| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Crop Sci., Univ. of Illinois, 1102 S. Goodwin Ave., Urbana, IL 61801
b Monsanto SAS, Route D'Epincy, Louville, La Chenard 28150, France
c 311 N. Osborn Lane, Aberdeen, MD 21001
* Corresponding author (jdudley{at}uiuc.edu).
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: BLUP, best linear unbiased predictor • IHO, Illinois High Oil • IHP, Illinois High Protein • ILO, Illinois Low Oil • ILP, Illinois Low Protein • MR, multiple regression • QTL, quantitative trait loci • RFLP, restriction fragment length polymorphism • SF, single factor • SSR, simple sequence repeat • TC, testcross
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Dudley (1993) suggested the use of randomly mated progeny to increase recombination between markers, between QTL, and between markers and QTL to increase resolution. Liu et al. (1996) showed empirically that much higher resolution could be obtained among tightly linked markers after four generations of randomly mating an F2. Darvasi and Soller (1995), using simulation, demonstrated a reduction in the confidence interval surrounding a QTL by successive generations of random mating. Lee et al. (2002), with 190 RFLP loci, showed a nearly four-fold increase in genetic map distance following five generations of intermating the cross of maize inbreds B73 and Mo17. Recombinant inbreds were derived from this population after four generations of randomly mating the F2 population. Lee et al. (2002) suggested this resource should connect research on dense genetic maps, physical mapping, gene isolation, comparative genomics, and analysis of QTL. Winkler et al. (2003) showed the relationship between crossover rates in an F2 and expected recombination in a population of recombinant inbreds developed from a randomly mated population tracing to the F2. Lu et al. (2003) used random mating to break up repulsion linkages in an attempt to separate pseudooverdominance from true overdominance for grain yield in maize.
Crosses between very divergent parents are desirable to determine the genetic architecture of a trait and identify many of the QTL affecting the trait. The IHP and ILP strains of maize are ideally suited for this purpose. These strains were divergently selected for high and low protein concentration in the grain beginning in 1896 (Dudley and Lambert, 2004). At the time of initiation of the experiment reported in this paper, ILP had reached a lower limit of protein concentration, although progress was still being made in IHP. In a Design III study with the cross of generations 70 of IHP and ILP, Dudley (1994) demonstrated a significant reduction in additive genetic variance for protein concentration following four generations of random mating. This result is what is expected with coupling phase linkages (Comstock and Robinson, 1948).
Increasing protein concentration in maize grain usually results in reduced starch concentration (Dudley and Lambert, 2004). Thus, in generations 70, the generations from which the parents of this study came, IHP (440 g kg–1 starch, 261 g kg–1 protein) and ILP (745 g kg–1 starch, 58 g kg–1 protein) (Dudley and Lambert, 1992) were low and high starch strains as well as being high and low protein strains. In a previous study, Goldman et al. (1993) found 19 significant marker–QTL associations for starch, of which 16 were also significant for protein. Whether these shared associations were the result of linkage among trait-specific QTL or single QTL with pleiotropic effects could not be determined. Information on the genetic control of starch and protein is desirable because of the importance of increased starch concentration, with accompanying reduced protein concentration, for increasing efficiency of ethanol production, and because of the importance of increased protein concentration in grain for cattle feeding.
The objectives of this study were (i) to determine the effects of random mating on the ability to identify marker–QTL associations in the cross of generations 70 of IHP and ILP; (ii) to compare S1 per se performance with performance of TCs to two different testers for ability to identify marker–QTL associations; and (iii) to use this information to identify chromosomal regions containing QTL controlling starch, protein, and oil concentration in maize grain.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Evaluation
In 1994 and 1995, the 200 Syn0 S1 and 200 Syn4 S1 lines were evaluated as lines per se in adjacent experiments at the Crop Sciences Research and Education Center near Urbana, IL. Each year, each set of lines was grown at two different planting dates approximately 3 wk apart to provide two environments in each year. To control environmental variation within a year–environment combination, families were arranged in a generalized lattice [
(0,1)] design with two replications of 20 incomplete blocks of 10 entries each (Carmer, 1985). Plots of the Syn0 experiment were grown adjacent to those of the Syn4 experiment. Plots consisted of a single 4.5-m row with 0.76 m between rows. Plots were overplanted and thinned to 15 plants per row. In each row, six plant-to-plant crosses were made to produce seed for chemical analysis. Ears from the hand-pollinated plants were harvested by hand.
In 1996 and 1997, TCs to FR1064 and FR616 of the 200 Syn0 and Syn4 lines were grown. Testcrosses were grown at two locations near Urbana, IL, each year and the Syn0 and Syn4 TCs to a particular tester were grown in adjacent experiments. Each year–location combination was arranged in a generalized lattice [(0,1)] design with two replications of 20 incomplete blocks of 10 entries each. Plots consisted of 2 rows 5.3 m long with 0.76 m between rows. Plots were overplanted and thinned to 23 plants per row. Plots were machine-harvested and a sample of grain collected from the combine for chemical analysis. Grain was weighed and moisture data collected on the combine.
Protein, starch, and oil concentrations in the grain were measured with a DICKEY-john near-infrared analyzer (Hymowitz et al., 1974), as described by Dudley and Lambert (1992). Thus, the phenotypic data set consisted of 14400 data points.
Molecular Marker Data
Molecular marker genotypes for the 200 S1 lines from the Syn0 and the 200 lines from the Syn4 were obtained for 44 RFLP and 20 SSR markers. DNA was isolated from bulks of tissue from 25 random plants from each S1 line according to established procedures (Mikkilineni, 1997). The RFLP procedures (Berke and Rocheford, 1995; Dijkhuizen et al., 1998) and SSR procedures (Wong et al., 2003) used have been described previously and are based on protocols reported by Hoisington et al. (1989) and Senior et al. (1996). The RFLP and SSR probes are part of larger sets listed in the Maize Genomics and Genetics Database (http://www.maizegdb.org/; verified 4 Mar. 2004). In addition, the cDNA clone of shrunken2 (sh2) (Bhave et al., 1990) was used as a probe.
Markers were chosen on the basis of polymorphism between IHP and ILP (based on differences in allelic frequencies between IHP and ILP as measured on a sample of plants from generations 70), previous association with QTL for kernel composition in the cross of generations 76 of IHP and ILP (Goldman et al., 1993), and coverage of all chromosome arms in the maize genome. Because the parents were not homozygous and the allelic frequencies in the parents were not known, it was not possible to assign alleles on the basis of their origin as being from IHP or ILP. Consequently, the fragments were simply scored by relative molecular weight and given a letter designation of A, B, C from highest molecular weight to lowest. For some markers, different single alleles were fixed in IHP and ILP and designations were simply A and B, scored in the same manner.
Statistical Analysis
Analyses of variance were done and relevant variance components estimated for each generation (Syn0 and Syn4) of each progeny type (S1 progeny, TC to FR1064, TC to FR616). PROC MIXED in SAS (SAS Institute, 1999) was used for statistical analysis assuming years and environments (planting dates for S1's; locations for TCs) (E) were fixed effects. Blocks (B), reps (R), and lines (L) were considered random. The interactions of lines with years (LY), lines with environment (LE), and lines x years x environments (LYE) were considered random. Variance components and their 95% confidence intervals were estimated from PROC MIXED by the = 0.05 option (SAS Institute, 1999).
Heritabilities (H2), for each generation of each progeny type, were calculated from the equation H2 = 
2g/
(Bernardo, 2002). Confidence intervals (90%) for H2 were calculated as described by Knapp et al. (1984). Best linear unbiased predictor (BLUP) estimates for protein, starch, and oil concentration were obtained for each line for each progeny type by PROC MIXED.
A procedure described by Zehr et al. (1992) was used to test for significance of marker–QTL associations because many of the markers were segregating for multiple alleles and allelic frequencies were not 0.5 in the F1 generation. Briefly, each allele for each marker was analyzed separately. Homozygotes for allele A were designated AA. All families heterozygous for allele A were designated AX regardless of the designation of the alternative allele. Genotypes not AA or AX were designated XX. The BLUP estimates were then used in a SF ANOVA to identify significant associations of markers with QTL for protein, oil, or starch. This procedure tests for a QTL linked to allele A. A similar procedure was used for alleles B, C, etc. for each marker. When one or more alleles showed a significant effect at the 0.05 significance level for a marker, the allele with the lowest probability for significance was selected to represent that marker with the restriction that there had to be at least 10 individuals in each marker genotype class. For markers with only two alleles, the probability for each allele was the same and the A allele was selected in every case. Because the parents were not homozygous and the allelic frequencies in the parents were not known, it was not possible to assign alleles on the basis of their origin as being from IHP or ILP.
Multiple regression models were developed only for starch and protein because of the low genetic variance and small number of significant effects from the SF analysis for oil. All markers with at least one allele significant at the 0.05 level for starch in at least one of the progeny types were selected as the set to be used in MR analysis. The same set of markers was used for analysis of starch and protein because the markers selected based on starch included all the markers with significant effects for protein. This procedure reduced the set of markers included in the regression analysis to 42.
The stepwise option in PROC REG in SAS (SAS Institute, 1999), with a probability of entering or dropping a marker as 0.15, was used in MR analysis. Because missing marker data reduced the number of families available for MR analysis, a sequential procedure was used in that analysis. First, an analysis was done for each chromosome separately. The markers selected from each chromosome in this step were then entered into a similar analysis across all chromosomes to produce the final model. This procedure reduced the number of markers entered into any one analysis, thus maximizing the number of families available for QTL identification.
Effects of random mating were measured by comparing means, genetic variances, heritabilities, and proportion of genetic variance accounted for by the best multiple regression model (p) measured in the Syn0 and Syn4. Estimates of p were obtained by dividing R2 (proportion of phenotypic variability accounted for by the multiple regression model) by H2. Effects of random mating on marker–QTL associations were measured by comparing the number of marker–QTL associations significant at the 0.01 probability level in the Syn0 with the number significant in the Syn4 for each trait and progeny type. Significance of changes in number of significant marker–QTL associations between the Syn0 and Syn4 was measured by a contingency table 
2 test (P < 0.05) adjusted for continuity (Snedecor and Cochran, 1989). Marker allelic frequencies in the Syn0 and Syn4 were compared with standard errors of allelic frequencies calculated as described by Weir (1996). Chi-square analysis (Weir, 1996) was used to test for agreement of genotypic frequencies with random-mating expectations in the Syn0 and Syn4. For tests of changes in allelic frequency between generations, the marker loci and alleles used in the MR analysis were used and the genotypes associated with those alleles were used in the 2 analysis. Linkage disequilibrium among markers in each generation was measured by the correlations among markers within each generation. This procedure is equivalent to the method described by Weir (1996)(p. 137). Correlations were calculated for the set of 42 markers used in MR analysis.
Phenotypic correlations among traits were calculated for all possible combinations of traits for each progeny type and generation by PROC CORR in SAS. BLUP estimates of entry values were used in the calculations. Genotypic correlations were calculated with a SAS program developed by Holland (personal communication, 2003, available at http://www4.ncsu.edu/%7Ejholland/correlation/correlation.html). This procedure uses PROC MIXED with the ASYCOV option to calculate genetic variances and covariances between traits. The output is then used in an IML subroutine to calculate genetic correlations and their standard errors by the procedure of Mode and Robinson (1959). Markers showing significant associations for both starch and protein in either the SF or MR analysis were identified to determine the extent of common genetic control of these traits.
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
|
|
Heritabilities for protein and starch ranged from 0.81 to 0.93 with no significant difference between the Syn0 and Syn4 and no difference among progeny types (Table 4) based on overlapping 90% confidence intervals. For oil, Syn0 and Syn4 heritabilities did not differ significantly. Thus, even though the genetic variance was reduced from the Syn0 to the Syn4 for starch and protein in the S1 and both TCs and for oil in the S1 progenies, heritabilities did not change because the interaction components in the denominator of the heritability equation were also reduced (data not shown).
Allele and Genotypic Frequencies, Linkage Disequilibrium
Significant (P < 0.05) differences in allele frequency between the Syn0 and Syn4 were found for 18 of the 42 markers (Table 5) used in MR analysis. Significant (P < 0.05) deviations from random mating as measured by 2 were found for 14 markers in the Syn0 and five markers in the Syn4. For 11 markers with significant 2 values in the Syn0, the deviation from random mating was not significant in the Syn4. This result might be expected with random mating because markers not in equilibrium in the Syn0 would be expected to go to equilibrium with random mating. On the other hand, there was only one marker with a significant deviation from random-mating equilibrium in the Syn4 which had a nonsignificant deviation in the Syn0.
|
QTL Analysis Results
Syn0 vs. Syn4—Number of Significant Marker–QTL Associations
A major objective of this study was to compare results of QTL analysis from the Syn0 and Syn4 generations. The Syn4 was expected to have fewer significant marker trait associations than the Syn0. The reasons for this can be illustrated by three examples. In the first example, assume marker density is low. Where marker density is low, a marker may be loosely linked with a QTL having a large effect. In the Syn0, linkage disequilibrium is high and this marker will show a significant effect. Because random mating reduces linkage disequilibrium and the size of unbroken linkage blocks (Hanson, 1959), disequilibrium between a loosely linked marker and a QTL may be reduced to the point that the marker QTL association is no longer detectable. In the second example, assume a region with 10 markers in, for example, a 20-cM distance. Within that region, assume there are two QTL linked in coupling, each with a relatively small effect. In the Syn0 all 10 markers may show significant marker QTL effects. After random mating, perhaps only four of the markers (two flanking each QTL) may show significant effects. Thus, the number of significant marker effects detected is reduced. As a third example, QTL effects measured in the Syn0 could be the result of blocks of genes linked in coupling with a marker located near this block. The effect of each individual gene may be too small to be detectable. When the block of genes is broken up by random mating, the marker QTL association is no longer detectable. The second and third examples are based on coupling phase linkages as is expected in the IHP x ILP cross. If two QTL are linked in repulsion and their effects cancel, random mating may break up the linkage and allow identification of two QTL where none were identified before random mating.
For comparisons of the Syn0 with the Syn4, the number of marker trait associations significant at P < 0.01 was used. This P value was chosen arbitrarily as a compromise between a very stringent P value and a more relaxed P value. The expectation of fewer significant associations in the Syn4 was realized because for protein there were 38, 32, and 32 markers significantly (0.01 probability level) associated with QTL in the Syn0 for the S1, FR1064, and FR616 progenies, respectively, but only 13, 7, and 3 in the Syn4 (Table 6). Similarly, for starch there were 35, 30, and 29 significant associations in the Syn0, and 9, 3, and 6 in the Syn4 for the S1, FR1064, and FR616 progenies, respectively. For each progeny type for starch and protein, the contingency table 2 value (Table 6) comparing the Syn0 and Syn4 was highly significant (P < 0.01), indicating a significant reduction in the number of significant marker QTL associations. For oil, only 6, 3, and 8 markers were significant in the S1, FR1064, and FR616 progenies, respectively, in the Syn0, and only 3, 2, and 0 for the Syn4. Chi-square values for oil were not significant for the S1 and FR1064 progenies. In evaluating these results, it is important to remember that the 200 S1 lines used in the Syn4 are independent of those used in the Syn0 because the sample of parents used to continue random mating from the Syn0 was not the set of lines used to measure effects in the Syn0. In addition, the fact that the standard errors of the means of the Syn0 and Syn4 were very similar (Table 1) and the number of progenies used was the same in both generations suggests that the smaller number of QTL detected in the Syn4 are not the result of a difference in precision between the Syn0 and Syn4.
|
Syn0 vs. Syn4—Variability Accounted for in Multiple Regression Analysis
Results from the MR analysis agreed with the SF results. For protein, the final genetic models in the Syn0 accounted for 68 to 73% of the genetic variance. For starch, 52 to 75% of the genotypic variance in Syn0 was accounted for (Table 7). However, for the Syn4 the final regression models accounted for only 12 to 18% of the genetic variance for protein and 3 to 20% of the genetic variability for starch. Thus, with the limited number of genetic markers available in this study, no marker model accounted for a high proportion of the genetic variance in the Syn4. In addition, the numbers of markers included in the final models for starch and protein in the Syn4 were less than half the numbers included in the models for the Syn0 (Tables 8 and 9).
|
|
|
2 tests or between the FR1064 and FR616 progenies (results not shown). The ability of the TC progenies to detect a similar number of significant effects as the S1 progenies, when the magnitude of the effects is expected to be half that of the S1's, resulted from greater precision in the TC data as evidenced by the standard error of means of the TCs being approximately half the standard errors of S1 means (Table 1). In the Syn0, agreement among progenies as to the markers found significantly associated with traits was high. For both protein and starch, 30 of the 38 markers significant (P < 0.01) in the Syn0 for S1 progenies were also significant in the FR1064 and FR616 progenies (Tables 8 and 9). Only two markers significant in either the FR1064 or FR616 progenies were not significant in the S1 progenies. Twenty-four markers were significant for all three sets of progenies in the Syn0. Agreement was not as good for the Syn4 progenies likely because of the small number of significant markers in the Syn4 for any progeny type.
Identification of QTL
Criteria for declaring a marker associated with a QTL included significance (P < 0.01) of marker trait associations in the SF analysis of the Syn0 for all three progeny types, retention in the Syn0 multiple regression model, significance in the SF analysis in the Syn4 of markers found significant in the Syn0, or inclusion in the Syn4 multiple regression model. The number of markers used was too small to make the Syn4 results definitive, thus the Syn4 results were used to reinforce the interpretation of the Syn0 results. On the basis of these criteria, markers npi262 (bin 1.04), bmc1643 (bin1.08), npi337 (bin 2.06/07), bmc1144 (bin3.03), bnl5.37b (bin 3.06), umc63a (bin 3.09), php20020 (bin 7.06), npi260b (bin 8.03), umc2c (bin8.05), php20075b (bin 9.03), umc130 (bin 10.03), and bmc1185 (bin 10.07) are considered highly likely to be linked to QTL for protein. In addition to being significant for all three progeny types in the Syn0, markers npi262, bmc1643, bnl5.37b, php20020, and umc2c were included in the final Syn0 multiple regression models for all three progeny types (Table 8). For bmc1643, the Syn4 data were also significant (P < 0.01) in the SF analysis for the S1 progenies. Marker npi337 was in the Syn0 final multiple regression model for the S1. Marker umc63a was included in the final Syn0 regression models for the S1 and FR616 progenies and was significant in the Syn4 SF analysis for FR1064. Markers umc130 and bmc1185 were also in the final multiple regression models for FR616 and FR1064 progenies, but not in the S1 progenies. Marker npi260b was included in the list of marker QTL associations highly likely to be linked to QTL for protein, even though it was not significant in all three progeny types in the Syn0 because it was significant in all three progeny types in the Syn4.
With the same criteria as for protein, markers bmc1953 (bin 1.02), npi262 (bin 1.04), umc67a (bin 1.06), bmc1643 (bin 1.08), umc6a (bin 2.03), npi298 (bin 2.08), bnl17.14 (bin2.10), bmc1144 (bin 3.02), umc154 (bin 3.04), bnl5.37b (bin 3.06), php20726 (bin 3.09), umc63a (bin 3.09), php20020 (bin 7.06), bnl9.08 (bin 8.03/04), npi260b (bin 8.03), umc2c (bin 8.05), umc130 (bin 10.03), and bmc1185 (bin 10.07) were considered highly likely to be linked to QTL for starch. Markers bmc1953, npi262, bmc1643, and php20020 were included in the final multiple regression models in the Syn0 for all three progeny types. Marker npi260b was significant in the Syn4 SF analysis for all three progeny types.
Another indication of whether a marker is highly likely to be linked to a QTL is agreement with results of other studies. Eighteen of the RFLP markers used in this study were also used in a study of the cross of generations 76 of IHP and ILP (Goldman et al., 1993). Five of those markers were significantly associated with protein concentration in the Goldman et al. (1993) study. Of those five, three (php10080, sh2, and umc63a) were significantly (P < 0.01) associated with protein in the S1 and both TC progenies in the Syn0 of this study. All three markers are in bin 3.09 (Table 8). Of three markers significantly associated with starch, in the Goldman et al. (1993) study, two (sh2 and umc63a) were also significant for the S1 and both TCs in the Syn0 (Table 9) of our study.
In a study to compare the relationship between a micro wet-milling procedure and the near-infrared reflectance method used in this study, Dijkhuizen et al. (1998) presented results for 26 RFLP markers used in this study. The phenotypic data were the 1995 S1 progeny data. Thus, the Dijkhuizen et al. (1998) results were from a subsample of the data reported herein. Twenty-six RFLP markers were used in both studies. Of the 26 markers in common, 20 showed significant effects in both studies, four were not significant in either, and one was significant in this study and not in Dijkhuizen et al. (1998), while one was significant in the Dijkhuizen et al. (1998) study and not here. Thus results from the one year of data were consistent with the total set of two year's data reported in this study.
Agreement between Traits
The same markers are expected to be associated with significant effects for both starch and protein because of the high negative correlations between starch and protein. In the Syn0, 23 markers were significant at the 0.01 level for both traits for all three progeny types (Tables 8 and 9). To select for increased starch and protein simultaneously, the signs of the additive effects should be in the same direction. As expected, based on the genetic correlations between traits, most of the signs of the additive effects for protein were in the opposite direction from the signs for starch and this result was consistent across all three types of progeny. However, for umc67a in bin 1.06, npi238 in bin 1.11, and bnl17.14 in bin 2.10, the signs were the same for the additive effects for both protein and starch for all three progeny types. Thus, these regions may be of interest for developing high-protein, high-starch genotypes.
|
CONCLUSIONS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
To the authors' knowledge, this is the first report of the effects of random mating on identification of QTL in plants. From a cross for which the parents differed drastically for protein and starch, large numbers of QTL should have been segregating. Use of such a cross assured the presence of coupling phase linkages for protein and, given the high negative correlation between protein and starch, for starch as well. Random mating drastically reduced the number of significant marker–QTL associations. This result was expected because of the breakup of marker QTL linkages and of linkages between QTL. Accompanying the reduction in number of significant effects was a drastic reduction in the proportion of the genetic variability accounted for by multiple regression models. This reduction is likely because of the small number of markers available. This observation points to the value of emerging common resource populations such as the Intermated B73 x Mo17 (IBM) population that has more than 1200 molecular markers. This type of new resource with extensive marker coverage of the genome will enable more precise identification of QTL in randomly mated populations.
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Received for publication September 2, 2003.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J.W. Dudley From Means to QTL: The Illinois Long-Term Selection Experiment as a Case Study in Quantitative Genetics Crop Sci., December 18, 2007; 47(Supplement_3): S-20 - S-31. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. W. Dudley, D. Clark, T. R. Rocheford, and J. R. LeDeaux Genetic Analysis of Corn Kernel Chemical Composition in the Random Mated 7 Generation of the Cross of Generations 70 of IHP x ILP Crop Sci., January 22, 2007; 47(1): 45 - 57. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Clark, J. W. Dudley, T. R. Rocheford, and J. R. LeDeaux Genetic Analysis of Corn Kernel Chemical Composition in the Random Mated 10 Generation of the Cross of Generations 70 of IHO x ILO Crop Sci., February 24, 2006; 46(2): 807 - 819. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. A. Barrett, I. J. Baird, and D. R. Woodfield A QTL Analysis of White Clover Seed Production Crop Sci., August 1, 2005; 45(5): 1844 - 1850. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. M. Wilson, S. R. Whitt, A. M. Ibanez, T. R. Rocheford, M. M. Goodman, and E. S. Buckler IV Dissection of Maize Kernel Composition and Starch Production by Candidate Gene Association PLANT CELL, October 1, 2004; 16(10): 2719 - 2733. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
and heritability (H2) estimates for protein, starch, and oil.