (1993) Use of near-isogenic lines derived by
backcrossing or selfing to map qualitative traits. Theoretical and Applied
Genetics 87:233-237
Kato K, Mirura H, Akiyama M, Kuroshima M, Sawada S (1998) RFLP mapping of the
three major genes, Vrn1, Q and B1, on the long arm of chromosome 5A of
wheat. Euphytica 101:91-95
Kato K, Miura H, Sawada S (2000) Mapping QTLs controlling grain yield and its
components on chromosome 5A of wheat. Theoretical Applied Genetics
101:1114–1121
King RW, Richards RA (1984) Water uptake in relation to pre-harvest sprouting
damage in wheat: ear characteristics. Australian Journal of Agricultural
Research 35: 327-336
Korol A, Mester D, Frenkel Z, Ronin Y (2009) Methods for Genetic Analysis in the
Triticeae. In Feuillet C, Muehlbauer GJ (eds) Genetics and genomics of the
Triticeae. Springer, New York, pp 163-199
Kosambi DD (1944) The estimation of map distance from recombination values.
Annals of Eugenics 12:172-175
Kumar J, Mir RR, Kumar N, Mohan A, Prabhu KV, Balyan HS, Gupta PK (2010)
Marker-assisted selection for pre-harvest sprouting tolerance and leaf rust
resistance in bread wheat. Plant Breeding DOI:10.1111/j.14390523.2009.01758.x
Lande R, Thompson R (1990) Efficiency of marker-assisted selection in the
improvement of quantitative traits. Genetics 124:743-756
Lander ES, Botstein D (1989) Mapping Mendelian factors underlying quantitative
traits using RFLP linkage maps. Genetics 121:185-199
70
Laperche A, Brancourt-Hulmel M, Heumez E, Gardet O, Hanoeq E, Devleen-Barret F,
Gouis J (2007) Using genotype x nitrogen interaction variables to evaluate the
QTL involved in wheat tolerance to nitrogen constraints. Theoretical Applied
Genetics 115:399-415
Li JZ, Huang XQ, Heinrichs F, Ganal MW, Röder MS (2006) Analysis of QTLs for
yield components, agronomic traits, and disease resistance in an advanced
backcross population of spring barley. Genome 49: 454-466
Li S, Jia J, Wei X, Zhang X, Li L, Chen H, Fan Y, Sun H, Zhao X, Lei T, Xu Y, Jiang
F, Wang H, and Li L. (2007) A intervarietal genetic map and QTL analysis
for yield traits in wheat. Molecular Breeding 20:167-178
Loomis RS, Rabbinge R, Ng E (1979) Explanatory models in crop physiology. Annual
Review of Plant Physiology 30:339-367
Marza F, Bai GH, Carver BF, Zhou WC (2005) Quantitative trait loci for yield and
related traits in the wheat population Ning7840 · Clark. Theoretical Applied
Genetics 21:1-11
Mi X, Eskridge K, Wang D, Baenziger PS, Campbell BT (2008) Multiple-trait QTL
mapping using a structural equation model [abstract]. American Society of
Agronomy-Crop Science Society of America-Soil Science Society of America
Annual Meeting, November 5-9, Houston, Texas
Murphy LR, Santra D, Kidwell K, Yan G, Chen X, Campbell KG (2009) Linkage
maps of wheat stripe rust resistance genes Yr5 and Yr15 for use in markerassisted selection. Crop Science 49:1786-1790
Nerson H (1980) Effects of population density and number of ears on wheat yield and
its components. Field Crops Research 3:225-234
Paux E, Sourdille P (2009) A toolbox for Triticeae genomics. In Feuillet C,
Muehlbauer GJ (eds) Genetics and genomics of the Triticeae. Springer, New
York, pp 255-283
Quarrie1 SA, Quarrie SP, Radosevic R, Rancic D, Kaminska A, Barnes JD,
Leverington M, Ceoloni C, Dodig D (2006) Dissecting a wheat QTL for yield
present in a range of environments: from the QTL to candidate genes. Journal
of Experimental Botany 57:2627–2637
71
Sherman JD, Yan L, Talbert L, Dubcovsky J (2004) A PCR marker for growth habit in
common wheat based on allelic variation at the VRN-A1 gene. Crop Science
44:1832-1838
Slafer GA, Calderini DF, Miralles DJ (1996) Yield components and compensation in
wheat: opportunities for further increasing yield potential. In Reynolds MP,
Rajaram S, McNab A (eds) Increasing Yield Potential in Wheat: Breaking the
Barriers. CIMMYT, Mexico, pp 101–132
Singh RP, Nelson JC, Sorrells ME (2000) Mapping Yr28 and Other Genes for
Resistance to Stripe Rust in Wheat. Crop Science 40:1148-1155
Snook SC, Gorsuch RL (1989) Component analysis versus common factor analysis: A
Monte Carlo study. Psychological Bulletin 106:148-154
Tanguma J (2000) Determining the number of factors to retain. Paper presented at the
annual meeting of the southwest educational research association (Dallas, TX,
January 27-29, 2000)
Upadyayula N, Wassom J, Bohn MO, Rocheford T R (2006) Quantitative trait loci
analysis of phenotypic traits and principal components of maize tassel
inflorescence architecture. Theoretical Applied Genetics 113:1395-1407
Van Ooijen JW, Voorrips RE (2006) JoinMap 4.0. Plant Research International,
Biometris, Wageningen, Netherlands, (http://www.joinmap.nl/)
Voet D, Voet JG (2004) Biochemistry, 3rd edition. Wiley, New York
Walton PD (1971) The use of factor analysis in determining characters for yield
selection in wheat. Euphytica 20:416-421
Wang S, Basten CJ, Zeng ZB (2010) Windows QTL cartographer 2.5. Department of
Statistics, North Carolina State University, Raleigh, NC,
(http://statgen.ncsu.edu/qtlcart/WQTLCart.htm)
Washmon CN, Solie JB, Raun WR, Itenfisu DD (2002) Within field variability in
wheat grain yields over nine years in Oklahoma. Journal of Plant Nutrition
25:2655-2662
Worland AJ, Sayers EJ, Börner A (2006) The genetics and breeding potential of
Rht12, a dominant dwarfing gene in wheat. Plant Breeding 113:187-196
72
Wu R, Ma CX, Casella G (2007) Statistical genetics of quantitative traits. Springer,
New York
Xu S (2003) Estimating polygenic effects using makers of the entire genome. Genetics
163:789-801
Yan L, Loukoianov A, Tranquilli G, Helguera M, Fahima T, Dubcovsky J (2003)
Positional cloning of the wheat vernalization gene VRN1. Proceedings of the
National Academy of Sciences 100:6263-6268
Yin X, Chasalow SC, Dourleijn CJ, Stam P, Kropff MJ (2000) Coupling estimated
effects of QTLs for physiological traits to a crop growth model: Predicting
yield variation among recombinant inbred lines in barley. Heredity 85:539-549
Yin X, Stam P, Kropff MJ, Schapendonk HCM (2003) Crop Modeling, QTL Mapping,
and Their Complementary Role in Plant Breeding. Agronomy Journal 95:90-98
Yin X, Struik PC, van Eeuwijk FA, Stam P, Tang J (2005) QTL analysis and QTLbased prediction of flowering phenology in recombinant inbred lines of barley.
Journal of Experimental Botany 56:967-976
Young ND (1999) A cautiously optimistic vision for marker-assisted breeding.
Molecular Breeding 5:505-510
Zeng Z-B (1994) Precision mapping of quantitative trait loci. Genetics 136:1457-1468
Zhou L, Bai G, Ma H, Carver BF (2007) Quantitative trait loci for aluminum
resistance in wheat. Molecular Breeding 19:153-16
73
APPENDICES
74
1A
gwm636
tPt-2481
wPt-5245
345256cluster
barc124cluster
FS/S-C
S/S-P
S/FS-P
S/FS-C
S/S-C
wPt-1596cluster
wPt-1888cluster
gwm666
wmc169
305077cluster
wPt-2698cluster
wPt-2127
127.5
wPt-8658
157.6
161.9
168.0
196.4
204.0
206.2
218.1
229.0
237.0
244.9
barc346
gwm2
barc356
344008cluster
wPt-0797cluster
wPt-1652cluster
wPt-2866
wmc11
wPt-1111
345152cluster
S/FS-P
GFD-C
wPt-6711cluster
wPt-5029cluster
wPt-1657
gwm515
gwm95
gwm372cluster
wPt-1142cluster
348413cluster
gwm294
gwm312
wPt-7024
381146
wPt-9277cluster
wPt-1615cluster
wPt-9797
wPt-1499cluster
wPt-6687cluster
0.0
9.6
29.6
35.2
38.6
56.7
68.4
FS/S-P
74.3
81.0
86.2
107.4
108.4
110.7
113.8
120.6
142.9
145.5
146.8
155.7
191.0
201.9
202.2
204.0
208.6
3A
HT-P
wPt-0164cluster
wPt-5577
345666cluster
wPt-5316
cfa2147
0.0
1.3
3.1
3.7
4.6
S/M2-P
191.5
199.9
214.7
217.8
223.9
gdm33cluster
wPt-6358
304069cluster
344950cluster
377889cluster
378503
cfd15cluster
304991cluster
wPt-7326
wPt-3904
barc148
cfd59
wmc278
gwm357
wPt-4765cluster
S/M2-C
0.0
19.8
31.2
32.9
35.7
39.3
44.0
47.9
54.8
77.5
91.7
95.1
96.1
97.9
99.9
2A
Figure 1 Genetic linkage map for the Tubbs x NSA 98-0995 population (TxN) with
QTL positions. Distances are in Kosambi cM. “Cluster” at the end of a marker name
denotes the use of a delegate marker. Solid boxes indicate the position of a QTL that
has a positive additive effect for the Tubbs allele. Open boxes indicate a negative
additive effect for the Tubbs allele. The trait abbreviations are “YLD” for grain yield,
“PROT” for grain protein, “TWT” for test weight, “FLW” for growing degree days to
flowering, “RIPE” for growing degree days to ripening, “GFD” for growing degree
days of grain fill duration, “HT” for height, “S/M2” for spikes per square meter,
“FS/S” for fertile spikelets per spike, “SS/S” for sterile spikelets per spike, “S/FS” for
seeds per fertile spikelet, “S/S” for seeds per spike, and “SW” for seed weight. At the
end of the trait abbreviations “-C” indicates the QTL is for Corvallis and “-P”
indicates the QTL is for Pendleton.
75
4A
5A
196.1
200.4
217.8
306004cluster
wPt-0687cluster
312808cluster
234.0
wPt-6768
255.6
gwm332
0.0
5.2
7.9
9.0
10.8
12.1
80.9
81.4
82.8
104.2
107.1
131.0
136.4
139.9
140.9
146.4
149.4
174.9
177.9
194.6
196.1
202.1
212.6
215.2
217.1
221.9
344418cluster
wPt-0049cluster
343718cluster
305231
379274cluster
wPt-0471
wPt-0047
305461cluster
348369
wPt-0950cluster
wPt-0189cluster
345055
wPt-2430
wPt-5044
wPt-2327cluster
wPt-4417cluster
wPt-6471
345421cluster
wPt-3561
wPt-9668
wPt-0746cluster
wPt-8760
wPt-0477cluster
wPt-1813
gwm614
wPt-0100cluster
SW-C
348652cluster
tPt-1772cluster
barc156cluster
wPt-1500
wPt-1568
wPt-0983cluster
wPt-2654cluster
wPt-0044cluster
372565
wPt-2859cluster
barc8cluster
344609cluster
345641
344206cluster
barc119.1
wPt-0441cluster
wPt-1684cluster
gwm413cluster
wmc216
wPt-6690
wPt-2315
wPt-3579
wPt-0705
wPt-1403
rPt-9074
wPt-0944cluster
345746
wPt-1973
303896cluster
304031cluster
378539
GFD-C
143.5
2B
YLD-P
wPt-1928
wPt-7785
barc127cluster
barc222
S/FS-P
97.1
100.0
107.4
125.6
0.0
1.0
2.2
4.2
7.3
8.7
12.3
15.9
18.2
32.8
36.5
60.2
62.8
67.2
71.5
75.2
77.8
93.8
106.7
111.0
111.2
112.3
137.5
152.5
154.2
163.4
192.3
194.6
197.5
198.5
TWT-C
gwm666
wPt-0031
wPt-9255
tPt-6794cluster
wPt-0008cluster
343497cluster
wPt-8418cluster
gwm635
305067cluster
wPt-4835cluster
FS/S-P
0.0
6.9
10.1
13.5
15.1
16.0
17.7
22.0
28.5
37.8
TWT-P
SS/S-C
Figure 1 Continued
TWT-P
gwm427cluster
344553
wPt-3247cluster
wPt-2216cluster
SS/S-P
164.1
164.7
170.7
171.5
1B
FLW-C
wPt-1377cluster
346131
tPt-6710
wPt-4270
wPt-0832
gwm334
wPt-0689cluster
wPt-9759
wPt-8266
wPt-3965cluster
wPt-0259cluster
375954cluster
wPt-7063
gwm518
348276
303843
RIPE-P
0.0
0.9
2.0
3.7
5.6
20.0
23.4
26.6
28.3
33.3
34.6
35.2
74.7
78.0
93.8
96.2
HT-C
7A
SW-P
RIPE-C
S/S-C
RIPE-P SW-C
6A
wPt-0373cluster
349403
gwm666
barc151
wPt-4262
345676
barc330
gwm186
373063
wPt-7769
348090cluster
tPt-9702cluster wPt-3884cluster
tPt-6183cluster
barc56
barc186
gwm293
gwm304
wPt-4131
tPt-4184cluster
gwm291
B1
wPt-4203cluster
wPt-7061
HT-P
PROT-C
wPt-2788
wmc617
wPt-7924
wPt-0817cluster
wPt-7939
wPt-6997
377680
311358cluster
wPt-5124cluster
wPt-2301
wPt-2291cluster
rPt-2478cluster
373576
wPt-1007cluster
wPt-6900
wPt-3491cluster
wPt-0610
wPt-7590
wPt-0032cluster
wPt-4620cluster
0.0
10.9
78.4
79.5
81.5
85.6
89.3
98.7
99.8
108.3
109.7
115.9
118.7
120.7
122.2
122.5
122.7
123.5
124.0
141.3
0.0
6.5
10.3
16.6
19.7
27.6
38.1
44.1
45.4
54.4
57.0
63.0
65.0
66.4
67.0
67.5
67.8
73.7
122.9
130.4
134.4
140.0
141.8
76
3B
4B
HT-P
gwm296
cfd51cluster
SS/S-C
wPt-3738
wPt-5320
gwm337
gwm458
wPt-3743
gwm642
0.0
8.2
SS/S-P
379913
wPt-4647cluster
gdm33cluster
wmc432cluster
cfd15
FLW-C
2D
FS/S-C
FLW-P
wPt-6156
wPt-3086cluster
78.2
83.1
88.5
97.1
99.5
102.6
S/S-C
159.7
160.3
0.0
2.7
25.2
37.4
41.4
RIPE-C RIPE-C
wPt-8920
wPt-6936
orw5
tPt-8569cluster
wPt-0577cluster
wPt-5377
375919
wPt-5138cluster
wPt-0217cluster
SW-P
0.0
3.1
11.9
14.2
15.5
17.4
18.5
22.0
27.6
FS/S-P
1D
GFD-C
wPt-1261cluster
wmc149cluster
wPt-0708cluster
wPt-1302
wPt-0419cluster
wPt-1784
barc4cluster
303942
310624
wPt-0318cluster
wPt-9467
378650cluster
wPt-2273cluster
wPt-0103cluster
349448
303987
303931cluster
wPt-1895cluster
305092cluster
wPt-3457cluster
wPt-3329
tPt-1253
wPt-3204cluster
wPt-4418
346078cluster
345374cluster
304364
311893cluster
311648cluster
wPt-0837
SW-C
TWT-P
Figure 1 Continued
0.0
1.1
20.6
22.8
24.2
27.1
39.1
72.6
78.9
80.7
123.4
125.7
127.1
128.7
136.3
140.9
144.9
159.8
162.0
172.1
175.6
178.6
180.0
182.4
188.7
199.8
224.7
232.9
235.6
249.9
SW-C
FS/S-C
7B
wPt-5234cluster
348774
wPt-1547cluster
344152cluster
wPt-3116
wPt-1241
wPt-5256cluster
wPt-0293cluster
381509cluster
wmc494cluster
343986cluster
wPt-8721
377899
346705
wPt-3657
gwm608
gwm193
wPt-2726
barc79cluster
wPt-4924
wPt-3581cluster
tPt-4990
wPt-8554
tPt-3689
wPt-9270
wPt-1048
cfd45
wPt-0696cluster
344410cluster
wPt-0171cluster
FS/S-P
HT-C
wmc617
348732
wPt-4535
barc90cluster
tPt-0602
gwm513
gwm495
gwm251
304777
wPt-0391
344980cluster
0.0
7.2
34.2
34.8
38.1
40.1
43.4
46.1
55.8
68.8
76.3
6B
0.0
1.3
3.3
7.0
7.9
35.6
38.2
47.0
54.3
59.7
61.4
64.2
65.3
70.6
72.5
74.5
75.2
78.1
81.7
85.0
85.5
87.8
91.3
93.2
101.1
103.8
128.1
147.3
156.3
156.8
5B
HT-P
GFD-P
wPt-7984
wPt-1867cluster
wPt-3921cluster
wPt-9066
wPt-5064
wPt-0013cluster
wPt-0688cluster
348055cluster
wPt-6239
wPt-6945
wPt-4209
381874cluster
wPt-5390
304025cluster
wPt-1171
wPt-3107
wPt-4428
wmc326cluster
wPt-4933
wPt-2491
wPt-8206
wPt-6834
gwm547cluster
wPt-0280cluster
wPt-1311
gwm247
0.0
2.7
3.7
11.9
21.1
36.1
36.2
50.1
54.7
57.0
58.6
63.1
64.6
66.9
92.9
104.2
118.8
135.5
137.6
141.5
171.1
174.6
177.0
179.8
180.2
181.8
77
3D
4D
wmc11
gwm161cluster
wPt-1034cluster
gwm533cluster
246.5
barc71
333.1
336.9
wmc656
barc42
cfd14
gdm46
barc97cluster
wPt-0695cluster
wPt-0493
265.6
wPt-3018cluster
289.2
301.3
303.8
wPt-0827cluster
wPt-3923cluster
wPt-5674cluster
Figure 1 Continued
TWT-P
123.5
139.3
140.4
144.3
161.4
GFD-P
wPt-1080cluster
gwm350cluster
gwm635
cfd31
gwm44
gwm111
gwm473
6D
0.0
gwm583
16.9
barc144
63.4
72.6
77.1
80.6
88.9
gwm272
wPt-0400cluster
347042
wPt-5870
gdm63
192.0
7D
0.0
1.2
3.2
4.1
5.8
9.1
15.3
SW-C
167.4
171.4
173.1
175.2
wmc720
wmc473cluster
wmc331
gdm125
wPt-4572
53.6
54.7
68.3
72.4
75.7
SW-P
gdm99
wPt-1994
wPt-8130
wPt-8463
wPt-6066
wPt-8356
wPt-2013cluster
gdm72
barc356
SS/S-P
57.0
63.0
63.6
67.9
69.3
70.2
71.6
92.2
103.5
barc217
0.0
SS/S-C
wPt-5313
0.0
5D
wmc357
0.0
65.6
70.1
wPt-1695
wmc773cluster
gwm469
121.1
gdm36
143.8
cfd45
78
1A
2A
57.7
cfd59
RIPE-C
FLW-C
tPt-6376
wPt-1888cluster
wPt-1036
305077cluster
wmc169
gwm666
wPt-5173
wPt-3697cluster
wPt-2127
wPt-1562cluster
tPt-3563cluster
wPt-11619
wPt-0714
wPt-2910cluster
gwm2
barc346
tPt-1143
gwm6661
barc356
FLW-P
gwm294
wPt-7024
gwm312
wPt-3244
FS/S-C
200.9
203.6
203.9
213.2
0.0
0.3
4.3
7.9
11.1
16.9
32.2
58.2
61.8
81.6
98.0
114.1
119.0
139.6
142.3
144.9
151.4
167.8
169.1
HT-P
wPt-1615cluster
wPt-1499cluster
wPt-6775
wPt-2696
wPt-1657
wPt-6711cluster
tPt-9405cluster
wPt-0003cluster
cfd36cluster
wPt-5738cluster
barc124cluster
wPt-0128cluster
345256cluster
gwm636
wPt-1142cluster
gwm372cluster
gwm95
gwm515
wPt-0094cluster
FS/S-P
wmc278
gwm357
barc83cluster
wPt-8797
wPt-9757
wPt-4384
wPt-5316
345666cluster
wPt-8016
wPt-4399
wPt-7339
wPt-5577
0.0
0.2
0.4
1.4
7.7
33.3
36.6
38.0
39.4
105.7
106.2
107.4
109.9
113.1
113.6
119.9
126.8
132.3
143.1
S/S-C SS/S-P
100.4
101.6
112.4
116.4
121.5
128.0
128.3
144.7
160.6
164.2
168.8
173.3
SW-P
377889cluster
304991cluster
wPt-2150
wPt-7326
wPt-3904
GFD-P
0.0
18.2
22.9
24.1
28.9
3A
Figure 2 Genetic linkage map for the larger Einstein x Tubbs population (ExT1) with
QTL positions. Distances are in Kosambi cM. “Cluster” at the end of a marker name
denotes the use of a delegate marker. Solid boxes indicate the position of a QTL that
has a positive additive effect for the Tubbs allele. Open boxes indicate a negative
additive effect for the Tubbs allele. The trait abbreviations are “YLD” for grain yield,
“PROT” for grain protein, “TWT” for test weight, “FLW” for growing degree days to
flowering, “RIPE” for growing degree days to ripening, “GFD” for growing degree
days of grain fill duration, “HT” for height, “S/M2” for spikes per square meter,
“FS/S” for fertile spikelets per spike, “SS/S” for sterile spikelets per spike, “S/FS” for
seeds per fertile spikelet, “S/S” for seeds per spike, and “SW” for seed weight. At the
end of the trait abbreviations “-C” indicates the QTL is for Corvallis and “-P”
indicates the QTL is for Pendleton.
79
4A
5A
wPt-2788
55.5
wPt-6748cluster
tPt-1772cluster
wPt-1568
wPt-2654cluster
wPt-8627
wPt-0983cluster
wPt-0044cluster
wPt-2859cluster
wPt-3465
344609cluster
wPt-4574
barc8cluster
wPt-0441cluster
barc1191
gwm413cluster
wPt-2461
wPt-6690 wPt-2315
wPt-0705
wPt-1201cluster
FLW-C
FS/S-P
Figure 2 Continued
wPt-4229cluster
wPt-4445
wPt-3247cluster
tPt-4178cluster
wPt-5310
gwm518
2B
RIPE-C
0.0
7.9
11.6
14.2
16.1
39.4
41.9
48.6
80.9
87.7
91.4
97.4
105.3
107.2
112.5
113.4
116.1
118.5
SW-C
FS/S-C
gwm635
304885cluster
gwm666
343497cluster
wPt-8418cluster
tPt-6794cluster
wPt-1885
wPt-0008cluster
wPt-9255
wPt-6959
wPt-0031
wPt-9651
barc70cluster
wPt-1179cluster
wPt-0321
barc127cluster
wPt-1446
barc222
117416cluster
wPt-0514
wPt-3992cluster
wPt-0433
wPt-5987
gwm332
wPt-11723
wPt-4831cluster
wPt-3425
wPt-1976cluster
wPt-0687cluster
306004cluster
YLD-C
0.0
1.4
4.9
53.5
54.6
55.3
64.2
75.1
77.7
122.1
125.2
130.4
151.2
171.0
171.3
184.9
225.7
228.1
230.0
245.2
249.1
251.5
255.0
255.4
255.9
256.4
258.6
259.4
261.9
265.7
1B
wPt-1285
wPt-4047 tPt-6710
wPt-1664cluster wPt-4270
117493cluster
gwm334
wPt-0689cluster
wPt-9759
wPt-8266
tPt-6278
TWT-C
7A
SS/S-P
wPt-1200
wPt-4203cluster
wPt-7061
gwm291
B1
SS/S-C
0.0
1.5
2.2
8.8
14.4
84.6
85.3
85.5
87.9
111.7
132.8
SW-P
wPt-0373cluster
FS/S-P
154.4
0.0
4.1
5.8
9.0
12.8
14.4
15.6
17.4
24.0
HT-C
S/FS-P
TWT-P
wPt-4544
rPt-2478cluster
wPt-6757
wPt-0421cluster
wPt-1007cluster
wPt-2291cluster
311358cluster
wPt-7096cluster
wPt-5124cluster
wPt-2301
wPt-6900
wPt-0610
tPt-9948cluster
wPt-7590 wPt-3491cluster
wPt-0032cluster
107.1
127.9
143.3
143.6
153.6
155.4
161.3
161.5
163.7
171.9
173.3
173.4
175.8
195.5
195.7
wPt-1165
wPt-4131
barc186
barc56
gwm293
wPt-3884cluster
gwm304
gwm186
rPt-3563
wPt-3187
barc330
wPt-4419
wPt-0707cluster
barc151
wPt-4262
gwm666
S/S-C
21.6
0.0
16.2
21.1
23.1
50.9
52.0
53.3
59.8
60.2
61.3
90.0
91.1
93.0
96.2
99.6
111.7
S/FS-C
wmc617
FS/S-C TWT-C
0.0
6A
0.0
17.1
20.8
53.9
59.4
83.0
86.7
88.6
89.0
90.0
108.1
109.9
112.6
113.8
117.6
128.7
142.9
159.3
166.8
173.2
182.3
193.9
205.7
208.5
214.7
215.6
226.6
251.1
259.3
273.6
wPt-5017
343718cluster
wPt-6643
wPt-0471
gwm526cluster
wPt-0694cluster
wPt-9257
wPt-0697
wPt-3383cluster
wPt-0473
wPt-8460
wPt-0489cluster
wPt-0189cluster
wPt-0950cluster
wmc175
wPt-2327cluster
wPt-2110
wPt-4417cluster
wPt-6471
wPt-2314
wPt-0408
wPt-7715
345421cluster
wPt-3561
wPt-1489cluster
wPt-8349
tPt-4627cluster
wPt-9859
wPt-0746cluster
wPt-0100cluster
80
3B
4B
7B
wPt-3284
wPt-4388cluster
wPt-2564
wPt-3309cluster
wPt-4119
wPt-1241
rPt-0029
wPt-6441
269.1
272.8
wPt-3657
gwm608
Figure 2 Continued
2D
0.0
5.8
wPt-2026
38.4
64.3
66.0
75.4
wPt-6963
barc229cluster
cfd92
FS/S-C
239.6
0.0
SW-C
wPt-3207cluster
wPt-9256
wPt-1541cluster
344410cluster
wPt-1264cluster
wPt-7636
TWT-C
168.1
168.9
171.8
172.6
173.0
188.0
1D
wPt-6936
wPt-8920
wPt-0837cluster
orw5
tPt-8569cluster
wPt-5216cluster
wPt-4620cluster
wPt-5377
wPt-0577cluster
wPt-5138cluster
wPt-4054
wPt-0217cluster
wPt-7720
wPt-8040
wPt-3939
wPt-2356
wPt-0194cluster
wPt-4258cluster
wPt-1266
gwm111
wPt-3723
wPt-1826
SW-P
0.0
1.0
2.8
56.8
65.0
66.9
75.2
81.5
82.5
85.6
86.2
91.1
93.4
109.4
112.1
123.1
125.6
128.2
130.6
136.7
142.0
143.8
TWT-C
gwm251
wPt-1261cluster
wmc149cluster
rPt-6127
barc216cluster
wPt-1784
barc4cluster
wPt-0318cluster
wPt-1505
348314cluster
378650cluster
wPt-0419cluster
wPt-6878cluster
305092cluster
wPt-1895cluster
wPt-3457cluster
wPt-9454
wPt-6014
wPt-3030cluster
wPt-1304
wPt-3204cluster
wPt-1881
wPt-0935cluster
tPt-1253
wPt-3329
wPt-2514cluster
wPt-7006
220.6
wPt-3855
295.3
cfd48
gwm296
gwm261cluster
gwm6
HT-P
0.0
6.9
10.8
22.9
38.6
39.4
55.5
165.9
0.0
28.9
32.5
35.7
43.2
44.7
48.3
51.8
62.8
63.1
82.8
89.5
94.8
97.7
112.1
134.6
158.0
176.2
182.9
190.2
192.1
198.8
202.9
211.6
219.8
250.8
FLW-C
6B
rPt-6847
HT-P
wPt-8021cluster
wPt-0367
wmc326cluster
140.1
HT-C
wPt-3342cluster
300.5
313.4
327.1
gwm149
gwm664
wmc617
barc90cluster
wPt-3991
wPt-8397
barc20cluster
tPt-0602
gwm513
gwm495
TWT-C
285.1
0.0
4.8
14.9
17.0
20.3
22.8
23.9
26.3
27.0
65.4
FS/S-P
wPt-8910
wPt-0013cluster
wPt-0688cluster
wPt-6945
wPt-4209
wPt-6020
381874cluster
348055cluster
barc068
tPt-4541cluster
wPt-7158
gwm547cluster
wPt-6834
wPt-8206
S/FS-C
141.6
166.0
172.5
179.4
186.7
187.4
204.7
205.8
211.0
240.1
245.5
248.1
251.5
255.5
RIPE-C
wmc11
wPt-3041
wPt-4352
barc57cluster
wPt-1867cluster
wPt-7984
wPt-3921cluster
barc133
wPt-9066
S/S-C
0.0
24.6
38.1
43.6
59.6
68.4
70.0
71.5
71.9
5B
110.6
wPt-0638cluster
152.1
163.0
wmc18
wPt-0298
81
3D
4D
gwm3
261.4
barc323
7D
0.0
11.7
barc126cluster
gwm44
87.0
88.3
gwm473
wPt-7508
123.0
136.1
146.1
GDM99
GDM67
wPt-1859
216.3
wPt-1080cluster
237.0
gwm635
285.0
cfd31
Figure 2 Continued
wPt-0472cluster
HT-C
242.3
23.2
YLD-C
gwm161cluster
gwm533cluster
wPt-1034cluster
wmc11
wmc473cluster
wPt-3058
FLW-C
144.1
146.4
147.1
148.7
0.0
0.7
HT-P
gwm52
GDM72
wPt-8356
wPt-1994
wPt-2013cluster wPt-6066
wPt-8130
wPt-7847
0.0
23.8
40.6
43.4
44.4
45.2
59.5
5D
6D
0.0
5.7
12.0
gwm583
wPt-2256
cfd8
65.2
GDM63
109.8
gwm292
173.5
174.8
wmc161
wPt-9169 wPt-0162
244.7
249.6
264.4
gwm272
wPt-0400cluster
wPt-1197
0.0
0.6
152.4
GDM127
cfd33
cfd49
82
1A
377889cluster
304991cluster
wPt-2150
wPt-7326
wPt-3904
wmc278
cfd59
gwm357
wPt-8797
barc83cluster
wPt-9757
wPt-4384
113.7
wPt-5316
139.4
145.8
148.1
150.5
345666cluster
wPt-4399
wPt-8016
wPt-7339
wPt-5577
wPt-0164cluster
gwm636
tPt-2481
345256cluster
wPt-6711cluster
wPt-5029cluster
88.1
92.3
94.2
96.0
116.3
119.5
128.8
129.4
gwm95
gwm515
wPt-1142cluster
gwm372cluster
gwm294
gwm312
wPt-7024
wPt-0094cluster
159.2
wPt-1499cluster
174.7
wPt-1615cluster
225.3
230.0
236.2
wPt-9797
wPt-6687cluster
wPt-7009
gwm666
wPt-1036
wPt-1888cluster
tPt-6376
wmc169
305077cluster
wPt-2698cluster
wPt-1277
wPt-3697cluster
wPt-2127
wPt-1562cluster
tPt-3563cluster
wPt-11619
gwm2
barc346
barc356
gwm6661
rPt-9057
tPt-1143
wPt-2910cluster
wPt-0714
wPt-1652cluster
wPt-1111
wPt-2866
wmc11
wPt-4352
345152cluster
barc57cluster
SS/S-P
24.8
27.4
30.7
34.0
0.0
18.5
19.8
27.9
32.7
38.0
52.1
57.5
83.8
92.3
93.3
95.2
95.5
95.9
101.3
119.2
120.6
147.3
156.9
170.1
175.9
176.4
200.6
203.8
219.1
220.6
238.4
242.3
S/M2-C
183.2
187.5
0.0
3A
S/S-P
0.0
3.0
5.8
8.8
21.9
44.4
46.4
53.6
57.4
61.9
70.4
71.1
2A
Figure 3 Genetic linkage map for the smaller Einstein x Tubbs population (ExT2)
with QTL positions. Distances are in Kosambi cM. “Cluster” at the end of a marker
name denotes the use of a delegate marker. Solid boxes indicate the position of a QTL
that has a positive additive effect for the Tubbs allele. Open boxes indicate a negative
additive effect for the Tubbs allele. The trait abbreviations are “YLD” for grain yield,
“PROT” for grain protein, “TWT” for test weight, “FLW” for growing degree days to
flowering, “RIPE” for growing degree days to ripening, “GFD” for growing degree
days of grain fill duration, “HT” for height, “S/M2” for spikes per square meter,
“FS/S” for fertile spikelets per spike, “SS/S” for sterile spikelets per spike, “S/FS” for
seeds per fertile spikelet, “S/S” for seeds per spike, and “SW” for seed weight. At the
end of the trait abbreviations “-C” indicates the QTL is for Corvallis and “-P”
indicates the QTL is for Pendleton.
83
4A
5A
wPt-2301
171.8
wPt-5124cluster
208.9
311358cluster
6A
0.0
gwm291
B1
wPt-1200
wPt-4203cluster
wPt-7061
tPt-4184cluster
151.2
156.7
166.4
167.8
170.5
186.3
SS/S-P
144.7
SS/S-C
wPt-0610
TWT-P
117.3
FLW-P
wPt-6900
FS/S-C
90.6
wPt-0373cluster
gwm666
wPt-0707cluster
barc330
wPt-4419
gwm186
wPt-3187
rPt-3563
barc186
348090cluster
tPt-6183cluster tPt-9702cluster
gwm304
wPt-4131
wPt-3884cluster
gwm293
barc56
0.0
16.3
54.8
55.7
56.6
66.2
66.8
70.1
85.1
87.2
89.8
90.7
92.6
94.7
96.9
99.3
S/S-P
wmc617
wPt-2788
wPt-6748cluster
wPt-7096cluster wPt-4544
rPt-2478cluster
wPt-2291cluster
wPt-1007cluster
wPt-0032cluster
wPt-7590
wPt-3491cluster
tPt-9948cluster
S/FS-P
0.0
2.0
9.4
10.6
11.3
14.0
14.4
15.3
20.4
22.3
55.3
7A
wPt-5310
wPt-2582
tPt-6661
wPt-3247cluster
wPt-4229cluster
93.5
wPt-4445
gwm518
S/M2-P
wPt-0008cluster
tPt-6794cluster
wPt-8418cluster
wPt-1885
343497cluster
wPt-0031
wPt-6959
wPt-9255
gwm635
gwm666
304885cluster
wPt-9651
wPt-6824
wPt-1179cluster
wPt-4835cluster
wmc283cluster
barc127cluster
wPt-1446
barc222
wPt-0321
117416cluster
wPt-3992cluster
wPt-0514
wPt-0433
wPt-5987
wPt-0961cluster
wPt-3425
wPt-6768
gwm332
HT-P
52.2
53.5
54.6
55.2
56.6
60.3
63.2
64.5
71.0
72.4
73.3
91.2
98.1
98.5
105.2
133.7
136.7
139.0
162.6
173.2
191.1
195.5
200.3
221.0
225.8
245.1
249.1
259.6
273.4
0.0
19.9
36.8
39.3
41.8
42.4
57.2
86.5
92.1
98.6
119.8
127.2
130.8
133.7
141.8
142.4
145.7
148.0
148.4
149.2
154.5
154.9
HT-C
Figure 3 Continued
306004cluster
wPt-0687cluster
wPt-1976cluster
FS/S-C
214.3
S/FS-P
tPt-6278
wPt-0689cluster wPt-8266
gwm334
tPt-6710
wPt-4047
wPt-1285
wPt-1664cluster
wPt-4270
117493cluster
wPt-9759
S/FS-C
135.3
142.4
144.6
150.0
152.6
158.0
160.2
163.7
165.7
170.0
0.0
1.8
7.1
barc156cluster
wPt-1500 wPt-0983cluster
wPt-2654cluster
wPt-8627
wPt-1568
tPt-1772cluster
wPt-2859cluster
wPt-0044cluster
344609cluster
wPt-4574
wPt-3465
barc8cluster
gwm413cluster
wPt-0441cluster
barc1191
wmc216
wPt-6690 wPt-2315
wPt-3579
wPt-1201cluster
wPt-0944cluster
rPt-9074
wPt-4651
GFD-P
54.5
61.5
66.4
69.8
1B
84
2B
3B
0.0
22.7
24.0
28.6
32.3
35.9
45.1
47.2
48.3
67.0
77.2
88.4
92.8
94.3
wPt-0049cluster
343718cluster
wPt-5017
0.0
5.2
15.1
wPt-0391
HT-C
HT-P
wPt-8206
wPt-7158
wPt-6834
gwm547cluster
wPt-0280cluster
6B
wPt-2514cluster
348314cluster
Figure 3 Continued
wPt-2175
wPt-0293cluster
cfd1cluster
381509cluster
wmc494cluster
gwm518
barc101cluster
wPt-7489
wPt-2726
tPt-3689
wPt-8554
wPt-4924
wPt-3581cluster tPt-4990
wPt-9231
barc79cluster
wPt-3657
gwm193
gwm608
wPt-6441
344410cluster
wPt-0171cluster
0.0
17.8
22.2
27.5
31.5
33.6
37.3
38.0
39.5
40.8
47.4
49.2
57.6
60.8
64.1
72.8
75.8
79.2
88.4
164.8
165.4
166.9
wPt-6936
wPt-8920
wPt-0837cluster
FS/S-C
171.9
181.9
132.4
150.7
151.2
152.7
157.6
159.6
162.9
165.7
169.5
171.6
171.7
173.8
176.9
178.4
179.3
184.0
189.1
192.4
197.3
226.7
232.7
1D
wPt-4620cluster
wPt-5216cluster
orw5
tPt-8569cluster
wPt-5377
wPt-0577cluster
wPt-5138cluster
wPt-4054
wPt-7720
wPt-0217cluster
wPt-8040
wPt-3939
wPt-2356
wPt-0194cluster
wPt-1266
wPt-3723
wPt-1826
gwm111
wPt-4258cluster
HT-C
wPt-1302
barc232cluster
wPt-1304
wPt-0935cluster
wPt-1881
wPt-1547cluster
wPt-3774cluster
RIPE-C
72.6
81.8
82.8
86.2
96.9
0.0
1.8
7B
PROT-P
wPt-0708cluster
GFD-C
0.0
S/M2-C
189.6
GFD-P
rPt-6847
SS/S-C
164.5
SS/S-P
gwm251
RIPE-C
114.9
S/S-P
gwm495
RIPE-P
91.6
HT-P
gwm149
gwm664
wmc617
wPt-0246cluster
wPt-8397
barc20cluster
wPt-3991
barc90cluster
tPt-0602
gwm513
wPt-0065
YLD-P
5B
212.4
217.9
223.2
224.6
233.6
0.0
26.4
31.9
33.5
34.3
34.7
39.2
42.4
43.4
44.5
S/FS-P
139.4
barc068
wPt-6945
wPt-4209
wPt-6020
381874cluster
348055cluster
wPt-8910
wPt-0688cluster
wPt-0013cluster
wPt-9066
barc133
wPt-7984
wPt-3921cluster
wPt-1867cluster
HT-C
wPt-0100cluster
wPt-0746cluster
wPt-9859
tPt-4627cluster
wPt-1489cluster
wPt-3561
wPt-9668
345421cluster
wPt-7715
wPt-2110
wPt-0408
wPt-2314
wPt-6471
wPt-4417cluster
wPt-2327cluster
wmc175
wPt-0189cluster
wPt-0489cluster
wPt-0950cluster
wPt-8460
wPt-9257
wPt-0697 wPt-0473
wPt-0694cluster
86.1
97.4
107.3
122.7
136.0
143.0
146.1
153.0
161.4
180.1
191.7
199.0
204.5
211.4
226.7
232.0
237.4
239.4
247.3
249.5
259.3
260.6
264.3
4B
0.0
15.8
17.1
wPt-2026
wPt-6963
barc229cluster
67.1
cfd92
85
2D
3D
gwm3
220.6
barc323
237.5
barc71
268.4
cfd60
5D
6D
wPt-2256
32.9
42.2
43.5
cfd8
gwm583
gwm292
76.5
gwm272
108.6
109.8
111.5
0.0
1.8
wPt-0400cluster
wPt-1197
GDM63
0.0
1.9
S/M2-C
0.0
7D
GDM127
cfd33
0.0
8.0
barc126cluster
gwm44
34.3
44.7
46.0
gwm111
gwm473
cfd14
71.9
73.8
82.5
95.7
GDM46
GDM99
GDM67
wPt-1859
183.5
gwm635
212.0
222.5
226.9
cfd31
wmc463
wPt-1100cluster
269.0
270.8
274.7
275.9
278.1
282.4
wPt-3923cluster
wPt-5674cluster
wPt-0493
wPt-3018cluster
wPt-0695cluster
barc97cluster
wmc161
wPt-9169 wPt-0162
Figure 3 Continued
HT-C
206.5
TWT-C
wPt-0619cluster
wPt-0638cluster
gwm52
GDM72
wPt-8356
wPt-1994
wPt-2013cluster
wPt-6066
wPt-8130
wPt-7847
wPt-3058
wmc473cluster
wPt-0472cluster
HT-P
201.0
205.2
81.9
95.2
106.3
109.6
110.7
112.0
116.0
130.8
0.0
0.6
16.9
RIPE-C
wPt-4691cluster
wPt-9580
gwm161cluster
wmc11
gwm533cluster
wPt-1034cluster
PROT-P
91.3
98.9
0.0
4.6
6.6
7.2
S/M2-P
gwm261cluster
gwm296
wmc18
wPt-0298
gwm608
SS/S-P
0.0
2.7
5.9
6.1
10.4
4D
Download