JXB Advance Access originally published online on February 14, 2005
Journal of Experimental Botany 2005 56(413):967-976; doi:10.1093/jxb/eri090
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RESEARCH PAPER |
QTL analysis and QTL-based prediction of flowering phenology in recombinant inbred lines of barley
1College of Agriculture, Jiangxi Agricultural University, Nanchang 330045, PR China
2Crop and Weed Ecology Group, Wageningen University, PO Box 430, 6700 AK Wageningen, The Netherlands
3Laboratory of Plant Breeding, Wageningen University, PO Box 386, 6700 AJ Wageningen, The Netherlands
* To whom correspondence should be sent at Wageningen. Fax: +31 317 485572. E-mail: Xinyou.Yin{at}wur.nl
Received 24 August 2004; Accepted 22 November 2004
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
Combining ecophysiological modelling and genetic mapping has increasingly received attention from researchers who wish to predict complex plant or crop traits under diverse environmental conditions. The potential for using this combined approach to predict flowering time of individual genotypes in a recombinant inbred line (RIL) population of spring barley (Hordeum vulgare L.) was examined. An ecophysiological phenology model predicts preflowering duration as affected by temperature and photoperiod, based on the following four input traits: fo (the minimum number of days to flowering at the optimum temperature and photoperiod),
1 and 2 (the development stages for the start and the end of the photoperiod-sensitive phase, respectively), and 
(the photoperiod sensitivity). The model-input trait values were obtained from a photoperiod-controlled greenhouse experiment. Assuming additivity of QTL effects, a multiple QTL model was fitted for the model-input traits using composite interval mapping. Four to seven QTL were identified for each trait. Each trait had at least one QTL specific to that trait alone. Other QTL were shared by two or all traits. Values of the model-input traits predicted for the RILs from the QTL model were fed back into the ecophysiological model. This QTL-based ecophysiological model was subsequently used to predict preflowering duration (d) for eight field trial environments. The model accounted for 72% of the observed variation among 94 RILs and 94% of the variation among the two parents across the eight environments, when observations in different environments were pooled. However, due to the low percentage (34–41%) of phenotypic variation accounted for by the identified QTL for three model-input traits (1, 2 and ), the QTL-based model accounted for somewhat less variation among the RILs than the model using original phenotypic input trait values. Nevertheless, days to flowering as predicted from the QTL-based ecophysiological model were highly correlated with days to flowering as predicted from QTL-models per environment for days to flowering per se. The ecophysiological phenology model was thus capable of extrapolating (QTL) information from one environment to another. Key words: Ecophysiological modelling, flowering time, genotype–phenotype relationship, Hordeum vulgare L., model-input traits, photoperiod, quantitative trait loci, temperatures
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
The advent of various molecular markers has enabled the unravelling of complex crop traits into the effect of individual quantitative trait loci, QTL (Paterson et al., 1988
). This has opened up opportunities to enhance the efficiency of plant breeding by the use of marker-assisted selection (Ribaut and Hoisington, 1998) and for concepts like ‘breeding by design’ (Peleman and van der Voort, 2003). However, QTL-by-environment interactions (QTLxE), i.e. when QTL expression is conditional on the environment, can be a major obstacle in the application of marker-assisted selection for manipulating complex traits of agronomic importance (Ribaut and Hoisington, 1998; Slafer, 2003). In the classical model for genotype-by-environment interactions (GxE), the regression on the mean model (Finlay and Wilkinson, 1963), the mean phenotypic value across genotypes per environment is used as a measure of environmental quality, the environmental index (Kearsey and Pooni, 1996). This approach does not relate phenotypic values to physical measures of the environment such as temperature. A reason to use the regression on the mean model might be that climatic and edaphic factors are hard to obtain, and therefore the plants' average performance in a particular environment is taken as an indicator of environmental quality. In the Finlay-Wilkinson model, GxE is represented by genotypic differences in sensitivity to the environmental index. The approach provides a summary of phenotypic data containing GxE and models the phenotypic responses in relation to the environment. However, the Finlay–Wilkinson approach does not allow a straightforward prediction of phenotypic responses for environments outside the set of environments used for calibrating the model.
With the development of automated weather station networks in the early 1980s and the availability of soil maps and satellite data, data for climatic and edaphic variables become now widely available (Weiss, 2003). To use the information of physical environments, the factorial regression was proposed as an ordinary linear model that allows the GxE to be modelled directly as a function of environmental variables (van Eeuwijk et al., 1996). The factorial regression approach has been adopted to analyse QTLxE, whereby environmental characterizations (often average or accumulated values over certain phases) are used as co-variables in models allowing for environment-dependent QTL expression (van Eeuwijk et al., 2002). The success of this approach relies on whether the correct physical environmental factors are included and whether incorporated values of these environmental factors match the relevant growth periods. Obviously, the correct choice of physical environmental factors and their values requires the knowledge of crop physiology for the traits under study. Even when the choice is correctly made, the power of the factorial regression to describe genotypic differences in relation to the environment can be limited because complex crop traits are often the result of non-linear interactions between genetic, environmental and managerial factors on multiple component processes, integrated over ontogenetic stages. The inability of the factorial regression to flexibly incorporate managerial factors and temporal (both seasonal and diurnal) dynamics over the growing season and spatial profiles of environmental variables within a crop canopy is the major limitation to it being used to resolve GxE on a biological basis.
The aim was to bring the link between crop physiology and genetics a step further, focusing on the GxE problem while studying genotype-to-phenotype relationships. This is possible because of the availability of ecophysiology-based crop growth models that relate elementary processes to environmental variables. In the context of QTL analysis, the complex trait was first dissected into component traits by the use of an ecophysiological model and then, instead of searching for QTL for the complex trait itself, QTL models for those component traits were found. The component traits often correspond to model-input parameters, reflecting effects with a genetic component. Another category of inputs for ecophysiological models is formed by weather, soil variables and management options. The ecophysiological model structure provides algorithms following physiological principles to quantify the interactions between ontogenetic component processes and environmental factors. Unlike factorial regression models, ecophysiological models use daily (or if needed, derived hourly) values of environmental variables during the whole growing season and consider spatial variation of these variables within the crop (e.g. the decline of radiation with the depth of the crop canopy) to drive simulation for predicting phenotypes. In addition, the effects of managerial factors (e.g. quantity and timing of fertilization) can be more flexibly considered as input variables in ecophysiological models.
In the first studies exploring the usefulness of this ecophysiological approach, QTL for various input traits of a model that predicts crop yields (Yin et al., 1999) were identified. QTL-based parameters were then fed back into the model to predict yield performance of individuals in the population under study (Yin et al., 2000a). Although the correlation between predicted yields using QTL-based parameters and those using the original measured, phenotypic parameters was high, the ability of current crop growth models is not yet sufficient to predict differences in complex traits like yield among individuals of a segregating population (Yin et al., 2000a, b).
Following a similar physiology-based approach, Reymond et al. (2003) demonstrated the potential value of combining ecophysiological modelling and genetic mapping in predicting GxE interaction on a simple trait, leaf elongation rate in maize (Zea mays L.). The QTL analysis was performed on parameters of a simple linear model for predicting the leaf elongation rate as affected by temperature and water deficit. It was shown that the rates of individuals, including lines not used for QTL analysis, of the mapping population were well predicted by the combined QTL- and ecophysiological model, for any climatic scenario.
In the present analysis, the same concept is applied to flowering time in barley (Hordeum vulgare L.). An ecophysiological model for phenology was used, where the daily rate of progress towards flowering is modelled as an interactive function of daily temperature and photoperiod (Yin et al., 2005). The QTL analysis has been performed for each of the model-input parameters. The aim was to test whether the flowering time of individuals in the mapping population under a range of field conditions can be predicted by combined use of QTL and ecophysiological modelling.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
Plant material and parameters of the phenology model
The plant material used in the present analysis consists of 94 individuals produced by eight generations of single seed descent from a cross of the two-row spring barley cultivars (‘Apex’ and ‘Prisma’). This is the same recombinant inbred line (RIL) population that was used in earlier research (Yin et al., 1999
, 2000a, b).
The ecophysiological phenology model used here was described in detail by Yin et al. (2005), so only summary information is presented here. In this model, daily development rate, 
i, at a developmental stage () is described as:
 | (1) |
 | (2a) |
 | (2b) |
1, 2, and . fo defines the minimum number of days from sowing to flowering when both temperature and photoperiod are at their optimum, representing the genotypic intrinsic earliness of flowering. 1 and 2 are the (dimensionless) development stage for the start and the end of the photoperiod-sensitive phase, respectively. is the parameter characterizing the photoperiod-sensitivity during the photoperiod-sensitive phase (h–1). Values of these traits for each RIL were estimated, via curve-fitting to data from a photoperiod-controlled greenhouse experiment, where plants were mutually transferred between long-day and short-day photoperiods at 10 d intervals (Yin et al., 2005). Because the function g(T) is non-linear and temperature fluctuates diurnally, g(T) was estimated on an hourly basis and hourly g(T) values were averaged to obtain the daily value. The hourly temperature was estimated from daily maximum and minimum temperature by a sine function. In this way, daily developmental rate, i, was calculated. The day when total daily i accumulated from sowing is just above or equal to 1.0 is the predicted time when an RIL flowers.
QTL detection
The QTL analysis was performed on the basis of the marker linkage map established by Yin et al. (1999) for the ‘ApexxPrisma’ RIL population, which contains 191 AFLP markers and one morphological marker, covering a total map length of 965 cM (Fig. 1). A mapping software MapQTL®4.0 (Van Ooijen et al., 2002) was used in identifying QTL positions in the genome for a given trait. QTL for each of the four traits fo, 1, 2, and were identified, using a composite interval mapping (or called the MQM mapping) method (Jansen, 1995) as implemented in MapQTL®4.0. In this method, background markers are selected to take over the role of the putative QTL as cofactors to reduce the residual variance. A two-stage MQM analysis was performed. In the first stage, a conventional interval mapping was performed at a 2 cM interval; the LOD profiles from interval mapping were inspected and the marker closest to each LOD peak was selected as the cofactor to perform the MQM mapping analysis further. The inclusion of cofactors may lead to new peaks in the LOD profiles, which suggested further cofactors to be included in the analysis. Several cycles were performed to obtain the potentially maximum number of cofactors for the MQM analysis. These cofactor markers were then subjected to backward elimination, as implemented in MapQTL®4.0, to select the best model for the second stage MQM analysis. Such a backward elimination procedure leaves out one cofactor at a time to create a subset of cofactors. The likelihood of each of these subset models is compared with the likelihood of the full model with all cofactors, and the subset model which causes the smallest change in likelihood is chosen as the starting set for a subsequent round of elimination. This process continues until the change in likelihood is significant according to the 0.05 P-value for the test. The then retained set of cofactors was used in the second stage of the MQM analysis. In the final LOD profile, QTL were declared according to the threshold LOD scores ranging from 2.8 to 3.0 (genome-wide false-positives rate 5%), depending on chromosome map length and the number of chromosome pairs (Van Ooijen, 1999).
|
|
Predicting flowering time using QTL-based parameters of the ecophysiological model
The authors wanted to test whether the flowering time could be accurately predicted on the basis of identified QTL for each of the model-input traits fo,
1, 2, and . To this end, QTL-based values for each of the traits had to be estimated for each RIL. Because in MapQTL®4.0, only additive gene actions are assumed for the RIL populations, no attempt was made to model epistatic QTL in this analysis. The genetic (additive effect) predictor of the i-th QTL genotype for the j-th RIL, gij, is obtained from the conditional allelic probabilities at QTL positions given the information at flanking markers (Jiang and Zeng, 1995, 1997), according to:
 | (3) |
The trait QTL loci identified within the MapQTL®4.0 modelling exercises, served as the basis for a last cycle of modelling in which multiple regression models were fitted to trait responses using the genetic predictors, gij, corresponding to the earlier identified QTL. From these regression models, predicted trait values were produced. The predicted value for the j-th RIL, 
is then:
 | (4) |
is the estimated intercept, and 
is the estimated additive effect of the i-th QTL on the trait (i=1, 2, ..., n). The coefficient of determination for equation (4), a measure for goodness of fit, is defined by the total percentage of phenotypic variation in the model-input trait that is accounted for by the QTL.
Predicted values for fo, 1, 2, and following from application of equation (4) were then used as input to the ecophysiological phenology model, equations (1) and (2). Predicted days to flowering using the QTL-based model input traits were compared with predictions using the observed phenotypic input traits. The comparison included the 94 RILs under eight independent field conditions during two growing seasons, created by using different sowing dates (Yin et al., 2005). Such a comparison was also performed for the two parents, which provided an additional independent test, since the QTL analysis used no information of either marker genotype or trait phenotype of the parents. Finally, predicted days to flowering by the QTL-based ecophysiological model were compared with those predicted by identified QTL per field environment for days to flowering per se.
|
Results and discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
The mean, minimum and maximum trait values, and the phenotypic correlation coefficients between the four traits are shown in Table 1. All correlations were statistically significant (P <0.01), and those between
1 and , and between 2 and were particularly so. Values of 1, 2, and were very similar for the two parents and the frequency distribution of trait values showed transgressive segregation (Fig. 2), which indicates that alleles of increasing and decreasing effect are dispersed over the parents, ‘Apex’ and ‘Prisma’. An ecophysiological model analysis (Yin et al., 2005) showed that all four traits were important, albeit to different extents, for predicting differences of flowering time among the individual genotypes of the population.
|
|
QTL for the model-input traits
The genetic analysis led to identified QTL for each of the four traits (Table 2). The two-stage MQM analysis successfully detected 4–7 QTL for a trait. Some genome positions were common to all four traits, whereas others were only specific to one or two traits (Fig. 1). Using the same ‘ApexxPrisma’ RIL population, Yin et al. (1999
, 2003) have shown that a major gene, the denso (or designated as sdw1) locus (with a dwarfing allele from ‘Prisma’) located at 126.4 cM on chromosome 3(3H) (Fig. 1) pleiotropically affects a number of traits including the preflowering duration. This gene is shown here to be important for, or closely linked to, all four traits (fo, 1, 2, and ), the modelling components of the preflowering duration (Fig. 1). The dwarfing allele had an increasing effect on fo, 1, and , and a decreasing effect on 2 (Table 2). A second genome position of interest that is associated with all traits was mapped at 94.4–99.2 cM on chromosome 5(1H) (Fig. 1). The direction of this locus on the four traits was just opposite to that of the denso locus. Both loci were detected in a previous QTL study for preflowering duration measured over two years (Yin et al., 1999). The two loci common to all traits provide the genetic basis for the significant phenotypic correlations among the traits (Table 1).
The locus close to the upper terminal marker of chromosome 3(3H) and the locus at 33.6–36.7 cM of chromosome 4(4H) are both shared by traits 1 and (Fig. 1). Both loci had an increasing allele from ‘Prisma’ (Table 2). Another region, which is at 66.9–67.6 cM on chromosome 4(4H), is shared by 2 and (Fig. 1), with an increasing effect on 2 and a decreasing effect on from ‘Prisma’ (Table 2). These loci gave an additional reason for the high phenotypic correlations between 1 and , and between 2 and (Table 1). None of these three loci were previously detected for preflowering duration (Yin et al., 1999).
Other mapped QTL are specific to only one trait (Fig. 1). Three specific loci (126.0 cM on chromosome 2(2H), 71.5 cM on chromosome 3(3H) and 144.1 cM on chromosome 5(1H)) were mapped for fo. Two specific loci (160.8 cM on chromosome 3(3H) and 40.6 cM on chromosome 6(6H)) were mapped for , both with the increasing allele from ‘Prisma’. The locus at 54.7 cM on chromosome 5(1H) and the locus at 25.8 cM on chromosome 7(5H) were mapped specifically for 1 and 2, respectively. Among these specific QTL, only fo loci on chromosomes 2(2H) and 5(1H) were in close proximity to minor QTL detected previously for preflowering duration (Yin et al., 1999). fo, among the four parameters, is the closest to the preflowering duration in biological meaning. It is, therefore, not surprising that QTL identified here for fo agree best with those found previously for preflowering duration.
This analysis gives information about the magnitude of the combined effect of identified multiple QTL for the trait, based on the additive multiple-QTL model, equation (4). The model accounted for 71.0, 37.5, 33.6, and 41.2% of the phenotypic variation in fo, 1, 2, and , respectively (Table 2).
Coupling QTL effects to ecophysiological model for QTL-based predictions
The performance of the ecophysiological model, equations (1) and (2), with QTL-based input traits as estimated by using equation (4) was examined. The model using QTL-based values accounted for only 20–43% of across-RIL phenotypic variation of days to flowering per field environment, and for 46% of the variation among the RILs in across-environment mean days to flowering (Table 3). The QTL-based parameters had less accurate predictability than the original, purely phenotypic model parameters, in terms of either their performance per environment or their prediction of across-environment means (Table 3). Since the two sets of predictions were using the same environmental (i.e. daily temperature and photoperiod) inputs, model performance in predicting the variation due to environment (as reflected by across-RIL mean for eight environments) was not much changed by using QTL-based inputs (Table 3). The model predicted about 96% of variation in the across-RIL means.
|
The overall variation caused by both genotypes and environments can be shown, by pooling all data points of the 94 RILs across the eight environments. The QTL-based parameters accounted for 72% of the overall variation, again lower than the percentage (81%) accounted for by the use of the original phenotypic parameters (Fig. 3). In both cases, the model seems to over-predict the lower end of and to under-predict the higher end of observed days to flowering, and as a result, the range of predicted values was narrower than that of the observed days. However, this effect is a direct consequence of the shrinkage effect that is inherent to the fitting of models to phenotypic data (Yin et al., 2005
). The overall average days predicted by use of the original and the QTL-based parameters were both 144.5 d, close to the observed average value (145.6 d).
|
As an independent test, the observed days to flowering of the two parents in the eight environments were compared to those predicted using the QTL-based parameters (Fig. 4). The QTL-based parameters accounted for 94% of the variation, compared with 93% accounted for by the model using the original parameters for the two parents. For the independent test of their model for predicting leaf elongation rates in maize, Reymond et al. (2003)
used not only two parents but also some RILs that had not been included in the QTL analysis. Relative to theirs, this RIL population is smaller. All RILs of the population were included in the mapping analysis for a maximum power of QTL detection.
|
Predictions for days to flowering using a QTL-based ecophysiological model were directly compared with predictions coming from using the original, phenotypic parameters (Fig. 5). The QTL-based predictions correlated well with the original predictions (r=0.92). Similar good correlation was also obtained from earlier QTL-based predictions for barley grain yields (Yin et al., 2000a
).
|
Comparison with earlier QTL-based modelling
The better performance of the model using the original input trait values than the QTL-based model (Fig. 3; Table 3) is not surprising, because variation accounted for by identified QTL was only about 35–40% for most of model-input traits (Table 2). The result reported here for days to flowering differed from the previous report (Yin et al., 2000a
) in predicting grain yields in the same ‘ApexxPrisma’ population. In that report, the QTL-based crop growth model performed somewhat better (r2=0.524) than the model using originally measured traits did (r2=0.376). The better performance of the QTL-based model was simply attributed to the fact that random errors in the original traits were properly removed by the QTL statistics (Yin et al., 2000a). The result of the current analysis for days to flowering indicates that the answer may not have been that simple.
In the phenology model for predicting flowering time, equations (1) and (2), four input parameters (fo, 1, 2, and ) were used and they were all important, although to different extents, to explain differences in flowering time among the RILs (Yin et al., 2005). In the crop growth model used in the earlier study (Yin et al., 2000a, b) for predicting grain yield, six model-input traits were examined. Among them, only two were found to be important for predicting grain yield differences among the RILs, because use of the across-RIL mean of the other four input parameters resulted in better model predictions than use of measured RIL-specific parameter values (Yin et al., 2000b). The better performance of the QTL-based model in predicting grain yields could simply be due to the fact that relative to the original, phenotypic values, the QTL-based input traits had a narrower range of values for those four unimportant parameters, as they were being shrunk to their across-RIL means. This reasoning is supported by the result of a new simulation, in which a two-parameter crop growth model was used (fixing the other four, unimportant, parameters at their across-RIL means). When the measured phenotypic inputs for the two parameters were used, 64.8% of the variation in grain yield was accounted for. By contrast, for the crop growth model using the QTL-based inputs of the two parameters, this percentage was slightly lower, 62.1%. Reymond et al. (2003) also showed that the QTL-based predictions of maize leaf elongation rates were somewhat more dispersed than the original predictions. It could be a general phenomenon that the gain from the QTL statistics that removes, at least part of, random noise in the original model-input traits, may not be sufficient to compensate for the loss due to the residual genetic variance that is not captured by the identified QTL.
Comparison of QTL-based model predictions with those using QTL for days to flowering per se
To what extent QTL-based model predictions are close to the predictions using QTL for field flowering dates per se was examined next. To this end, the observed days to flowering per environment were subjected to QTL analysis using the method described earlier. One to four QTL were detected for the flowering time in each environment (Table 4). The genome positions around 120 cM on chromosome 2(2H) and at 126 cM on chromosome 3(3H), that turned out to be important in most environments, were in close proximity to the two major QTL found for the model-input trait fo (Table 2). Some QTL were detected in one or several, but not in other environments, indicating a ‘QTLxenvironment interaction’, a common finding when data for a quantitative trait observed in multiple environments are subject to QTL analysis (Jansen, 1995; Jiang and Zeng, 1995). The identified QTL accounted for 79.4, 77.1, 65.7, 69.2, 45.2, 33.8, 15.4, and 53.4% of the phenotypic variation of days to flowering among RILs in the eight field environments, respectively (Table 4).
|
When all data points for both RILs and parents from the eight environments were pooled, QTL for field observed days to flowering per se accounted for 85.5% of overall variation (Fig. 6A), somewhat higher than 72.6%, the percentage of overall variation accounted for by the QTL-based phenology model (Fig. 6B). The slightly better predictions using QTL for field observed days to flowering itself per environment is not surprising because ecophysiological QTL-based model predictions under field environments were completely independent of the greenhouse experiment from which the phenotypic model-input traits had been derived. Another reason is that random noise may have influenced the curve-fitting of the original phenotypic values of ecophysiological model-input traits. Nevertheless, the correlation between two sets of QTL-based predictions was high (r=0.89, Fig. 7), indicating that the simple ecophysiological phenology model captures a large part of the photothermal responses of genotypes in the population under the environmental conditions studied. Therefore, a robust ecophysiological model is capable of extrapolating (QTL) information from one environment to another.
|
|
|
Concluding remarks |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
Following our earlier research (Yin et al., 1999
, 2000a, b), the present study continued to examine the feasibility of combining ecophysiological modelling and genetic mapping to predict performance of individuals in an RIL population under various environmental conditions. The interest in conducting this across-disciplinary research was to explore the utilization of complementary aspects of these two research areas (Yin et al., 2003). QTL mapping allows the dissection of a phenotype into individual genetic factors (Paterson et al., 1988), but its ability to extrapolate QTL information from one set of environment/management conditions to independent new conditions is limited (Stratton, 1998). Ecophysiological modelling can potentially give an insight into how GxE comes about (Tardieu, 2003), but it does not account for the genetic basis of differences in response to environmental changes. Combining ecophysiological modelling and genetic mapping into a QTL-based crop model could be powerful for resolving complex environment-dependent yield traits on a genetic basis. In view of the weakness of current crop-growth models in predicting differences in grain or seed yields among similar lines of a segregating population (Yin et al., 2000a, b), the focus in the current analysis was on a relatively simple trait, days to flowering. This analysis, together with the work of Reymond et al. (2003) on another simple trait (leaf elongation rate), highlighted a great potential of this combined approach in predicting the performance of plants in a mapping population carrying any combination of alleles under a wide range of climatic scenarios. While ecophysiological modelling and genetic mapping have evolved independently so far, modellers, physiologists, geneticists, and molecular biologists may work to reach a synergy, a blending of scientific disciplines, to face the challenges posed by the increasing availability of genomic information to predict plant or crop phenotypes for complex traits. |
Acknowledgements |
|---|
This research was funded by the National Natural Science Foundation of China (contract No. 30060036). We thank Hui Li and Peilian Zhang for their assistance in managing the experiments, and Dr MP Boer for providing his program for the estimation of allelic QTL genotypes.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResults and discussionConcluding remarksReferences |
|---|
Finlay KW, Wilkinson GN. 1963. The analysis of adaptation in a plant-breeding programme. Australian Journal of Agricultural Research 14, 742–754.[CrossRef]
Jansen RC. 1995. Genetic mapping quantitative trait loci in plants: a novel statistical approach. PhD dissertation Wageningen Agricultural University, The Netherlands.
Jiang CJ, Zeng ZB. 1995. Multiple trait analysis of genetic mapping for quantitative trait loci. Genetics 140, 1111–1127.[Abstract]
Jiang CJ, Zeng ZB. 1997. Mapping quantitative trait loci with dominant and missing markers in various crosses from two inbred lines. Genetica 101, 47–58.[CrossRef][Medline]
Kearsey MJ, Pooni HS. 1996. The genetical analysis of quantitative traits. London: Chapman and Hall.
Kleinhofs A, Kilian A. 1994. RFLP maps of barley. In: Phillips RL, Vasil IK, eds. DNA-based markers in plants. Dordrecht, The Netherlands: Kluwer Academic Publishers, 163–198
Paterson AH, Lander ES, Hewitt JD, Peterson S, Lincoln SE, Tanksley SD. 1988. Resolution of quantitative factors by using a complete linkage map of restriction fragment length polymorphisms. Nature 335, 721–726.[CrossRef][Medline]
Peleman JD, van der Voort JR. 2003. Breeding by design. Trends in Plant Science 8, 330–334.[CrossRef][Web of Science][Medline]
Reymond M, Muller B, Leonardi A, Charcosset A, Tardieu F. 2003. Combining quantitative trait loci analysis and an ecophysiological model to analyze the genetic variability of the responses of maize leaf growth to temperature and water deficit. Plant Physiology 131, 664–675.
Ribaut J-M, Hoisington D. 1998. Marker-assisted selection: new tools and strategies. Trends in Plant Science 3, 236–239.[CrossRef]
Slafer GA. 2003. Genetic basis of yield as viewed from a crop phyiologist's perspective. Annals of Applied Biology 142, 117–128.[CrossRef]
Stratton DA. 1998. Reaction norm functions and QTL-environments for flowering time in Arabidopsis thaliana. Heredity 81, 144–155.
Tardieu F. 2003. Virtual plants: modelling as a tool for the genomics of tolerance to water deficit. Trends in Plant Science 8, 9–14.[CrossRef][Web of Science][Medline]
Van Eeuwijk FA, Denis JB, Kang MS. 1996. Incorporating additional information on genetics and environments in models for two-way genotype by environment interaction tables. In: Kang MS, Gauch HG, eds. Genotype-by-environment interaction. Boca Raton, Florida: CRC Press, 15–49.
Van Eeuwijk FA, Crossa J, Vargas M, Ribaut J-M. 2002. Analysing QTL by environment interaction by factorial regression, with an application to the CIMMYT drought and low nitrogen stress programme in maize. In: Kang MS, ed. Quantitative genetics, genomics and plant breeding. Wallingford: CAB International, 245–256.
Van Ooijen JW. 1999. LOD significance thresholds for QTL analysis in experimental populations of diploid species. Heredity 83, 613–624.
Van Ooijen JW, Boer MP, Jansen RC, Maliepaard C. 2002. MapQTL®4.0, software for the calculation of QTL positions on genetic maps. Wageningen, The Netherlands: Plant Research International.
Voorrips RE. 2002. MapChart: software for the graphical presentation of linkage maps and QTLs. Journal of Heredity 93, 77–78.
Weiss A. 2003. Introduction to the ‘Crop modeling and genomics’ symposium papers. Agronomy Journal 95, 1–3.
Yin X, Stam P, Dourleijn CJ, Kropff MJ. 1999. AFLP mapping of quantitative trait loci for yield-determining physiological characters in spring barley. Theoretical and Applied Genetics 99, 244–253.[CrossRef]
Yin X, Chasalow S, Dourleijn CJ, Stam P, Kropff MJ. 2000a. 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.[CrossRef][Web of Science][Medline]
Yin X, Kropff MJ, Goudriaan J, Stam P. 2000b. A model analysis of yield differences among recombinant inbred lines in barley. Agronomy Journal 92, 114–120.
Yin X, Stam P, Kropff MJ, Schapendonk AHCM. 2003. Crop modeling, QTL mapping, and their complementary role in plant breeding. Agronomy Journal 95, 90–98.
Yin X, Struik PC, Tang J, Qi C, Liu T. 2005. Model analysis of flowering phenology in recombinant inbred lines of barley. Journal of Experimental Botany 56, 959–965.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N. C. Collins, F. Tardieu, and R. Tuberosa Quantitative Trait Loci and Crop Performance under Abiotic Stress: Where Do We Stand? Plant Physiology, June 1, 2008; 147(2): 469 - 486. [Full Text] [PDF]
|
|
|
|
 |
|
 X. Yin, P. C. Struik, J. Tang, C. Qi, and T. Liu Model analysis of flowering phenology in recombinant inbred lines of barley J. Exp. Bot., March 1, 2005; 56(413): 959 - 965. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||