Current Location:Home > Technique

Technique

Application of AMMT Model in Crossbreeding of Castor

 

Application of AMMT Model in Crossbreeding of Castor

ZHANG Bao-xian, WANG Guang-ming, TAN De-yun, LIU Hong-guang, SUN Li-juan

Zibo Agricultural Science Academy, Zibo of Shandong 255033

Abstract: [Objective] The paper analyzed production adaptability and stability of 4 hybrids of castor, so as to explore the application method of AMMI model in variety evaluation. [Method] AMMI model and biplot were used to carry out stability analysis on 4 castor combinations in multi-point test during 2008 and 2009, and further evaluated the stability and adaptability of various tested combinations. [Result] AMMI model analysis showed that the singular values IPCA1 and IPCA2 in principal component axis of interaction totally explained 92.24% of the interaction variance. AMMI model not only maximizes the reflection on variation of interaction, but also can accurately analyze the adaptability and yield stability of varieties. [Conclusion] AMMI model integrates variance analysis and principal component analysis together, which improves accuracy compared with the traditional method. Even the adaptability of female line without fertility could be indirectly evaluated through the identification on the genotype and environment interaction effects in F1 generation.

Key words: Castor; AMMI model; Biplot; Variety stability; Hybrid

Castor is an important oil crop with special purpose and high economic value. The crossbreeding and genetic research work for castor at abroad was carried out early, which began in 1950s [1]. Castor breeding in China started late, a batch of disease-resistant varieties with high yield and good quality were bred by means of system selection and crossbreeding during 1980s and 1990s [2-4]. The yield of “Zibo castor No.7” successfully bred by Zibo Agricultural Science Academy reached 4796.25 kg/hm2 in regional trail and production test in Shandong Province during 2008 and 2009[5]. Although China has made significant progress in the aspect of crossbreeding of castor, there are many deficiencies in varieties assessment and adaptation research of female line. As an indispensable part of castor breeding, years of multi-point tests for new varieties of castor aim at evaluating the production capacity of new varieties or combination and their adaptability to the environment, scientific analysis of test results is very important for evaluation on productivity and adaptability of tested varieties [6-9]. However, analysis of variance method still used for varieties evaluation can not well reflect the interaction effects between genotype and environment.

Additive main effects and multiplicative interaction model (AMMI) integrates variance analysis and principal component analysis into a model, which not only maximizes the reflection on variation of interaction, but also can accurately analyze the stability of varieties, being the current international popular model for analysis of regional test data of crops. Many domestic scholars have used AMMI model for analysis of regional test data of many crops such as rice, wheat, canola, cotton and corn, etc. AMMI model has thorough analysis on interaction between environment and genotype, which provides a better analysis method for the study of interaction between environment and genotype and stability of varieties, its application range is much wider and effective compared with variance analysis model and linear regression model [10-12].

The stability of varieties is mainly determined by the size of genotype and environment (G × E) interaction effects. Genotype and environment interaction effect is a common complex biological phenomenon, and there are many difficulties to reveal its rule. Using effective G × E analysis method has a crucial role in correct evaluation of stability of varieties. Taking test data of two years of multi-point variety comparison of castor hybrids during 2008 and 2009 as analysis targets, AMMI model and bisite were used to carry out stability analysis on varieties, so as to have a more objective evaluation on tested varieties[13 -14].

1 Materials and Methods

1.1 Test time and sites

The test was carried out from April to November during 2008 and 2009, and three test sites were set in Zhangdian District of Zibo City, Qingzhou City and Zouping County, respectively.

1.2 Test materials

A total of 4 combinations were used in the test, including 539, 428, 283 and 162, respectively. The male parent in various combinations was the same inbred line, and the female parents were different female lines. The tests were numbered as No.1, No.2, No.3 and No.4.

1.3 Experimental design

The tests were carried out according to a unified program, the treatments were randomly arranged, and each treatment contained three repeats. The plot was ​​45 m2 in area, four-row, with row length 9 m, row spacing 1.25 m, plant spacing 0.8 m, density 10005 plant/hm2. Protection lines were set around the test area. The test sites should be representative with consistent previous crops, the total area should be not less than 1200 m2. The field management of castor was in accordance with local habits, and the same management measure should be conducted within the same day.

1.4 Data processing

The interaction effect was significant after complex variance analysis[15]. On this basis, the interaction terms were decomposed, and analyzed by AMMI model. Then they were used to predict the yield performance of various tested crossbreed combinations in each site, and their stability was evaluated. Data analysis was conducted using Excel 2003 and DPS data processing system.

2 Results and Analysis

2.1 Determination results of yield

The survey results of yields in three sites within two years were shown in Table 1. The average yields of tested combinations No.1, No.2, No.3 and No.4 were 4825.18, 4107.40, 3902.22 and 4633.33 kg/hm2, respectively.

2.2 Statistical test

2.2.1 Complex variance analysis and linear regression model analysis

Variance analysis showed that genotype, site and interaction between them all reached a very significant level (Table 2). So the following results could be obtained: (1) the yields among varieties had real difference; (2) there were significant differences between sites, indicating that the choice of sites was representation; (3) square sum of interaction between genotype and sites (year) was extremely great, indicating that stability analysis was necessary.

Table 1   Survey results of crossbreed combination of castor in three sites within two years

Years

Test sits

No.1(kg/45m2)

No. 2(kg/45m2)

No. 3(kg/45m2)

No. 4(kg/45m2)

2008

Zhangdian

22.08

21.92

21.92

19.25

19.60

19.30

17.64

17.49

17.72

21.49

21.40

21.47

Qingzhou

21.63

21.81

21.50

17.72

17.91

17.88

18.26

18.22

18.35

22.36

22.43

22.46

Zouping

22.17

22.27

22.25

18.49

18.39

18.43

16.48

16.61

16.56

21.49

21.66

21.67

2009

Zhangdian

21.42

21.47

21.33

17.46

17.81

17.62

18.56

18.85

18.77

21.03

21.31

21.42

Qingzhou

20.92

20.87

21.03

19.23

19.10

19.31

16.47

16.37

16.40

19.85

19.90

19.92

Zouping

21.99

22.13

22.11

18.21

18.45

18.57

17.77

17.76

17.83

18.43

18.44

18.58

Average

21.70

21.75

21.69

18.39

18.54

18.52

17.53

17.55

17.60

20.77

20.86

20.92

Regression analysis in Table 2 showed that joint regression, gene regression and environment regression could explain 49.41% of square sum of genotype × environment interactions (proportion in square sum of interaction). The residual was still large, accounting for 50.59%, and it was extremely significant, indicating that regression model explained a little about interaction. Therefore, the regression model did not fit the experimental data well.

2.2.2 AMMI model analysis

AMMI model analysis showed that the singular value IPCA1 in principal component axis of interaction explained 57.68% of square sum of genotype × environment interactions (proportion in square sum of interaction), while IPCA2 explained 34.55% (Table 2). IPCA1 reached the extremely significant level at 0.01, and IPCA2 reached significant level at 0.05. IPCA1 and IPCA2 totally explained 92.24% of interaction variance, and the residual was only 7.76% of square sum of genotype × environment interactions. This indicated that AMMI model could greatly improve the accuracy of analysis compared with linear regression model.

Table 2  Joint variance analysis, linear regression analysis and AMMI model analysis results of yields in different sites

Source of variation

DF

SS

MS

F value

P value

Analysis of variance

Total variance

71

260.0055

 3.6620

 

 

Treatment

23

259.4253

11.2794

 933.1426

0.0001

Gene

3

205.4238

68.4746

5664.9098

0.0001

Environment

5

 10.5280

 2.1056

 174.1960

0.0001

Interaction

15

 43.4735

 2.8982

 239.7713

0.0001

Error

48

  0.5802

 0.0121

 

 

Regression analysis

Joint regression

1

 1.5599

1.5599

129.0481

0.0001

Gene regression

2

14.0635

7.0317

581.7363

0.0001

Environment regression

4

 5.8562

1.4640

121.1200

0.0001

Residual

8

21.9940

2.7493

227.4461

0.0001

Error

48

 0.5802

0.0121

 

 

AMMI model analysis

PCA1

7

25.0773

3.5825

3.1853

0.0075

PCA2

5

15.0222

3.0044

2.6713

0.0329

Residual

3

 3.3741

1.1247

 

 

Error

48

 0.5802

0.0121

 

 

2.2.3 Bisite analysis

Fig. 1 was AMMI1 bisite with average yield of varieties and sites as abscissa and IPCA1 value as vertical axis, the bisite simultaneously expressed the high yield of varieties and G × E information. As shown in Fig. 1, varieties was more scattered in horizontal direction than sites, indicating that the variation of varieties was greater than that of sites within the range of selected test sites. The vertical direction showed the difference of G × E, the closer the variety icon to zero value of IPCA1, the better the stability. The sequence of yield stability and adaptability of various combinations successfully was No.3 > No.1> No.2> No.4.

Fig. 1 also showed the adaptive information of varieties to different sites (years). The varieties around the zero horizontal line had positive interaction with the sites in the same side, which had negative interaction with the sites in the other side. The crossbreed combinations of No. 1 and No. 2 had positive interaction with Zhangdian site in 2008, Zouping and Qingzhou sites in 2009, which had negative interaction with Zhangdian site in 2009, Zouping and Qingzhou sites in 2008 (Fig.1). This indicated that the above combinations had good adaptability to climate conditions in Zhangdian site in 2008, Zouping and Qingzhou sites in 2009, while No.3 and No. 4 crossbreed combinations had positive interaction with Zhangdian site in 2009, Zouping and Qingzhou sites in 2008. The performance of the same combination in the same site between years was different, indicating that the meteorological factors among different years were different.

 

Fig. 1 Biplot of average yield with varieties and sites

 

Fig. 2 was the IPCA1 and IPCA2 bisite of variety and site. The length between test points and origin in Fig. 2 showed the size of interaction between varieties and sites, if the test point was farther away from the origin, it had greater contribution to the overall interaction. The test sites with large interaction in the test included Zouping in 2009, Qingzhou in 2009 and Qingzhou in 2009, while the test sites with smaller interaction were Zhangdian in 2009, Zouping in 2008 and Zhangdian in 2008. The closer the variety to origin was, the better stability of the variety was. So varieties of No. 1 and No. 3 were the most stable; variety No. 2 was far away from the origin, indicating it was more sensitive to environment; No. 4 was the furthest from the origin, which was most sensitive to environment.

Overall, the sequence of stability of 4 crossbreed combinations successfully was No.3> No.1> No.2 > No.4.

Fig. 2 Biplot of yields and sites with PCA1 and PCA2

 

3 Conclusions and Discussion

(1) AMMI model integrates variance analysis and principal component analysis together, which improves accuracy compared with the traditional method. It not only can better explain the gene-environment interaction during yield formation of castor, but also can better interpret the interaction between genotype of other relevant traits and environment. It also has important reference value for understanding the clear relationship between test sites, developing breeding objectives and demonstration of improved varieties.

(2) The performance of combinations in three test sites among different years is also different. Combinations No. 1 and No. 2 have positive interaction with Zhangdian site in 2009, while they have negative interaction with the same site in 2008; combinations No. 3 and No. 4 just have the opposite effect. This indicates that the performance difference among varieties mainly comes from the impact of meteorological factors in different sites, which is the response of genotype of variety to climate conditions. Various combinations have better adaptability to the sites and years with positive interaction, and the result is consistent with production practice.

(3) Four combinations used in the test are come from the same male parent and different female parents. Due to the poor fertility of female parent, they even have no fertility. Therefore, using traditional variety comparison methods can not identify their adaptability. The test has analyzed the genotype and environment interaction effects of hybrid combinations from the same male parent and different female parents to indirectly evaluate the adaptability of female line, which provides basis for breeding excellent parental combination with wide adaptability and high yield. This will be further verified in the next step.