免费文献传递   相关文献

Research Progress on Plant Genomic Selection(GS) and Its Application in Maize Breeding

植物全基因组选择技术的研究进展及其在玉米育种上的应用



全 文 :书!"#$%&
,2016,36(6):1269-1277
犃犮狋犪犅狅狋.犅狅狉犲犪犾.犗犮犮犻犱犲狀狋.犛犻狀.
  !"#$:10004025(2016)06126909               犱狅犻:10.7606/j.issn.10004025.2016.06.1269
%&(
:20151012;)*&%+(:20160428
,-./

()*+%,-
(31401457);./01234546789(SDAIT0204);:.%;<4=>?@
0123

A B(1978-),C,DE,FGHI,JKLMNOPQRSTU;:VW/,GHI,JKLMNOPQR456,789:;<=>?@A
BCDEFGHI=JK
! "1,#$%1,&(2,) *1,#$+1,,-.1,/011
(1./034+%XNOGHY,Z[250100;2\]^<_`a,./\]265200)
L M:b,cdef67Sgb,cdhijklmnopqrst(SNP)uvwx8kyz{|}~ BLUP
z€=‚ƒstkR<„

…†€=R<„
(GEBV),*‡ˆ‰Š‹Œk_st€=‡2Ž8R<„
‘ef

’“”•–—˜™š›œ,cdefžŸkJKc 
———
v¡x8k¢z£i¤

zk}~¥
¦

stk¢z§¨{©

rªPQ«

¬§­,cdefžŸkœ®¥¯kGH°‘±²

³´µb,
cdef67¶NOR<·¸Š¹º¬§­»—k°¼

NOP

b,cdef

NO

€=R<„
QRST$
:Q789 !UVWX:A
犚犲狊犲犪狉犮犺犘狉狅犵狉犲狊狊狅狀犘犾犪狀狋犌犲狀狅犿犻犮犛犲犾犲犮狋犻狅狀(犌犛)
犪狀犱犐狋狊犃狆狆犾犻犮犪狋犻狅狀犻狀犕犪犻狕犲犅狉犲犲犱犻狀犵
SUNQi1,LIWenlan1,CHENLitao2,ZHAOMeng1,LIWencai1,
YUYanli1,MENGZhaodong1
(1MaizeInstitute,ShandongAcademyofAgriculturalSciences,Jinan250100,China;2LaiyangCitySeedCorporation,Laiy
ang,Shandong265200,China)
犃犫狊狋狉犪犮狋:Markerassistedselection(MAS)technologycouldrealizedirectgeneticselection,butitmust
baseonQTLmapping.Genomicselection(GS),asthenewestMASmethod,hasmuchadvantagecom
paredtotraditionalMAStechnology,especialyQTLmappingnotnecessary.Inthispaper,thefactorsaf
fectingpredictionaccuracyofGSwerereviewed,includingtrainingpopulationtype,predictionmodel,
markernumber,populationsize,populationstructure,hereditaryoftraitsandsoon.Theapplicationof
GSinmaizebreedingwasalsointroducedaswelashybridsperformanceprediction.Wethenpredicated
thefutureresearchandapplicationofGSinmaizebreeding.
犓犲狔狑狅狉犱狊:genomicselection(GS);maize;GEBV
  Withrapiddevelopmentofthemolecularbiol
ogy and genomics, markerassisted selection
(MAS)emergedasthetimesrequire.MAStech
nologyisasakindofcropgeneticimprovement
methodcombingthephenotypicandgeneticvalue,
whichcanrealizegeneticdirectselectionandeffec
tivepolymerization[1].Whencomplextraitscon
troledby multiplegenesneedtobeimproved,
MAShastwoaspectsofflaws.First,selectionof
theprogenypopulationisestablishedonthequan
titytraitslocation(QTL)mapping.Buttheresult
ofQTLmappingbasingonthebiparentalpopula
tionshasnouniversalityandcouldn’tbeapplied
accuratelyinbreeding[2].Second,theimportant
traitswerecontroledbylotsofsmal effective
genes,lack ofappropriatestatistic methodand
breedingtechnology which wil apply quantity
genestocomplextraitsimprovement[3].NewMAS
technologygenomicselection(GS)emergedasthe
timesrequire.
1 Originationandadvantageofge
nomicselection(GS)
  Meuwissenfirstputforwardgenomicselection
(GS)breedingstrategy.GSusesa“trainingpopu
lation”ofindividualsthathavebeengenotypedand
phenotyped. Best linear unbiased prediction
(BLUP)modelisestablishedonthebasisofthe
genotypedresultofanindividualanditsbreeding
value (Meanperformanceofcrosseswithsame
tester)forthetrainingpopulation.Thebreeding
valueof“Candidatepopulation”isestimatedby
BLUPmodelandgenotypicdata.withoutcrossto
testerand phenotypesrecord[4].BLUP model
takesgenotypicdataofuntestedindividualsand
producesgenomicestimatedbreedingvalues(GEB
Vs).TheseGEBVssaynothingofthefunctionof
theunderlyinggenesastheidealselectioncriteri
on[5].GenomicselectionbasisofGEBVsissuperi
ortotraditionalbreedingforincreasinggainsper
unittimeevenifbothmodelsshowthesameeffi
ciency.Inprinciple,phenotypesvalueofthecan
didateindividualsisnonessentialfortheselection,
henceshorteningthelengthofthebreedingcy
cle[6].
Genomicselectionhaveseveralmeritscom
paredtothetraditionalMAS.(1)QTLmappingis
notnecessaryforGS.Genomicselectiondiffers
frompreviousstrategiessuchaslinkageandassoci
ationmappinginthatitabandonstheobjectiveto
maptheeffectofsinglegeneandinsteadoffocu
singontheefficientestimationofbreedingvalues
onthebasisofalargenumberofmolecularmark
ers,idealycoveringthefulgenome[5].(2)Ge
nomicselectionismoreprecisionespecialyforear
lyselection.Genotypinguseshighdensitymolecu
larmarkerswhichcanestimateal oftheQTL
effectsandexplainthegeneticvarianceformostof
thetraits.ButMASonlyusesseveralmarkersin
traitsselection.Sogenomicselectionismoreaccu
ratethan MAS[7].(3)Genomicselectioncan
shortengenerationinterval,accelerategeneticpro
gressandreduceproductioncost.Geneticprogress
ofGSismorethanphenotypicselection4% -
25%.CostofGSislessthantraditionalbreeding
26%-56%[8].(4)Selectionefficiencyoflowheri
tabilitytraitsishigherforGSthanMAS.(5)The
criterionofGSisbreedingvalue,sumofalofthe
alelegeneticeffectsforeachindividual.Itis
judgedbythemeanperformanceofitscrossproge
ny,nottheperformanceofitself.SoGSismore
accurate[9].
Genomic selection originated from animal
breedingduringlastcentury.Ithasbeenwidely
usedindairycattlebreedinginAmerica,Austral
ia,NewZealandandsoon[1011].Itwasalsoap
pliedinbroilerchickensandpigsbreeding[1213].
GS’applicationinplantbreedingwasdevelopedin
recentyears,whichfocusedonsimulationstudies.
Itisusedinmaize[14],wheat[15],tree[16],sugar
beet[17],Barley[18],triticale[19]andsoon.
Empiricalstudyisperformedinlargercompa
niessuchasMonsantoandPioneerDupond.Mark
SorrelsandJeanLucJanninkaretryingtouseGS
toincreasethespeedofvarietyimprovement3-4
times.TheworkiscarriedoutwithCYMMITand
performedfouraspectstoimprovetheyieldof
maizeandwheat[20].
Undertheabovecontext,theobjectiveofthis
studyistoreviewtheessentialfactorsaffectingthe
GSinplantbreeding.Maizeisessentialforglobal
foodsecurity.Moreresearchofgenomicselection
onmaizelauchedinrecentyears[2123].Thepaper
wilintroducetheadvanceontheapplicationofGS
0721 ! " # $ % &                   36½
inmaizebreeding.Wethanputforwardthefuture
research whichshouldbecarriedoutin maize
breedinginChina.
2 Affectingfactorsofgenomicselec
tion
  FactorsthataffectGSpredictionaccuracyof
includethenumberofmarkersusedforestimating
the GEBVs[10],trait heritability[7],calibration
populationsize[5],statisticalmodels[24],number
andtypeofmolecularmarkers[2526],linkagedise
quilibrium[27],effectivepopulationsize[28],rela
tionshipbetweencalibrationandtestset(TS)[2931]
andpopulationstructure[3234].
2.1 犜狉犪犻狀犻狀犵狆狅狆狌犾犪狋犻狅狀狅犳犵犲狀狅犿犻犮狊犲犾犲犮狋犻狅狀
Inanimalbreeding,weonlydiscussedGSin
thecontextofpopulationwidelinkagedisequilibri
um,wherethepopulationmightbedefinedasan
entirebreedofcattle,pig,orchicken.Theneed
forhighmarkerdensitiesinGSmaybereducedif
thecandidatepopulationconsistsofprogenyofthe
trainingpopulation.Inthatcase,anevenlyspaced
lowdensitysubsetofthemarkerstypedonthe
trainingpopulationcanbeusedonthecandidates,
andscoresforthefulcomplementofmarkerscan
beinferredbycosegregation[35].Becauseplantsof
tenproduceverylargefulsibships(anF2popula
tionderivedfromasingleF1byselfingisanexam
pleofsuchasibship),however,thereisalsoa
traditionofQTLdetection,MASandGSwithin
suchsibships[5].BernardocomparedF2,BC1,and
BC2populationsfromanadapted×exoticmaize
crossastrainingpopulationinthesimulationex
periment[14].Theresultindicatesthatgenomewide
selectionshouldstartatF2ratherthanbackcross
population,evenwhenthenumberoffavorableal
lelesissubstantialylargerintheadaptedparent
thanintheexoticparent.Comparedtonatural
populations,geneticbasisofF2populationsissim
plerbecauseF2populationsderivefromonlytwo
inbredlines.Sothebiparentalpopulationsize
mightbesmalerthanthatofnaturalpopulations.
Simulationstudieshavepreviouslyindicatedthat
forthreecyclesofgenomewideselectioninana
dapted×exoticcross,apopulationsizeof犖犆0=
144wasgeneralysufficient[21].Lowdensitymark
ersaresuitabletoF2populations[22].Buttwodis
advantagesofF2populationsexist.Biparentalpop
ulationrequiresseparatemodelfortrainingwithin
eachcross.TheBLUPmodelisonlysuitforthe
progeniesselectionfromthetwoparentallines.
TheprogenyofF2populationmustbeselectedby
thephenotypicvalueofF3testcrosses.Folowing
progenyselectionmaybeonlyaccordingtoBLUP
modelafterF3.
F2astrainingpopulationoftenbesuiltfor
crosspolinated plantsuch as maize.Yusheng
Zhaobasedonexperimentaldataofsixsegregating
populationsfromahalfdialelmatingdesignwith
788testcrossprogeniesfromanelitemaizebreed
ingprogram[23].InthestudyofVannesa犲狋犪犾.[36],
markereffectsestimatedin255diversemaizehy
bridswereusedtopredictgrainyield,anthesis
date,andanthesissilkingintervalwithinthediver
sitypanelandtestcrossprogeniesof30F2derived
linesfromeachoffivepopulations.
Wegenast犲狋犪犾.suggestedthatgenomicselec
tionwasappliedinplantbreeding,however,not
onlywithinaspecificbiparentalcrossorwithina
diversepanelofelitelinesbutalsoratherwithin
andamongcrosses[37].Selfpolinationplantoften
adoptnaturalpopulationsuchaswheatorsugar.
Würschumetalused924sugarbeetlinesastrain
ingpopulation.Theresultssuggestthatatraining
population derivedfrom intensively phenotyped
andgenotypeddiverselinesfromabreedingpro
gramdoesholdpotentialtobuilduprobustcalibra
tionmodelsforgenomicselection[17].Hans犲狋犪犾.
accessedtheaccuracyofGEBVsforrustresistance
in206hexaploidwheatlandraces[15].
2.2 犘狉犲犱犻犮狋犻狅狀犿狅犱犲犾狅犳犵犲狀狅犿犻犮狊犲犾犲犮狋犻狅狀
Genomicselectionmodelingtakesadvantageof
theincreasingabundanceof molecular markers
throughmodelingofmanygeneticlociwithsmal
effects[26,35,38].Overthelastdecade,simulation
andempiricalcrossvalidationstudiesin plants
haveshownGSismoreeffectivethantraditional
MASstrategiesthatuseonlyasubsetofmarkers
17216¾        A B,®:#$b,cdef67kGH°§¨¶NOR<·k¸Š(¿“)
withsignificanteffects[57,39].
Estimationmethodsofaleliceffectsinclude
leastsquaresregression[40],ridgeregressionBLUP
(RRBLUP),principlecomponentanalysis[4142]
and Bayesregression[43].In essenceforleast
squares,chromosomefragmentsormarkersarese
lectedassociatedtothetraitsbygenomewideas
sociationstudies(GWAS)atthesametimeand
thentheeffectofthefragmentsisestimated[44].
RRBLUPmethodregardsthefragmenteffectsas
randomeffects.Themarkereffectwasestimated
bylinearmixed models.Thesum offragments
effectisbreedingvalueforanindividual[43].Bayes
methodscombinesthepriordistributionofmarker
effectvarianceanddatacolection.Frenquently
usedBayesmethodsconcludeBayesAandBayes
B.MaindifferencebetweenBayesAandBayesBis
thatBayesApermitsdifferentvariancefordiffer
entmarkersandBayesBpermitsthatthevariance
ofsomemarkersiszero[45].
Simulationstudiesshowthattheprediction
accuracyofBayesmethodisbestandleastsquares
isweakest.TheaccuracyrateofRRBLUPis
slightlysmalerthanBayesA.Evenso,RRBLUP
hasfouraspectssuperiorto Bayesian method.
First,Bayesianmethodiscomplexandneedsuper
computer.Butcomputerrequirementislowerand
calculationspeedishigherforRRBLUP.Marker
effectsareestimatedbyRRBLUPinSASPROC
IML[46].Second,prediction withinfamilieswas
moreaccurateinBLUPthanBayesB.Regression
coefficientbofRRBLUPisnearerto1thanBayes
A[47].Habier犲狋犪犾.showedthatRRBLUPismore
effectiveatcapturinggeneticrelationshipsbecause
itfitsmoremarkersintothepredictionModel[27].
Incontrast,BayesBismoreeffectiveatcapturing
LDbetweenmarkersandQTL.Third,RRBLUP
ismoreaccuratethanothermethodwhenthenum
berofQTLsincreasesortheheredityishigher[18].
Fourth,BLUPledtolowerinbreedingandasmal
lerreductionofgeneticvariancecomparedtoBayes
andPLS[48].From above,wecanconcludthat
BLUPmethodsisbetterthanBayesianregression
forplantmodels.
Inaddition,machinelearning methodsalso
canbeusedtopredictthemarkereffect,including
supportvectormachine(SVM),bootingandran
domforest(RF).Ogutu犲狋犪犾.comparedthese
methodsforgenomicselection.Theresultshows
thatthecorrelationbetweenthepredictedandtrue
breedingvaluesis0.547forboosting,0.497for
SVMs,and0.483forRF,indicatingbetterper
formanceforboostingthanforSVMsandRF[49].
2.3 犗狋犺犲狉犳犪犮狋狅狉狊犪犳犳犲犮狋犻狀犵狆狉犲犱犻犮狋犻狅狀犪犮犮狌狉犪犮狔
Ingenomewideselectionmethods,prediction
accuracyisaffectedbypopulationsize(犖),aver
agehereditaryoftraits(犺2)andmarkernumbers
(犖犕)[50].Simulationstudiesshowedthatthepop
ulationstructureisalsocrucialfortheprediction
accuracyingenomicselection[27].
Predictionaccuracyincreases with markers
density.Markersnumberonacertainlengthge
nomealsodirectlyaffectstotalinformationofge
neticmarkers.IfSSR markersdensityincreases
from0.25犖犲/犿狅狉犵犪狀 (犖犲,effectivepopulation
size)to2犖犲/犿狅狉犵犪狀,predictionaccuracywilbe
improvedfrom0.63to0.83.IfSNPmarkersden
sityincreasesfrom1犖犲/犿狅狉犵犪狀to8犖犲/犿狅狉犵犪狀,
predictionaccuracywilbeimprovedfrom0.69to
0.86.Evenatthehighesttesteddensitiesof2犖犲
SSRmarkersperMorganor8犖犲SNPmarkersper
Morgan,accuracyhadnotreachedaplateau[5].
Meanwhile,moremarkersnumber,moreeasyto
getthe Linkagedisequilibrium (LD)markers.
Emilyfoundthatinthebiparentalpopulations,
therewasnoconsistentgainingenomewidepre
diction(狉犿狆)fromincreasingmarkerdensitya
boveonemarkerper12.5cM[22].Zhao犲狋犪犾.re
vealedthattheaccuracywasnearlyreachingaplat
eauat800SNPswhenthenumberofmarkersvar
iedfrom100to800[23].Thereasonisthatgenome
issufficientlysaturated with markerswhenthe
predictionaccuracyarrivesataplateau[28,50].The
numberofmarkersneededforaccuratepredictions
ofgenotypicvaluesdependsontheextentoflink
agedisequilibrium (LD)between markersand
QTL[4]andalsoonthegermplasmunderconsidera
tion[18].
2721 ! " # $ % &                   36½
Differentmarkertypehasdifferentpolymor
phisminformationcontent(PIC).ComparingSSR
andSNPmarkers,theyfoundthatforsimilarac
curacies,theSNPmarkersrequiredadensityof2
to3timesthatoftheSSR[5].
Simulationstudiesshowedthatthepopulation
sizeiscrucialforthepredictionaccuracyingenom
icselection[27].TheresultofEmily犲狋犪犾.indicated
thatpredictionaccuracy狉犿狆increasedaspopula
tionsizeNincreased.Inthebiparentalmaizepop
ulationandwiththehighestmarkersnumber犖犕,
(1213markers)andhereditary犺2= 0.30,the
predictionaccuracyforgrainyieldwas狉犿狆=0.19
with犖=48,狉犿狆=0.26with犖 =96,and狉犿狆
=0.33with犖 =192[22].ZhaoYushengobserved
amonotonicincreaseinthepredictionaccuracyfor
grainyieldwithincreasingpopulationsizewithout
anysubstantialdecreaseintheslope[23].Thestudy
ofBernardoalsoindicatedthatlagerpoluationsize
wouldgethigherpredictionprecision[14].ButF2
populationsizeof犖犆0=144wasgeneralysuffi
cient[21].
Trainingpopulationstructureisalsoanim
portantfactoraffectingpredictionaccuracyofge
nomicselectionfor multiparentalpopulations.
Trainingpopulationstructuresetmethodsconclude
randomsampling,unidirectionalsampling(selec
tingindividualswithhighestgenotypicvalues),bi
directionalsampling (selectingindividuals with
highestorlowestgenotypicvalues)[5051].Thisbi
directionalselectionshowedtobemuchmorepow
erfulthanrandomsampling[52].YushengZhaoob
servedasubstantiallossintheaccuracytopredict
genomicbreedingvaluesinunidirectionalselected
populations.Bidirectionalselectionisavaluable
approachtoefficientlyimplementgenomicselection
inappliedplantbreedingprograms[53].
Forthesametraitwithinthesamepopulation,
predictionaccuracy(狉犿狆)wilremainunchanged
fordifferentcombinationsofpopulationsize(犖)
andtraithereditary(犺2).Decreaseonh2canbe
compensatedbyaproportionalincreasein犖 (and
viceversa)sothat狉犿狆ismaintained.Ontheoth
erhand,traitswithinitialylowh2canbeevalua
tedwithlarger犖orthe犺2forasubsetoftraits
canbeincreasedbytheuseofadditionaltestingre
sources.Differenttraits,however,varyintheir
predictionaccuracyeven when 犖,犺2,and 犖犕
(markersnumber)areconstant.Yieldtraitshad
lowerpredictionaccuracythanothertraitsdespite
theconstant犖,犺2,and犖犕.Simulationresults
indicatedthat狉犿狆isalsolowestforyieldtraitse
venwhenits犺2isashighasothertraits.Plant
heightandlodgingarealwayspredictedmostaccu
ratelyfolowedbyfloweringtime[22].Empiricalev
idenceandexperienceonthepredictabilityofdif
ferenttraitsarenecessaryindesigningtraining
populations.
3 Genomicselectioninmaizebreed
ing
3.1 犗狉犻犵犻狀犪狋犻狅狀狅犳犌犛犻狀犿犪犻狕犲
ThekeytechnologyofGSisthemaizehybrid
predictionbyBLUPmodelwithmarkerseffectsor
coefficientofparentage.Itwasusedtopredictthe
singlecrossperformanceinmaizehybridbreeding
atfirst.TheBLUPmodelisestablishedbasedon
thetestedhybridsdataandthemarkersinforma
tionoftheirparents.Theperformanceofuntested
hybridsispredictedbytheBLUPmodelandthe
markersdataoftheparents[54].
Bernardodevotedhimselftohybridsprediction
byBLUPmodelinmaize[5558].Thecoefficientof
relativebetweentheoryandactualobservationwas
0.688~0.800byRFLPmarkers[54].BLUPissuit
ableforhybridperformancepredictionsincethe
traitonlyhasmoderateheritability.Predictionac
curacyofmolecularmarkereffectsishigherthan
phylogeneticrelationship[58].Withthedevelop
mentofmolecularmarkers,newmolecularmarker
typeemerged.Simplesequencerepeats(SSR)and
singlenucleotidepolymorphism(SNP)werewide
lyused.ManjeGowda犲狋犪犾.foundthatprediction
accuracyofflowertimeandplantheightwasabove
0.8withSSR markersinmaize[19].Researchof
Massman犲狋犪犾.indicatedthatpredictionaccuracy
ofgrainyieldwas0.8,androotloggingratiowas
0.87usingSSRmarkers[59].Buttheprediction
37216¾        A B,®:#$b,cdef67kGH°§¨¶NOR<·k¸Š(¿“)
effectofgrainyieldwasonly0.50~0.66,androot
loggingratiowasonly0.31~0.45withcoefficient
ofparentage[55].Thenitindicatedthatmolecular
markerswasmoresuitableforhybridperformance
predictionthancoefficientofparentage.
ThenscientistsfoundthatBLUPwasnotonly
usedtohybridperformanceprediction,butalsothe
breedingvalueofindividualsamongthemaizepop
ulation.SoBLUPwasusedtoindividualsselection
ofF2populationinselectionandbreedingofinbred
lines.Hybridperformancepredictionlaythefoun
dationforthegenomewideselectioninmaize.
3.2 犃狆狆犾犻犮犪狋犻狅狀狅犳犵犲狀狅犿犻犮狊犲犾犲犮狋犻狅狀犻狀犿犪犻狕犲
Bernardo’slaboratorybegantostudyapplying
GStomaizebreedinginMinnesotaUniversityof
America[21].Theydidplentyofsimulationandem
piricalexperiments.PiephoinGermanandRobert
inBrazilalsotriedtostudyusingGSin maize
breeding[6061].GSutilityinmaizebreedingconsist
oftwosides,hybridsperformancepredictionand
improvementofinbredlines.Hedevotedtoinbred
linesimprovementusingGS.TheBLUPmodelof
biparentalpopulationsfromtwoinbredlinesison
lysuitfortheprogenyoftheparents.Genomewide
selection as proposed in maize involves two
steps[21].First,asegregatingmaizepopulationis
genotypedandevaluatedfortestcrossperformance
ofF3family.Basedonthegenotypicandphenotyp
icdata,breedingvaluesassociatedwithalargeset
ofmarkers(e.g.,256to512markers)arecalcu
latedforthetraitsofinterest.Significancetests
formarkersarenotused,andtheeffectsofal
markersarefittedasrandomeffectsinalinear
modelbybestlinearunbiasedprediction(BLUP).
Second,twoorthreegenerationsofselectionbased
onalmarkersareconductedinayearroundnurs
ery(e.g.,HawaiorPuertoRico)orgreenhouse.
Traitvaluesarepredictedasthesumofanindivid
ualplant’smarkervaluesacrossal markers,and
selectionissubsequentlybasedonthesegenome
wideprediction.Accordingtothesteps,Emily
(2013b)introgressedsemidwarfgermplasmtoU.
S.Cornbeltinbredandfoundthatgenomewidese
lectionfromCycle1untilCycle5eithermaintained
orimprovedonthegainsfromphenotypicselection
achievedinCycle1[62].
TheresultsofBernardoindicatedthatauseful
strategyfortherapidimprovementofanadapted×
exoticcrossinvolves7to8cyclesofgenomewide
selectionstartingintheF2[14].Benjamin犲狋犪犾.
demonstratedthatprogressiveselfinghadasignifi
cantand positiveimpacton genomicselection
gains.Inparticular,selfingtotheF8produceda
72%increaseoverF2gains[63].However,mostof
thegainsarerealizedbytheF5generation(95%of
theF8gains).AlsonotethattheF8andDHper
formedsimilarly,consistentwithpreviousobser
vations[64].
IntheresearchofBernardo,thetrainingpop
ulationisthespecificbiparentalpopulationsfrom
thetwoparentallines,sotheBLUPmodelissuit
fortheprogenyofthetwoinbredlines.Otherex
perimentsofGSinmaizeareaboutmultiparental
populationsastrainingpopulation.StudyofYush
engZhaowasbasedonexperimentaldataofsix
segregatingpopulationsfromahalfdialelmating
design.Asformaizeuptothreegenerationsare
feasibleperyear,selectiongainperunittimeis
highand,consequently,genomicselectionholds
greatpromisefor maizebreeding programs[23].
Theseresultofthestudymightbeasgenomicpre
dictionmodelforfurtherbreedingelitemaizelines
betweenthesixpopulations.InthestudyofVan
essa犲狋犪犾.,markereffectsestimatedin255diverse
maizehybridswereusedtopredictgrainyield,an
thesisdate,andanthesissilkingintervalwithinthe
diversitypanelandtestcrossprogeniesof30F2de
rivedlinesfromeachoffivepopulations[36].Poten
tialusesforgenomicpredictionin maizehybrid
breedingarediscussedemphasizingtheneedof(1)
acleardefinitionofthebreedingscenarioinwhich
genomicpredictionshouldbeapplied(i.e.,predic
tionamongorwithinpopulations),(2)adetailed
analysisofthepopulationstructurebeforeper
formingcrossvalidation,and(3)largertraining
setswithstronggeneticrelationshiptothevalida
tionset.
4721 ! " # $ % &                   36½
4 Futureresearchinmaizebreeding
GSisjustbeginningtobeimplemented,butit
wiltakelongtimetobeusedinmaizebreeding.In
previousstudy,trainingpopulationwasonlyfrom
severalinbredlines,eveniftwoinbredlines.It
couldntbeimplementedbyotherbreedingpro
gram.Futureresearchshouldfocusontwosidesof
work.First,weshouldcommittobuildageneral
izedpredictionmodelforsomekindsofinbredlines
suchasyield,qualityandsoon.Butthesetraits
werecomplexcomposedofagreatdealofgenes.
TraditionalMAStechnologycouldntrealizethe
traitsselectioninmaizebreeding.973Plan“Basic
studyonbreedingofgenomewideselectionofyield
andqualitytraitsinmaize”hasbeencarriedoutin
2014.Theplanwilsystematiclyanalyzethege
neticbasisofmaizeyieldandquality,andthen
buildgenomewideselectionbreeding model.It
wil affordnewtechnologyfor maizebreeding.
Seond,inChina,abioticstresstolerancealsore
ducesthe yield seriously in maize especialy
droughttolerance.Droughtistheforemostfactor
restrictingmaizeproduction,oftenresultingin20
-50% maizeyieldreductioneveryyearinChi
na[65].Ifweestablishpredictionmodelofdrought
tolerance,itwilaffordthetheoryandtechnology
supportofmaizebreeding.Consequently,ourre
searchteamwilcarriedoutstudyonthegenomic
selectionprogramofdroughttolerance.
犚犲犳犲狉犲狀犮犲狊:
[1] STUBERCW,POLACCOM,SENIORML.Synergyofem
piricalbreeding,markerassistedselection,andgenomicsto
increasecropyieldpotential[J].犆狉狅狆犛犮犻犲狀犮犲,1999,39:
15711583.
[2] MOOSESP,MUMMRH.Molecularplantbreedingasthe
foundationfor21stcenturycropimprovement[J].犘犾犪狀狋
犘犺狔狊犻狅犾狅犵狔,2008,147:969977.
[3] BERNARDOR.Molecularmarkersandselectionforcomplex
traitsinplants:learningfromthelast20years[J].犆狉狅狆犛犮犻
犲狀犮犲,2008,48:16491664.
[4] MEUWISSENTH,HAYESBJ,GODDARD ME.Predic
tionoftotalgeneticvalueusinggenomewidedensemarker
maps[J].犌犲狀犲狋犻犮狊,2001,157:18191829.
[5] JANNINKJL,LORENZAJ,IWATAH.Genomicselection
inplantbreeding:fromtheorytopractice[J].犅狉犻犲犳犻狀犵狊犻狀
犉狌狀犮狋犻狅狀犪犾犌犲狀狅犿犻犮狊,2010,9(2):166177.
[6] HEFFNEREL,JANNINKJL,IWATAH,犲狋犪犾.Genomic
selectionaccuracyforgrainqualitytraitsinbiparentalwheat
populations[J].犆狉狅狆犛犮犻犲狀犮犲,2011,51:25972606.
[7] HEFFNEREL,SORRELLSME,JANNINKJL.Genomic
selectionforcropimprovement[J].犆狉狅狆犛犮犻犲狀犮犲,2009,49:
112.
[8] MAYORPJ,BERNARDOR.Genomewideselectionand
markerassistedrecurrentselectionindoubledhaploidversus
F2populations[J].犆狉狅狆犛犮犻犲狀犮犲,2009,49:17191725.
[9] MASSMANJM,JUNGHJG,BERNARDOR.Genome
wideselectionversusmarkerassistedrecurrentselectiontoim
provegrainyieldandstoverqualitytraitsforcelulosicethanol
inmaize[J].犆狉狅狆犛犮犻犲狀犮犲,2012,53(1):5866.
[10] SCHAEFFERLR.Strategyforapplyinggenomewideselec
tionindairycattle[J].犑狅狌狉狀犪犾狅犳犃狀犻犿犪犾犅狉犲犲犱犻狀犵犌犲狀犲狋犻犮,
2006,123:218223.
[11] GODDARDME,HAYESBJ.Genomicselection[J].犑狅狌狉
狀犪犾狅犳犪狀犻犿犪犾犅狉犲犲犱犻狀犵犌犲狀犲狋犻犮狊,2007,124:323330.
[12] DAETWYLER H D,VILLANUEVA B,BIJMA P.In
breedingingenomewideselection[J].犑狅狌狉狀犪犾狅犳犃狀犻犿犪犾
犅狉犲犲犱犻狀犵犌犲狀犲狋犻犮,2007,124:369376.
[13] TUL,WOOLLIAMSJA,SIGBJORNL.Theaccuracyof
genomicselectioninnorwegianredcattleassessedbycross
validation[J].犌犲狀犲狋犻犮狊,2009,183:11191126.
[14] BERNARDOR.Genomewideselectionforrapidintrogres
sionofexoticgermplasminmaize[J].犆狉狅狆犛犮犻犲狀犮犲,2009,
49:419425.
[15] HANSDD,BANSALUK,BARIANAHS,犲狋犪犾.Genom
icpredictionforrustresistanceindiversewheatlandraces[J].
犜犺犲狅狉狔犪狀犱犃狆狆犾犻犲犱犌犲狀犲狋犻犮狊,2014,127:17951803.
[16] MARIED,BOUVETJM.Genomicselectionintreebreed
ing:testingaccuracyofprediction modelsincludingdomi
nanceeffect[J].犅犕犆犘狉狅犮犲犲犱犻狀犵狊,2011,5(Supply7):12.
[17] WRSCHUMT,REIFJC,KRAFTT,犲狋犪犾.Genomicse
lectioninsugarbeetbreedingpopulations[J].犅犕犆犌犲狀犲狋犻犮狊,
2013,14:8592.
[18] ZHONGSQ,DEKKERSJCM,FERNANDORL,犲狋犪犾.
Factorsaffectingaccuracyfromgenomicselectioninpopula
tionsderivedfrommultipleinbredlines:abarleycasestudy
[J].犌犲狀犲狋犻犮狊,2009,182(1):355364.
[19] GOWDAM,ZHAOYS,MAURERHP,犲狋犪犾.Bestlinear
57216¾        A B,®:#$b,cdef67kGH°§¨¶NOR<·k¸Š(¿“)
unbiasedpredictionoftriticalehybridperformance[J].犈狌
狆犺狔狋犻犮犪,2013,191:223230.
[20] ÀÁÂ,ÃÄÅ,ÆÇ],®.#${jrªb,cdef
GH°
[J].![34%&,2012,25(4):15101514.
WUYS,SHAOJM,ZHOURY,犲狋犪犾.Reviewsofge
nomewideselectionforquantitativetraitsinplants[J].
犛狅狌狋犺狑犲狊狋犆犺犻狀犪犑狅狌狉狀犪犾狅犳犃犵狉犻犮狌犾狋狌狉犪犾犛犮犻犲狀犮犲狊,2012,25
(4):15101514.
[21] BERNARDOR,YUJ.Prospectsforgenomewideselection
forquantitativetraitsinmaize[J].犆狉狅狆犛犮犻犲狀犮犲,2007,47:
10821090.
[22] EMILYC,BERNARDOR.Accuracyofgenomewideselec
tionfordifferenttraitswithconstantpopulationsize,herita
bility,andnumberofmarkers[J].犘犾犪狀狋犌犲狀狅犿犲,2013a,6
(1):17.
[23] ZHAOYS,GOWDAM,LIU WX,犲狋犪犾.Accuracyofge
nomicselectioninEuropeanmaizeelitebreedingpopulations
[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱 犃狆狆犾犾犻犲犱 犌犲狀犲狋犻犮狊,2012a,124:
769776.
[24] HESLOTN,YANGHP,SORRELLSME,犲狋犪犾.Genomic
selectioninplantbreeding:acomparisonofmodels[J].犆狉狅狆
犛犮犻犲狀犮犲,2012,52:146160.
[25] CHENX,SULLIVANPF.Singlenucleotidepolymorphism
genotyping:biochemistry,protocol,costandthroughput
[J].犘犺犪狉犿犪犮狅犌犲狀犲狋犻犮狊,2003,3:7796.
[26] POLANDJ,RIFETW.Genotypingbysequencingforplant
breedingandgenetics[J].犘犾犪狀狋犌犲狀犲狋犻犮狊,2012,5:92102.
[27] HABIERD,FERNANDORL,DEKKERSJCM.Theim
pactofgeneticrelationshipinformationongenomeassisted
breedingvalues[J].犌犲狀犲狋犻犮狊,2007,177:23892397.
[28] DAETWYLER H D,VILLANUEVAB,WOOLLIAMSJ
A.Accuracyofpredictingthegeneticriskofdiseaseusinga
genomewideapproach[J].犘犔狅犛犗狀犲,2008,3:3395.
[29] ALBRECHTT,WIMMERV,AUINGERHJ,犲狋犪犾.Ge
nomebasedpredictionoftestcrossvaluesinmaize[J].犜犺犲狅
狉犲狋犻犮犪犾犪狀犱犃狆狆犾犾犻犲犱犌犲狀犲狋犻犮狊,2011,123:339350
[30] CLARKS,HICKEYJ,WERFJ.Differentmodelsofgenetic
variationandtheireffectongenomicevaluation[J].犌犲狀犲狋犻犮
犛犲犾犲犮狋犻狅狀犈狏狅犾狌狋犻狅狀,2011,43:18.
[31] PSZCZOLAM,STRABELT,MULDERHA,犲狋犪犾.Relia
bilityofdirectgenomicvaluesforanimalswithdifferentrela
tionshipswithinandtothereferencepopulation[J].犑狅狌狉狀犪犾
狅犳犇犪犻狉狔犛犮犻犲狀犮犲,2012,95z:389400.
[32] SAATCHIM,MCCLUREMC,MCKAYSD,犲狋犪犾.Accu
raciesofgenomicbreedingvaluesinAmericanAngusbeef
cattleusingkmeansclusteringforcrossvalidation[J].犌犲
狀犲狋犻犮犛犲犾犲犮狋犻狅狀犈狏狅犾狌狋犻狅狀,2011,43:40.
[33] WINDHAUSENVS,ATLINGN,CROSSAJ,犲狋犪犾.Ef
fectivenessofgenomicpredictionofmaizehybridperformance
indifferentbreedingpopulationsandenvironments[J].犌犲狀犲狊
犌犲狀狅犿犲狊犌犲狀犲狋犻犮,2012,2:14271436.
[34] GUOZ,TUCKERDM,BASTENCJ,犲狋犪犾.Theimpactof
populationstructureongenomicpredictioninstratifiedpopu
lations[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱犃狆狆犾犾犻犲犱犌犲狀犲狋犻犮狊,2014,127:
749762
[35] HABIERD,FERNANDORL,DEKKERSJCM.Genomic
selectionusinglowdensity markerpanels[J].犌犲狀犲狋犻犮狊,
2009,182:343353.
[36] VANESSASW,ATLINGN,HICKEYJM,犲狋犪犾.Effec
tivenessofgenomicpredictionofmaizehybridperformancein
differentbreedingpopulationsandenvironments[J].犌犲狀狅犿犻犮
犛犲犾犲犮狋犻狅狀,2012,2(14):14271436.
[37] WEGENASTT,LONGINCFH,UTZHF,犲狋犪犾.Hybrid
maizebreedingwithdoubledhaploidsIV.Numberversussize
ofcrossesandimportanceofparentalselectionintwostage
selectionfortestcrossperformance[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱犃狆
狆犾犾犻犲犱犌犲狀犲狋犻犮狊,2008,117:251260.
[38] SOLBERGTR,SONESSONAK,WOOLLIAMSJA,犲狋
犪犾.Genomicselectionusingdifferentmarkertypesanddensi
ties[J].犑狅狌狉狀犪犾狅犳犃狀犻犿犪犾犅狉犲犲犱犻狀犵犌犲狀犲狋犻犮狊,2008,86
(10):24472454.
[39] LORENZANARE,BERNARDOR.Accuracyofgenotypic
valuepredictionsfor markerbasedselectioninbiparental
plantpopulations[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱 犃狆狆犾犻犲犱 犌犲狀犲狋犻犮狊,
2009,120:151161.
[40] WOLDH,JOHNSONNL,KOTZS.Partialleastsquares
[C].EncyclopediaofStatisticalScience.NewYork:Wiley,
1985:58191.
[41] SOLBERGTR,SONESSONAK,WOOLLIAMSJA,犲狋
犪犾.Reducingdimensionalityforpredictionofgenomewide
breedingvalues[J].犌犲狀犲狋犻犮犛犲犾犲犮狋犻狅狀犈狏狅犾狌狋犻狅狀,2009,41
(1):29.
[42] CROSSAJ,CAMPOSG,P?REZP,犲狋犪犾.Predictionofge
neticvaluesofquantitativetraitsinplantbreedingusingpedi
greeandmolecularmarkers[J].犌犲狀犲狋犻犮狊,2010,186(2):
713724.
[43] MEUWISSENTHE,SOLBERGTR,SHEPHERDR,犲狋
犪犾.AfastalgorithmforBayesBtypeofpredictionofgenome
wideestimatesofgeneticvalue[J].犌犲狀犲狋犻犮犛犲犾犲犮狋犻狅狀犈狏狅犾狌
狋犻狅狀,2009,41:2.
[44] WANG W YS,BARRATTBJ,CLAYTONDG,犲狋犪犾.
Genomewideassociationstudies:theoreticalandpractical
concerns[J].犖犪狋狌狉犲犚犲狏犻犲狑犌犲狀犲狋犻犮狊,2005,6(2):109118.
[45] ÈÉÊ,ËÌÍ,Až“.,cdef§¨¸Š[J].PQ,
2011,33(12):13081316.
LIHD,BAOZM,SUNXW.Genomicselectionanditsap
plication[J].犎犲狉犲犱犻狋犪狊,2011,33(12):13081316.
[46] SASInstitute.TheSASsystemforWindows.Release9.2.
6721 ! " # $ % &                   36½
SASInst.,Cary,NC,2009.
[47] LUNDMS,SAHANAG,KONINGDJ,犲狋犪犾.Comparison
ofanalysesoftheQTLMASXIIcommondataset.I:Genomic
selection[J].犅犕犆犘狉狅犮犲犲犱犻狀犵狊,2009,3(Suppl.1):S1.
[48] BASTIAANSENJW,COSTER A,CALUSM P,犲狋犪犾.
Longtermresponsetogenomicselection:effectsofestima
tionmethodandreferencepopulationstructurefordifferent
geneticarchitectures[J].犌犲狀犲狋犻犮狊犛犲犾犲犮狋犻狅狀犈狏狅犾狌狋犻狅狀,2012,
4:316.
[49] OGUTUJO,PIEPHOHP,TORBENSS.Acomparison
ofrandomforests,boostingandsupportvectormachinesfor
genomicselection[J].犅犕犆犘狉狅犮犲犲犱犻狀犵狊,2011,5(Suppl3):
S11.
[50] DAETWYLERHD,WONGRP,VILLANUEVAB,犲狋犪犾.
Theimpactofgeneticarchitectureongenomewideevaluation
methods[J].犌犲狀犲狋犻犮狊,2010,185:10211031.
[51] JULIOI,JANNINKJL,AKDEMIRD,犲狋犪犾.Trainingset
optimizationunderpopulationstructureingenomicselection
[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱 犃狆狆犾犾犻犲犱 犌犲狀犲狋犻犮狊,2014,128(1):
145158.
[52] NAVABIA,MATHERDE,BERNIERJ,犲狋犪犾.QTLde
tectionwithbidirectionalandunidirectionalselectivegenoty
ping:markerbasedandtraitbasedanalyses[J].犜犺犲狅狉犲狋犻犮犪犾
犪狀犱犃狆狆犾犾犻犲犱犌犲狀犲狋犻犮狊,2009,118:347358.
[53] ZHAOYS,GOWDA M,LONGINFH,犲狋犪犾.Impactof
selectivegenotypinginthetrainingpopulationonaccuracy
andbiasofgenomicselection[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱犃狆狆犾犾犻犲犱
犌犲狀犲狋犻犮狊,2012b,125:707713.
[54] ÎÏÐ,Ñ}Ò.¸Š{ÓPQ%[M]."Ô:h34+%
67ÕÖ
,2007:185204.
ZHAIH Q,WANGJK.Applied QuantitativeGenetics
[M].Beijing:ChinaAgriculturalScienceandTechnology
PublishingHouse,2007:185204.
[55] BERNARDOR.Predictionofmaizesinglecrossperformance
usingRFLPsandinformationfromrelatedhybrids[J].犆狉狅狆
犛犮犻犲狀犮犲,1994,34:2025.
[56] BERNARDOR.Geneticmodelsforpredictingmaizesingle
crossperformanceinunbalancedyieldtrialdata[J].犆狉狅狆
犛犮犻犲狀犮犲,1995,35:141147.
[57] BERNARDOR.Bestlinearunbiasedpredictionofmaizesin
glecrossperformance[J].犆狉狅狆犛犮犻犲狀犮犲,1996,36:5056.
[58] BERNARDOR.Markerassistedbestlinearunbiasedpredic
tionofsinglecrossperformance[J].犆狉狅狆犛犮犻犲狀犮犲,1999,39:
12771282.
[59] MASSMANJM,GORDILLO A,LORENZANARE,犲狋
犪犾.Genomewidepredictionsfrommaizesinglecrossdata[J].
犜犺犲狅狉犲狋犻犮犪犾犪狀犱犃狆狆犾犾犻犲犱犌犲狀犲狋犻犮狊,2013,126:1322.
[60] PIEPHOHP.Ridgeregressionandextensionsforgenome
wideselectioninmaize[J].犆狉狅狆犛犮犻犲狀犮犲,2009,49:1165
1176.
[61] ROBERTOFN,JULIOCD,?DERCML,犲狋犪犾.Genome
wideselectioninfortropicalmaizeroottraitsunderconditions
ofnitrogenandphosphorusstress[J].犃犮狋犪犛犮犻犲狀狋犻犪狉狌犿,
2012,34(4):389395.
[62] EMILYC,BERNARDOR.Genomewideselectiontointro
gresssemidwarfmaizegermplasmintoU.S.CornBeltin
breds[J].犆狉狅狆犛犮犻犲狀犮犲,2013b,53:14271436.
[63] BENJAMINM,COMBEJL,TANKSLEYSD.Selfingfor
thedesignofgenomicselectionexperimentsinbiparental
plantpopulations[J].犜犺犲狅狉犲狋犻犮犪犾犪狀犱 犃狆狆犾犾犻犲犱 犌犲狀犲狋犻犮狊,
2013,126:29072920.
[64] BORDESJ,CHARMETG,VAULXRD,犲狋犪犾.Doubled
haploidversussingleseeddescentandS1familyvariationfor
testcrossperformanceinamaizepopulation[J].犈狌狆犺狔狋犻犮犪,
2007,154:4151.
[65] HU,RF,MENGEC,ZHANGSH,犲狋犪犾.Prioritization
formaizeresearchanddevelopmentinChina[J].犛犮犻犲狀狋犻犪
犃犵狉犻犮狌犾狋狌狉犪犛犻狀犻犮犪,2004,37:781787.

!"

#$%
)  
77216¾        A B,®:#$b,cdef67kGH°§¨¶NOR<·k¸Š(¿“)