GxE Interactions Analysis of Wheat Genotypes Evaluated Under Peninsular Zone of the Country by AMMI Model

: AMMI analysis of wheat genotypes had highlighted significant effects of environments, interactions and genotypes for the 2017-18 and 2018-19. Number of adaptability measures had been studied as per utilization of number of significant interaction principal components (IPCs). Total of interaction variations exploited by Type-1, 2, 3, 4 & 5 measures were 45.5%, 66.3%, 75.9% & 88.4% respectively. Type-1 measures EV1, D1, ASTAB1 identified (G7, G6, G12) genotypes while SIPC1 selected (G14, G17, G2). EV2, D2, ASTAB2, ASV and ASV1 measures found (G7, G6, G4) as desirable genotypes. Analytic measures based on all significant IPCA’s i.e. MASV and MASV1 settled for G6, G7, and G3. Adaptability measures GAI, HM, PRVG & MHPRVG observed G13, G4, and G12 genotypes would be of stable adaptations. Biplot analysis seen largest cluster comprised D3, D5, EV2, EV3, EV5, ASTAB3, ASTAB5, MASV1, MASV and Standard deviation measures. Genotypes were ranked G9, G11, and G6 by values of EV1, D1 & ASTAB1 for second year of study. D2, ASV, ASV1, EV2 & ASTAB2 observed (G9, G6, and G11) as adaptable genotypes. MASV & MASV1 measures also supported G9, G6, G11 genotypes for the considered locations of the zone. Studied measures were clustered in three groups in graphical analysis. Three clusters were observed among studied measures by biplot analysis. Measures EV1, EV2, EV3, D1, D2, D3, ASV, ASV1, MASV, MASV1, ASTAB1, and ASTAB2 & ASTAB3 formed largest cluster.


Introduction
Very well wide adaptation of wheat made it possible to cultivate an important cereal in most of the countries [18]. Multi location trials have been conducted to evaluate the yield performance of several genotypes simultaneously [7]. Field evaluation of genotypes have required an efficient analytic estimation procedure for GxE interactions [1]. Presence of cross over type GxE interactions mask the real potential of deserving genotypes for their specific and general adaptations [17]. Quite large number of analytic approaches have developed especially for adaptation behavior of genotypes [15,2]. Univariate parametric, nonparametric, multivariate models for additive and multiplicative nature of factors. Over exploited method Additive Main effects and Multiplicative Interaction (AMMI) advocated in agricultural research field experiments [8]. Large portion of the interactions sum of squares had been utilized by AMMI analysis to discriminate environments, adaptability of the genotypes to specific and general environmental conditions to harvest well yield [5]. This analysis mechanism has proved as an effective analysis with possible opportunities to research workers [6]. The current study was planned with clear objectives (i) Study number of AMMI based measures as per utilization of number of significant interaction principal components (ii) similarity & dissimilarity among adaptability measures.

Materials and Methods
Peninsular zone comprises mainly of Maharashtra and Karnataka states of our country. Major three species of wheat viz T. aestivum, T. durum, and T. dicoccumare cultivated in this zone. Bread wheat cultivation is concentrated under irrigated environments, whereas, the cultivation of durum and dicoccum wheat is generally confined to rainfed/ Peninsular Zone of the Country by AMMI Model restricted irrigation situation. Genotype by environment (GxE) interactions of seventeen advanced wheat genotypes at thirteen major locationsduring 2017-18 and eleven genotypes at elevenlocations in 2018-19 cropping season of the zone by AMMI model. Research field trials were conducted at centers of AICRP by randomized complete block designs with four replications. Recommended agronomic practices were followed to harvest good yield. Details of genotype parentage along with environmental conditions were reflected in tables 1 & 2 for ready reference. AMMI first calculate genotype and environment additive effect using analysis of variance (ANOVA) and then analyse residual from these model using principal components analysis (PCA). AMMI stability value (ASV) proposed by Purchase [11] to quantify the stability measure by considering relative weight of IPCA 1 and IPCA 2 scores. In certain cases where more than two IPCAs were significant, ASV failed to encompass all the variability explained by GxE interactions. Zali [20] attempted modified version ASV which would cover all available Interaction Principal Components. But in doing so, Zali interpreted the formula of ASV incorrectly compared to the original formula of Purchase [11,12]. In the present study the original MASV formula of Zali [20] and a revised version of MASV [2] were compared with other AMMI based measures of interaction effects. The description of widely used measures based on AMMI analysis was mentioned for completeness.    AMMI analysis was performed using AMMISOFT version 1.0, available at https://scs.cals.cornell.edu/people/ hughgauch/ and SAS software version 9.3. AMMI based measures were compared with recent analytic measures of adaptability calculated as the relative performance of genetic values (RPGV) and MHGV (Harmonic mean of Genetic Values), based on the harmonic mean of the genotypic values across different environments. Another harmonic mean based measure of the relative performance of the genotypic values (MHRPGV) for the simultaneous analysis of stability, adaptability and yield [14]. GV ij is the genotypic value of the i genotype, in the j environment, expressed as a proportion of the average in this environment. Geometric adaptability index (GAI) [10] was calculated as ,∏ X = > ? > @ ; in which X = 1 , X = 2, X = 3, … X = m are the mean yields of the first, second and mth genotype across environments and n is number of environments. Genotypes with higher values of GAI are desirable.

Results and Discussion
Better understanding of the GxE interaction had been provided by AMMI analysis as this facilitated identification of general and specific adaptations of genotypes anddiscriminate environments. In fact AMMI exercised family of models with retaining 0, 1, 2, or more significant interaction principal components (IPCs).

First Year of Study (2017-18)
Estimated sums of squares for GxE signal and GxE noise were 69.9% and 30.2% respectively Sum of squares for GxE signal is 2.33 times that for genotypes main effects, implied, narrow adaptations are important for trials research dataset. First IPC1 alone is 1.52 times the genotypes main effects whereas GxE noise is1.01 times the genotypes main effects. Discarding noise improves accuracy, increases repeatability, simplifies conclusions, and accelerates progress [6]. Highly significant environments (49.9), GxE interaction (21.4) and genotypes (6.4) were observed by ANOVA analysis. Diversity of considered locations had justified the selection of environments [3]. Explained variation of GxE interaction accounted by each of highly significant IPCA's, as type-1 measures benefited 45.5%, type-2 measures utilized 66.3%, type 3 measures used up to 75.9%, type-5 measures used up to 88.4% of interaction variations, tough IPCA5, IPCA6 and IPCA7 contributed to the tune of 5.5, 3.9 and 2.9% respectively ( Table 4). Use of AMMI derived measures upto first five IPCAs had been justified [9].     ASV and ASV1 observed suitability of (G7, G6, G4) along with unsuitable performance forG15, G11 (Table 6). Considering first two IPCAs in ASV & ASV1 measures utilized 66.3% of GxE interaction sum of squares. The two IPCAs have different values and meanings and the ASV and ASV1 parameters using the Pythagoras theorem and to get estimated values between IPCA1 and IPCA2 scores to produce a balanced measure between the two IPCA scores [11,12].

Second Year of Study (2018-19)
Estimated sums of squares for GxE signal and GxE noise were 66.6% and 33.4% respectively. Note that the SS for GxE signal is 1.26 times that for genotypes main effects.
Hence, narrow adaptations are important for this dataset. Even just IPC1 alone is 0.97 times the genotypes main effects. Also note that GxE noise is0.63 times the genotypes main effects. Highly significant environments (44.1), GxE interaction (20.4) and genotypes (10.8) were observed by ANOVA analysis. Extent of GxE interaction variation accounted by each of highly significant IPCA's for AMMI based measures, as type-1 benefited 51.2%, type-2 measures utilized 66.3%, type 3 measures used up to 79.3%, contributions of other non significant IPCA4, IPCA5, IPCA6 and IPCA7 were of 8.5, 5.8, 3.6 and 1.7% respectively ( Table  4) (Tables  7 and 8). ASV observed (G9, G6, G11) and ASV1 recommended (G9, G11, G6) as with stable performance and unsuitable performance forG1, G2 (Table 5) (Tables 7 & 8). Composite measures MASV & MASV1selected G9, G6, G11 genotypes for stable performance and G1, G2 would not be recommended for cultivation due to unstable yield behavior. Average yield was considered as an important measure to assess the genotypes potential as highly significant yield differences were exhibited. G4, G8, G6 genotypes maintained higher yields as compared to G1 & G11. GAI pointed towards G4, G7, G8 as of stable adaptation and G1 & G11 would be unstable. HM, PRVG and MHPRVG measures identified G4, G7, G8 and G1 & G11 for general and specific adaptations respectively. Lower values of standard error anticipated consistent yield performance forG4, G10, G1 genotypes as far as considered locations were considered. Three clusters were observed among studied measures by biplotanalysis by considering first two significant principal components (80.3% total variation) (

Conclusions
AMMI analysis has been proved as an effective tool to explore complex GxE interaction under multi environmental trials. Large number of AMMI based measures had been studied as each measures related to a different concept of stability. Recent analytic measures for adaptability of wheat genotypes exhibited affiliation withAMMI based measures exploiting number of significant IPC scores.

Conflict of Interest
All the authors do not have any possible conflicts of interest.