Hybrid Maize Selection Through GGE Biplot Analysis
https://doi.org/10.1590/1678-4499.20170438T. R. A. Oliveira et al.PLANT BREEDING - ArticleHybrid maize selection through GGE biplot analysisTâmara Rebecca Albuquerque de Oliveira1*, Hélio Wilson Lemos de Carvalho2,Gustavo Hugo Ferreira Oliveira3, Emiliano Fernandes Nassau Costa2, Geraldo de Amaral Gravina1,Rafael Dantas dos Santos4, José Luiz Sandes de Carvalho Filho51.Universidade Estadual do Norte Fluminense Darcy Ribeiro - Centro de Ciências e Tecnologias Agropecuárias Campos dos Goytacazes (RJ), Brazil.2.Embrapa Tabuleiros Costeiros - Aracaju (SE), Brazil.3.Universidade Federal de Sergipe - Núcleo de Graduação de Agronomia - Nossa Senhora da Glória (SE), Brazil.4.Embrapa Centro de Pesquisa Agropecuária do Trópico Semiárido - Petrolina (PE), Brazil.5.Universidade Federal Rural de Pernambuco - Departamento de Engenharia Agronômica - Recife (PE), Brazil.ABSTRACT: The cultivation of genotypes non-adapted to thewas efficient on data interpretation and represented 63.73% of thecultivation region of interest is among the main factors responsibletotal variation in the first two main components, it also allowedfor low yield. The aim of the present study is to select hybrid maizeclassifying the ten environments into three macro-environments.through GGE biplot analysis and to assess its adaptability andMost environments were positively correlated. Hybrids 2 B 604 HX,stability in different environments in Northeastern Brazil. Twenty-five30 A 95 HX, 2 B 587 HX and 2 B 710 HX were responsive and stable.hybrid maize cultivars were assessed in ten different environmentsHybrid 30 A 16 HX was recommended for macro-environmentsin Northeastern Brazil in 2012 and 2013 based on the randomized2 and 3. Cultivar 30 A 68 HX was recommended to environment 1.block design, with two replications. The analysis of variance andSão Raimundo das Mangabeiras and Nova Santa Rosa countiesassessment of genotype adaptability and stability were made throughwere discriminating and representative. Nossa Senhora das Dores,GGE biplot analysis, based on grain yield. Analysis of varianceUmbaúba, Teresina, Brejo, Frei Paulo, Colinas and Balsa countiesresults showed different performances depending on the genotype,were ambiguous and non-recommended for further evaluations.as well as genotype/environment interaction. The biplot analysisKey words: G E interaction, multivariate analysis, Zea mays L.*Corresponding author: tamara rebecca@hotmail.comReceived: Dec. 26, 2017 – Accepted: Sept. 18, 2018166Bragantia, Campinas, v. 78, n. 2, p.166-174, 2019
Multivariate analysis for maize selectionINTRODUCTIONMATERIALS AND METHODMaize (Zea mays L.) is an important culture worldwidedue to its several applications and economic relevance.Farmers in Northeastern Brazil count on low technologicallevel during maize cultivation because, in this region, itis mainly grown in small farms (Carpentieri-Pípolo et al.2010). The lack of genotypes adapted to the soil and weatherin the region has resulted in low and instable yield (Oliveiraet al. 2017).It is difficult to identify superior genotypes due tothe genotypes environments (G E) interaction, butassessing such interaction is extremely important, since itis the number one factor responsible for changing genotypeperformance in different environments. This feature impairsthe recommendation of adaptable and stable cultivars(Mohamed 2013; Oliveira et al 2017). G E studies allowidentifying the ideal location to each genotype, which wouldmaximize the grain yield potential and reduce productioncosts (Oyekunle et al. 2017).Maize producers use different analyses to assess cultivaradaptability and stability based on biometric concepts(Camargo-Buitrago et al. 2011; Silva and Benin 2012).The GGE biplot analysis derives from the first two maincomponents (MCs), the first one regards the yield ratio,which is associated with genotypic characteristics; andthe second one concerns the yield related to the G Einteraction (Yan and Holland 2010; Yan 2001).The GGE biplot analysis is efficient because it enablespredicting the mean genotype yield per specific environment,as well as helps identifying the most stable genotype for theregion of interest (Santos et al. 2017; Yan 2014). Accordingto Badu-Apraku et al (2012), this analysis is more versatileand flexible than other models based on simple linearregression and on segmented linear regression, as well asthan non-parametric methods, because it allows betterunderstanding the G E interaction.Studies about the adaptability and stability of differentcultures based on biplot graphics corroborate the efficiency ofthe analysis to recommend genotypes and to group favorableand unfavorable environments (Silva et al. 2011; Santoset al. 2016; Paramesh et al. 2016; Yokomizo et al. 2017).The aim of the present study was to select hybridmaize through the GGE biplot analysis and to evaluateits adaptability and stability in different environments inNortheastern Brazil.We assessed 25 hybrids from private and public companiesduring the agricultural years 2012 and 2013 (Table 1) inMaranhão (Balsas, Brejo, Colinas and São Raimundo dasMangabeiras counties), Piauí (Nova Santa Rosa, Teresinaand Uruçuí counties) and Sergipe states (Nossa Senhoradas Dores, Frei Paulo and Umbaúba counties) (Table 2).Our study followed the randomized block design, with tworepetitions. The plots comprise five lines (5.0 m long) each,and the pits in the rows were spaced 0.70 m and 0.20 m fromeach other. Fertilization procedures were based on results ofsoil analyses applied to the soil in each experimental area.Were sowed 15 seeds per linear meter (75 plants perline). The seedlings were thinned 15 days after emergence(five plants per linear meter remained after thinning) andassessed all lines in each plot at harvest time in order to findthe yield rates.The needs of the culture (in each region) guided the weedand pest control procedures, but we did not irrigate the plants.The analysis of variance was conducted to each location in2012 and 2013 in order to assess residual variance homogeneity.The multivariate analysis of variance included genotypes,year and local, finding the GE matrix. The analysis assessedeach location as an environment.Information about the phenotypic mean substantiatedthe multivariate GGE biplot analysis. We took the followingmodel into consideration (Eq. 1):Ȳij – µ Gi Ej GEij(1)where Ȳij is the phenotypic mean of genotype i in environment j;μ is the general constant; Gi is the random effect of genotype i;Ej is the fixed effect of environment j; and GEij is the randomeffect of the interaction between genotype i and environment j(Yan 2001).The GGE biplot model does not dissociate the genotypeeffect (G) from the genotypse environments effect (GE).It keeps G and GE together in two multiplicative terms inthe Eq. 2:Yij – µ – βj gi1e1j gi2ej2 εij(2)where Yij is the expected performance of genotype i inenvironment j; µ is the general constant of observations; βjis the main effect of environment j; g1i and e1j are the mainBragantia, Campinas, v. 78, n. 2, p.166-174, 2019167
T. R. A. Oliveira et al.scores of the ith genotype in the jth environment, respectively;and εij is the non-explained residue of both effects (“noise”).The biplot graphs in the GGE model were generatedthrough the simple dispersion of gi1 and gi2 to assess theTable 1. List of hybrid maize cultivars and their respective origins, types, cycles, colors, grain textures and companies.14No.CultivarTransgenic/ conventionalType1Cycle2Grain color3Grain texture4Seed company120 A 55 HXTransgenicTHEORSMHARDMORGAN220 A 78 HXTransgenicSHEORSMHARDDOW32 B 433 HXTransgenicTHEEY / ORSMDENTDOW42 B 587 HXTransgenicSHEY / ORSMDENTDOW52 B 604 HXTransgenicSHmEORSMHARDDOW62 B 688 HXTransgenicTHEORSMHARDDOW72 B 707 HXTransgenicSHEORSMHARDDOW82 B 710 HXTransgenicSHEY / ORSMHARDDOW930 A 16 HXTransgenicSHEORSMHARDMORGAN1030 A 37 HXTransgenicSHEEY / ORSMHARDMORGAN1130 A 68 HXTransgenicSHEEORSMHARDMORGAN1230 A 91 HXTransgenicSHmEY / ORSMHARDMORGAN1330 A 95 HXTransgenicTHEORSMHARDMORGAN1430 F 53 HRTransgenicSHEORSMHARDDU PONT1530 K 73 HTransgenicSHEY / ORSMHARDDU PONT16AG 8041 YGTransgenicSHEY / ORSMHARDSEMENTES17AS 1555 YGTransgenicSHEORSMHARDAGROESTE18AS 1596 R2TransgenicSHERSMDENTAGROESTE19BM 820ConventionalSHERHARDBIOMATRIX20BRS 2020ConventionalDHEORSMHARDEMBRAPA21BRS 2022ConventionalDHEORSMDENTEMBRAPA22DKB 330 YGConventionalSHEER / ORSMDENTDEKALB23DKB 370ConventionalSHmEY / ORSMHARDDEKALB24P 4285 HTransgenicSHEY/ ORHARDDU PONT25STATUSVIPTransgenicSHEHARDSYNGENTAORDH Double hybrid; TH Triple hybrid; SHm Modified single hybrid; EE Extra early; E Early;SMDENT Semi-dent; SMHARD Semi-hard.23OR Orange; R Reddish; Y Yellow;Table 2. Geographic coordinates of counties where the experiments were installed in Northeast Brazil, 2012 and 2013.CountyLatitude (S)Longitude (W)Altitude (m)Soil typeMean temperature ( C)Colinas/MA06 0144 4141Argisol DR27São Raimundo dasMangabeiras/MA07 22’45 36’225Argisol Y26Brejo/MA03 4142 4555Latosol Y27Balsas/MA07 32’46 02’247Argisol Y29Uruçuí/PI03 1141 3770Argisol Y25Teresina/PI05 05’42 49’72Argisol Y28Nova Santa Rosa/PI08 2445 55469Latosol Y23Frei Paulo/SE10 55’37 53’272Cambisol26Nossa Sra das Dores/SE10 3037 13200Latosol Y25Umbaúba/SE12 22’37 40’109Argisol Y24DR dark red; Y yellow.168Bragantia, Campinas, v. 78, n. 2, p.166-174, 2019
Multivariate analysis for maize selectiongenotypes; and of ej1 and ej2 to evaluate the environmentsbased on the Singular Value Decomposition (SVD), in Eq. 3:Yij – μ – βj λ1ξi1η1j λ2ξi2η2j εij(3)where λ1 and λ2 are the highest self-values of the first andsecond main components: ACP1 and ACP2, respectively; ξi1and ξi2 are the self-vectors of the ith genotype of ACP1 andACP2, respectively; and η1j and η2j are the self-vectors ofthe jth environment of ACP1 and ACP2, respectively (Yan2001). The R (R Development Core Team 2014) softwareand the GGEbiplotGUI package (Wickham 2009) were ourtools to conduct the GGE biplot analysis.RESULTS AND DISCUSSIONBased on the coefficient of variation (CV%), the recordedmeasurements evidenced good experimental precision,because they stayed inside the acceptable limits set for maizecultures (Fristche-Neto et al. 2012) (Table 3).The significant environment year result (p 0.01) showedweather and soil differences both in 2012 and in 2013. On theother hand, based on the significant cultivar environment,and cultivar year effect (p 0.01) interaction, cropspresented different behavior in the assessed environmentsduring the evaluated years. According to this result, genotypesrecorded different performances due to environmentalchanges. Therefore, maize cultivar classification can changedepending on the assessed environment, corroborating theresults recorded by Faria et al. (2017) and Oliveira et al. (2017).Table 3. Joint analysis of the mean grain production of 25 hybrids maizecultivars tested in 10 locations in Northeastern Brazil, 2012 and 2013.Variation sourcesDFMean squaresRepetition (Environment 112119936**Years11175065583**Genotype Environments2162667319**Genotype Years247132006**Environments Years921294325**Genotype Environments Years2161606177**Error480694624CV(%)10.08Mean (kg ha–1)8623.77** indicates 1% significance.The genotype classification instability caused byenvironmental variations highlighted the need of futureand detailed studies about the behavior of this cultivarduring the selection of the best genotypes.The first two main components (MCs) of the biplotanalysis applied to the genotypes environments GGE biplotanalysis explained 60.01% of the total variation (Fig. 1).This result suggests that biplot graphics explain most sumsof squares and GE interaction in the genotype. This outcomemade it possible to have a safe genotype selection based onthe multivariate analysis (Yan 2001).A set of perpendicular lines divided the which-wonwhere biplot into many groups. Genotypes in the vertexof the biplot were farther from the origin than all othergenotypes inside the sector limited by them. Therefore,these genotypes were classified as the ones showing the bestperformance in one or more environments (Yihunie andGesesse 2018). These genotypes could be used to identifypossible macro-environments (Santos et al. 2017; Yan 2001);the ones located inside the polygon were less responsive toenvironment stimuli (Fig. 1).Environments grouped inside the same polygon hadsimilar influence on the genotypes. Environment groupsderiving from the 10 assessed environments highlightedthree macro-environments; the first one encompassed SãoRaimundo das Mangabeiras, Nossa Senhora das Dores, FreiPaulo, Colinas and Balsas (presented genotype 11 in thevertex), and recorded the highest mean grain yield. Thisoutcome corroborated the results recorded by Cardoso et al.(2014). Hybrid 9 stayed in the vertex of macro-environments2 (Nova Santa Rosa, Umbaúba, Teresina and Brejo) and 3(Uruçuí) and reached the best yield rate in environmentsinside these macro-environments – similar results werefound in previous studies (Carvalho et al. 2013; Carvalhoet al. 2017).Genotypes from the polygon vertex did not group inany environment and were not favorable for the testedenvironment groups – which recorded low yield. Genotypes12, 14, 18 and 25 were non-responsive because, based on theenvironments, they were outside the groups.The assessed genotype yield and stability based on themean environment coordination (MEC), was representedby the circle around the image in Fig. 2 (means stabilities).The line cutting the origin and passing cross the idealenvironment was the axis of the ideal environment. We usedthe main-component scores of all environments to defineBragantia, Campinas, v. 78, n. 2, p.166-174, 2019169
T. R. A. Oliveira et al.4Uruçuí3PC2 16.15%210-1-225BrejoTeresina14169UmbaúbaNova Santa Rosa621915513 14 810324 2322São R. Mangabeiras11 7Nossa S. das DoresFrei PauloColinas-32117201218Balsas-2-40PC1 43.86%2Figure 1. GGE biplot representing the which-won-where graph indicating the yield rankings of 25 hybrids.4Uruçuí3PC2 16.15%210-1-225BrejoTeresina14169UmbaúbaNova Santa Rosa621915513 1481017243 2322São R. Mangabeiras11 7Nossa S. das Dores18Frei PauloColinas-3212012Balsas-4-20PC1 43.86%2Figure 2. GGE biplot representing the means stabilities indicating the yield rankings of 25 hybrids, and their respective production stabilities.the ideal environment (the arrow highlighted the highestgenotypic value).The axis of the coordinate, perpendicular to the abscissas,indicated the genotypes mostly affected by the genotypes environments interaction, as well as their lower stability.It also separated the genotypes presenting records belowand above the average (Yan and Tinker 2006). Therefore,genotypes 11, 9, 5, 13, 4, 7, 8, 24, 6, 10, 2, 3, 14, 1, 16 and23 recorded yield rates higher than the general mean. Theyield of Hybrid 22 was close to the general mean; the othergenotypes recorded lower yield. These results were confirmedby the mean test (Table 4).The higher the genotype projection on the axis, the betterthe genotype and environment interaction and, consequently,170the more stable the genotype. Genotypes 25 and 18 were themost instable ones.Based on the GGE biplot analysis, the ideal genotypeis the one presenting the longest vector and no G Einteraction (arrow in the center of the smallest circle – Fig. 3).Although this genotype was just symbolic, it was the referenceto assess other genotypes. Plant breeding programs searchfor genotypes close to the ideal. The other concentriccircles in the figure helped visualizing the distance betweengenotypes. Yet, ideal hybrids were the ones presenting highPC1 (high yield) and low PC2 (high stability) values in thebiplot generated to estimate the best genotype.Based on the genotypes for grain yield ranking,cultivars 5, 13, 4 and 8, were the closest ones to the idealBragantia, Campinas, v. 78, n. 2, p.166-174, 2019
Multivariate analysis for maize selectiongenotype, because they recorded both high yield meansand phenotypic stability. Genotypes 12 and 25 were thelesser productive and most unstable ones.Discriminating the tested environments is an importantmeasure, since environments unable to discriminatethemselves do not provide information about genotypes;therefore, they are useless. It is also essential measuringthe ability of a tested environment to represent the targetenvironment. When the environment is not representative,it is not just useless but also biased, because it can provideinaccurate information about the assessed genotypes. Thegraphic Discrimination Representativeness showedthe environment ability to discriminate itself and itsrepresentativeness (Fig. 4).Concentric calculations in the biplot helped visualizingenvironment vector length – which is proportional to thestandard deviation inside the respective environment – andenvironment ability to discriminate itself. Thus, the testenvironments showing long vectors were the ones betterseparating the genotypes, whereas the shorter vectorsprovided little, or none, information about differencesbetween genotypes (Yan et al. 2007). The mean environment(represented by the small circle at the end of the arrow onthe line) held the mean coordinates of all test environments(Fig. 4). The axis of the mean environment (AME) is theline crossing the mean environment and the origin ofthe biplot. The test environments presenting shorter angleswith AME were the most representative ones (Yan andTinker 2006).Table 4. Means yield values of 25 maize hybrids.GenótypesYield*30A68HX9461.9 a30A16HX9318.2 ab2B707HX9257.1 abc30A95HX9229.1 abc2B604HX9182.0 abc2B710HX9096.9 abcd2B587HX9087.3 abcdP4285H9020.0 abcd30A37HX8979.2 abcd2B433HX8862.7 abcde30F53HR8816.1 abcde20A55HX8810.0 abcde2B688HX8745.6 bcdefAG8041YG8736.8 bcdefDKB3708694.5 bcdef20A78HX8677.8 bcdefDKB330YG8637.8 bcdef30K73H8612.2 cdefAS1596R28431.0 defBM8208261.8 efgAS1555YG8081.3 fghSTATUSVI7610.6 ghiBRS20227488.4 hiBRS20207384.2 i30A91HX7352.1 i* Means with different letters in the column differ by 5% probability by theTukey test.4Uruçuí3PC2 16.15%210-1-225BrejoTeresina14169UmbaúbaNova Santa Rosa2615191513481017243 2322São R. Mangabeiras11 7Nossa S. das Dores18Frei PauloColinas-3212012Balsas-4-202PC1 43.86%Figure 3. GGE biplot comparing 25 hybrids evaluated according to the estimate of an ideal genotype.Bragantia, Campinas, v. 78, n. 2, p.166-174, 2019171
T. R. A. Oliveira et al.4Uruçuí3PC2 16.15%210-1-225BrejoTeresina14169UmbaúbaNova Santa Rosa6219151513481017243 2322São R. Mangabeiras117Nossa S. das Dores18Frei PauloColinas-3212012Balsas-4-20PC1 43.86%2Figure 4. GGE biplot comparing 25 hybrids evaluated according to the discrimination and representativeness of environments for grainyield (kg·ha–1).Uruçui and Balsas was highly discriminating and nonrepresentative. Such condition could have been usefulfor genotype selection, mainly to select the adaptablegenotypes. Target environments could be divided into macroenvironments or help excluding unstable genotypes, whenthe environment of interest was a single macro-environment.Genotype 7 had good adaptability to environments inmacro-environment 1 (Balsas County).São Raimundo das Mangabeiras and Nova SantaRosa counties recorded the best representativeness andgenotypes with good discriminating ability; this outcomefavored the selection of widely adaptable genotypes.Colinas County was non-representative and had the lessdiscriminating genotype; therefore, it was excluded (Yanand Tinker 2006; Yan et al. 2007).Lines connecting the origin of the biplot to environmentmarkers are the environment vectors. The angle betweenvectors of the two environments is related to the coefficientof correlation of the two angles. The cosine betweenvectors of the two environments indicates their coefficientof genetic correlation. Sharp, obtuse and straight ang
The GGE biplot analysis derives from the first two main components (MCs), the first one regards the yield ratio, which is associated with genotypic characteristics; and the second one concerns the yield related to the G E interaction (Yan and Holland 2010; Yan 2001). The GGE biplot analysis is efficient because it enables
The analysis was conducted and biplots generated using the “GGEbiplot” software developed by Yan (2 001). The model for GGE Biplot A GGE biplot isconstructed by subjecting the GGE matrix i.e., the environment-centred data, to singular value decomposition (SVD) as devised by Eckart and Young (1936). The GGE matrix is
Analysis Software‐GGE biplot GGE ‐biplot method is a new analysis way of multi‐factor interactions. GGE‐biplot can clearly identify high‐yield, stable yield varieties and the test sites with effective resolution using chart, so GGE‐biplot is considered to be the ideal method of analysis of regional test data.
Maize Farming Techniques Manual PART II: 5. Basic 5.1 V (C) Tasseling stage TECHNICAL Know ledge for Maize C arious stages of Maize plant (A) Seedling stage It is the initial stage of maize plant with 2-4 leaves after the emergence. It takes 1 plantation (B) Growth stage of maize
MAIZE MILLING BUSINESS PLAN - SAMPLE iMvubu Holdings, Ltd Siavonga District Zambia. 1. Executive Summary 1. Executive Summary a) Business Concept iMvubu Holdings, Ltd is proposing a maize mill to service the remote Siavonga district of Zambia. iMvubu will buy maize and produce maize meal with bran as a by-product.
of the crop production. The research focus on identification Maize Lethal Necrosis Disease, Gray Leaf Spot Disease and Maize streak virus for maize diseases using a mobile vision approach. In this research, the total of 3 groups of maize diseases with statistical, texture and color features are considered. The neural network
Directorate Marketing Tel: 012 319 8456 Private Bag X 15 Fax: 012 319 8131 . Yellow maize is mostly used for animal feed production while white maize is primarily produced for human consumption. Maize is the second largest produced crop in South Africa after sugar cane. The maize industry is important to the economy both
Yan (2001) developed GGE biplot to visually analyze these two types of MET data. They reported that GGE biplot was originally designed for graphical analysis of MET data and it was later extended to graphical analysis of other types of two-way data, Table 2. such as genotype by trait data and other types of two-way data.
The Adventure Tourism Development Index (ATDI) is a joint initiative of The George Washington University and The Adventure Travel Trade Association (ATTA). The ATDI offers a ranking of countries around the world based on principles of sustainable adventure tourism
