exploring specific combining abilies in incomplete factorial cross … · 2019-01-22 · (yates,...

1
Exploring specific combining abilies in incomplete factorial cross designs with LABKEY soware Soula, Julie¹*; Vincourt, Patrick²; Bernard, Jean-Marc³; Bardet, Sébasen¹; Royer, Frédéric³ 1 Doriane SAS, 31 avenue Jean Médecin, 06000 Nice, France ; 2 Open-Source-Biology, 12 avenue V. Segoffin, 31400 Toulouse, France ; 3 Biosearch Data Management, 39 boulevard Dubouchage, 06000 Nice, France Results and discussion The first two components of SCA are accounng for the major part of the {female * male} interacon (55% of the total variance): see Figure 1. Based on these first two components, 2 there is a good adequacy (R = 0.68) between the observed and calculated SCA for observed combinaons, as shown in Figure 2. We therefore propose to use the same predictor for unobserved combinaons. In this way, we obtain a table of (posive and negave) predicted interacons: see Figure 3. TM Using the crosses matrix available in LABKEY , it is possible to display all the informaon on the parents and on the hybrids (crosses done, predicons for the non-observed combinaons) to help the breeder to choose his future crosses and to automacally TM generate them in LABKEY : see Figure 4. As pointed out by Technow et al. (2014), hybrid breeders are oen tesng the combining ability with a limited number and range of testers, thus liming the possibility to explore a wider range of combinaon for the detecon of parcular hybrid combinaon. This example shows how, with the same investment in hybrid producon and experimentaon, it becomes possible to explore a wider space of genec diversity and to propose inferences for unobserved combinaons. References 1. Soula J, Duminil T, Bardet S, Bernard JM, Royer F (2014) Consideraon Of Field Heterogeneies In The Calculaon Of Variety Means From Agronomic Trial Data: Comparisons Between A Nearest Neighbor Adjustment Method And Experimental Designs Like Laces And Lanized Alpha-Designs. Plant Animal genome Conf., San Diego 2. Technow F, Schrag TA, Schipprack W, Bauer E, Simianer H, Melchinger AE. (2014) Genome Properes and Prospects of Genomic Predicon of Hybrid Performance in a Breeding Program of Maize. GENETICS 197 (4): 1343-135 3. Yates F. (1933). The analysis of replicated experiments when the field results are incomplete. Emp. Jour. Exp. Agr. 1, 129-142 Introducon One of the biggest challenges for plant breeders is to deal with interacons. Parcularly, they aim at idenfying which {female * male} combinaon is offering the most promising specific combining ability (SCA). Aer the pioneer works of Tukey (1949) and Mandel (1971), the AMMI (for “Addive main effects and mulplicave interacon”) framework became very popular during the last decade. Based on a praccal example TM analyzed using LABKEY , this study is intending to show how this soware is providing the users with tools helping them to predict and display specific combining abilies (SCA) for non-observed hybrid combinaons. These predicons are made using Principal Component Analysis (PCA) results obtained on the observed crosses, under the hypothesis that they represent a connected, unbiased subset of the complete set of possible combinaons. General Framework The AMMI approach is modeling the result of the combinaon of two factors (Female and Male, hereaer named I and J) as the sum of addive, main effects and of mulplicave effects of both factors, those being actually calculated through a PCA on interacon esmates: Where m is the grand mean, a and b are the main, addive effects of the factors I and J, i j and ab represents their interacon: c and d are the coordinates of the component k of ij ik jk th the interacon, and U is the k component eigenvalue. k Material & Method 25 female lines were crossed with 8 male lines to produce 90 hybrids, i.e. with a high proporon of missing data. Each female line was crossed with a minimum of two male lines. Interacons between male and female parents (= SCA) were obtained in subtracng the esmates of addive effects of the male parent and of the female parent from the TM observed yield. In the LABKEY soware, these esmates are obtained through Yates' (Yates, 1933) method of esmaon of missing data, a method which implicitly assumes a null interacon between the unobserved {female parent * male parent} combinaons. These esmates are also those produced by the LSMEANS statement of the GLM procedure under SAS© with a model including only addive effects. PCA was performed on the FxM interacon esmates and we then used the SCA components on the first principal components to predict the SCA for unobserved combinaons. « With the same investment [..], explore a wider space of genec diversity and propose inferences for unobserved combinaons. » Conclusion LABKEY is currently developing tools to help breeders in their aim to deal with various types of interacon. This study shows an example dealing with Female x Male interacons and mainly using two LABKEY tools, the PCA and the crosses matrix, configured to compute and display SCA predicons for a subset of non-observed hybrids. Figure 1 (le): Male and female parents on the first two principal components [graphs TM obtained with LABKEY soware]. Figure 2 (right): Modeling the specific combining abilies (SCA): the first two components are accounng for 68% of the SCA variaon on observed combinaons [graph obtained with R soware]. - 8 , 0 0 -6, 0 0 - 4 , 0 0 - 2 , 0 0 0 , 0 0 2,0 0 4 , 0 0 6 , 0 0 8 , 0 0 PCA Axis 1 : 30,74 % -8,00 -6,00 -4,00 -2,00 0,00 2,00 4,00 6,00 8,00 PCA Axis 2 : 24,39 % FAN FASHION FAST FATE FEBRUARY FELIX FIDJI FIRST FISH FJORD FLAME FLAMENCO FLORA FLOWER FLUOR FOG FORK FOSSIL FOX FREE FRESH FRIEND FUCHSIA FUN FURY C0100---MILK C0100---MISTER C0100---MEGA C0100---MILES C0100---MAD C0100---MAGNET C0100---MAGIC C0100---MICRO Figure 3: Predicted interacon values using the first two principal components. Black: Predicted interacon values for tested hybrids. Red (negave), green (posive): predicted interacon values for non tested hybrids. Grey: General Combining Abilies (GCA). [Report generated from LABKEY soware]. MAD MAGIC MAGNET MEGA MICRO MILES MILK MISTER GCA - F FAN -1,36 0,37 -1,09 2,56 1,01 -1,83 1,73 -1,38 2,49 FASHION -0,04 0,03 -0,1 0,17 -0,01 -0,18 0,16 -0,03 -0,86 FAST -1,26 -0,07 0,45 0,19 1,9 1,15 -0,8 -1,55 -4,27 FATE 0,03 0,04 -0,15 0,2 -0,13 -0,29 0,24 0,06 1,88 FEBRUARY -0,89 0,01 0,09 0,47 1,18 0,36 -0,19 -1,05 6,16 FELIX 0,57 -0,07 0,17 -0,64 -0,62 0,2 -0,25 0,64 -2,53 FIDJI -0,08 -0,02 0,07 -0,06 0,15 0,16 -0,12 -0,1 0,54 FIRST 6,89 -0,75 1,61 -7,1 -7,66 1,66 -2,33 7,69 2,11 FISH -0,5 0,12 -0,34 0,85 0,4 -0,56 0,54 -0,51 -1,00 FJORD 0,77 0,18 -0,74 0,58 -1,47 -1,61 1,26 1,04 -2,74 FLAME -0,15 0,18 -0,63 1,04 -0,22 -1,19 1,03 -0,06 -2,69 FLAMENCO 0,41 0,17 -0,66 0,7 -0,96 -1,37 1,1 0,61 -3,51 FLORA 1,06 -0,13 0,29 -1,16 -1,15 0,34 -0,43 1,18 8,82 FLOWER -0,06 -0,02 0,08 -0,08 0,14 0,17 -0,14 -0,09 3,11 FLUOR 0,11 -0,39 1,37 -2,12 0,77 2,65 -2,25 -0,13 6,33 FOG -0,03 0,11 -0,37 0,57 -0,2 -0,71 0,61 0,03 -6,00 FORK 1,22 -0,49 1,56 -3,16 -0,5 2,78 -2,51 1,11 -0,36 FOSSIL -3,82 -0,75 3,22 -2,22 6,98 7,12 -5,49 -5,05 3,71 FOX 0,47 0,56 -2,06 2,76 -1,96 -4,12 3,42 0,94 -5,00 FREE 0,01 0 0 -0,01 -0,01 0 0 0,01 -7,10 FRESH 2,09 0,11 -0,7 -0,37 -3,12 -1,83 1,26 2,56 -2,41 FRIEND -3,04 0,65 -1,83 4,81 2,63 -2,93 2,88 -3,17 -0,85 FUCHSIA -0,02 -0,46 1,62 -2,41 1,1 3,16 -2,67 -0,33 4,54 FUN -2,23 0,61 -1,82 4,24 1,61 -3,08 2,9 -2,24 -3,21 FURY -0,16 0,02 -0,05 0,18 0,17 -0,06 0,07 -0,17 2,83 GCA - M -2,30 1,67 4,25 1,24 2,04 -2,63 -3,40 -0,87 TM Figure 4: Generaon of new crosses in LABKEY (picture on the boom right) from the crosses matrix (background picture). The GCA of the parents is always displayed whereas the predicted SCA for non tested hybrids (tested hybrids are in grey) are displayed on the toolp. Hybrids for which the SCA predicon is larger than 0.5 or smaller than -0.5 are highlighted by colored squares: red colors for negave predicons and green colors for posive predicons. www.doriane.com -8 -6 -4 -2 0 2 4 6 8 10 -8 -6 -4 -2 0 2 4 6 8 R2 = 0.677 Predicted interaction values with the first two principal components Observed interaction values

Upload: others

Post on 11-Mar-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Exploring specific combining abilies in incomplete factorial cross … · 2019-01-22 · (Yates, 1933) method of esmaon of missing data, a method which implicitly assumes a null

Exploring specific combining abili�es

in incomplete factorial cross designs withLABKEY�� so�ware

Soula, Julie¹*; Vincourt, Patrick²; Bernard, Jean-Marc³; Bardet, Sébas�en¹; Royer, Frédéric³1 Doriane SAS, 31 avenue Jean Médecin, 06000 Nice, France ; 2 Open-Source-Biology, 12 avenue V. Segoffin, 31400 Toulouse, France ; 3 Biosearch Data Management, 39 boulevard Dubouchage, 06000 Nice, France

Results and discussionThe first two components of SCA are accoun�ng for the major part of the {female * male} interac�on (55% of the total variance): see Figure 1. Based on these first two components,

2there is a good adequacy (R = 0.68) between the observed and calculated SCA for observed combina�ons, as shown in Figure 2. We therefore propose to use the same predictor for unobserved combina�ons. In this way, we obtain a table of (posi�ve and nega�ve) predicted interac�ons: see Figure 3.

TMUsing the crosses matrix available in LABKEY , it is possible to display all the informa�on on the parents and on the hybrids (crosses done, predic�ons for the non-observed combina�ons) to help the breeder to choose his future crosses and to automa�cally

TMgenerate them in LABKEY : see Figure 4.As pointed out by Technow et al. (2014), hybrid breeders are o�en tes�ng the combining ability with a limited number and range of testers, thus limi�ng the possibility to explore a wider range of combina�on for the detec�on of par�cular hybrid combina�on. This example shows how, with the same investment in hybrid produc�on and experimenta�on, it becomes possible to explore a wider space of gene�c diversity and to propose inferences for unobserved combina�ons.

References 1. Soula J, Duminil T, Bardet S, Bernard JM, Royer F (2014) Considera�on Of Field Heterogenei�es In The Calcula�on Of Variety Means From Agronomic Trial Data:

Comparisons Between A Nearest Neighbor Adjustment Method And Experimental Designs Like La�ces And La�nized Alpha-Designs. Plant Animal genome Conf., San Diego 2. Technow F, Schrag TA, Schipprack W, Bauer E, Simianer H, Melchinger AE. (2014) Genome Proper�es and Prospects

of Genomic Predic�on of Hybrid Performance in a Breeding Program of Maize. GENETICS 197 (4): 1343-135 3. Yates F. (1933). The analysis of replicated experiments whenthe field results are incomplete. Emp. Jour. Exp. Agr. 1, 129-142

Introduc�onOne of the biggest challenges for plant breeders is to deal with interac�ons. Par�cularly, they aim at iden�fying which {female * male} combina�on is offering the most promising specific combining ability (SCA). A�er the pioneer works of Tukey (1949) and Mandel (1971), the AMMI (for “Addi�ve main effects and mul�plica�ve interac�on”) framework became very popular during the last decade. Based on a prac�cal example

TManalyzed using LABKEY , this study is intending to show how this so�ware is providing the users with tools helping them to predict and display specific combining abili�es (SCA) for non-observed hybrid combina�ons. These predic�ons are made using Principal Component Analysis (PCA) results obtained on the observed crosses, under the hypothesis that they represent a connected, unbiased subset of the complete set of possible combina�ons.

General FrameworkThe AMMI approach is modeling the result of the combina�on of two factors (Female and Male, herea�er named I and J) as the sum of addi�ve, main effects and of mul�plica�ve effects of both factors, those being actually calculated through a PCA on interac�on es�mates:

Where m is the grand mean, a and b are the main, addi�ve effects of the factors I and J, i j

and ab represents their interac�on: c and d are the coordinates of the component k of ij ik jkththe interac�on, and U is the k component eigenvalue.k

Material & Method25 female lines were crossed with 8 male lines to produce 90 hybrids, i.e. with a high propor�on of missing data. Each female line was crossed with a minimum of two male lines. Interac�ons between male and female parents (= SCA) were obtained in subtrac�ng the es�mates of addi�ve effects of the male parent and of the female parent from the

TMobserved yield. In the LABKEY so�ware, these es�mates are obtained through Yates' (Yates, 1933) method of es�ma�on of missing data, a method which implicitly assumes a null interac�on between the unobserved {female parent * male parent} combina�ons. These es�mates are also those produced by the LSMEANS statement of the GLM procedure under SAS© with a model including only addi�ve effects. PCA was performed on the FxM interac�on es�mates and we then used the SCA components on the first principal components to predict the SCA for unobserved combina�ons.

« With the same investment [..], explore a wider space of gene�c diversity and propose inferences for unobserved combina�ons. »

ConclusionLABKEY�� is currently developing tools to help breeders in their aim to deal with various types of interac�on. This study shows an example dealing with Female x Male interac�ons and mainly using two LABKEY�� tools, the PCA and the crosses matrix, configured to compute and display SCA predic�ons for a subset of non-observed hybrids.

Figure 1 (le�): Male and female parents on the first two principal components [graphs TMobtained with LABKEY so�ware].

Figure 2 (right): Modeling the specific combining abili�es (SCA): the first two components are

accoun�ng for 68% of the SCA varia�on on observed combina�ons [graph obtained with R so�ware].

-8,00

-6,00

-4,00

-2,00

0,00

2,00

4,00

6,00

8,00

PCA Axis 1 : 30,74 %

-8,00

-6,00

-4,00

-2,00

0,00

2,00

4,00

6,00

8,00

PCA Axis 2 : 24,39 %

FAN

FASHION

FAST

FATE

FEBRUARYFELIXFIDJI

FIRST FISH

FJORDFLAME

FLAMENCO

FLORAFLOWER

FLUOR

FOG

FORK

FOSSIL

FOX

FREE

FRESHFRIEND

FUCHSIA

FUN

FURY

C0100---MILK

C0100---MISTER

C0100---MEGA

C0100---MILES

C0100---MAD

C0100---MAGNET

C0100---MAGIC

C0100---MICRO

Figure 3: Predicted interac�on values using the first two principal components. Black: Predicted interac�on

values for tested hybrids. Red (nega�ve), green (posi�ve): predicted interac�on values for non tested hybrids. Grey: General Combining Abili�es (GCA). [Report generated from LABKEY�� so�ware].

MAD MAGIC MAGNET MEGA MICRO MILES MILK MISTER GCA - F FAN -1,36 0,37 -1,09 2,56 1,01 -1,83 1,73 -1,38 2,49

FASHION -0,04 0,03 -0,1 0,17 -0,01 -0,18 0,16 -0,03 -0,86 FAST -1,26 -0,07 0,45 0,19 1,9 1,15 -0,8 -1,55 -4,27 FATE 0,03 0,04 -0,15 0,2 -0,13 -0,29 0,24 0,06 1,88

FEBRUARY -0,89 0,01 0,09 0,47 1,18 0,36 -0,19 -1,05 6,16 FELIX 0,57 -0,07 0,17 -0,64 -0,62 0,2 -0,25 0,64 -2,53 FIDJI -0,08 -0,02 0,07 -0,06 0,15 0,16 -0,12 -0,1 0,54 FIRST 6,89 -0,75 1,61 -7,1 -7,66 1,66 -2,33 7,69 2,11 FISH -0,5 0,12 -0,34 0,85 0,4 -0,56 0,54 -0,51 -1,00

FJORD 0,77 0,18 -0,74 0,58 -1,47 -1,61 1,26 1,04 -2,74 FLAME -0,15 0,18 -0,63 1,04 -0,22 -1,19 1,03 -0,06 -2,69

FLAMENCO 0,41 0,17 -0,66 0,7 -0,96 -1,37 1,1 0,61 -3,51 FLORA 1,06 -0,13 0,29 -1,16 -1,15 0,34 -0,43 1,18 8,82

FLOWER -0,06 -0,02 0,08 -0,08 0,14 0,17 -0,14 -0,09 3,11 FLUOR 0,11 -0,39 1,37 -2,12 0,77 2,65 -2,25 -0,13 6,33

FOG -0,03 0,11 -0,37 0,57 -0,2 -0,71 0,61 0,03 -6,00 FORK 1,22 -0,49 1,56 -3,16 -0,5 2,78 -2,51 1,11 -0,36

FOSSIL -3,82 -0,75 3,22 -2,22 6,98 7,12 -5,49 -5,05 3,71 FOX 0,47 0,56 -2,06 2,76 -1,96 -4,12 3,42 0,94 -5,00

FREE 0,01 0 0 -0,01 -0,01 0 0 0,01 -7,10 FRESH 2,09 0,11 -0,7 -0,37 -3,12 -1,83 1,26 2,56 -2,41 FRIEND -3,04 0,65 -1,83 4,81 2,63 -2,93 2,88 -3,17 -0,85

FUCHSIA -0,02 -0,46 1,62 -2,41 1,1 3,16 -2,67 -0,33 4,54 FUN -2,23 0,61 -1,82 4,24 1,61 -3,08 2,9 -2,24 -3,21

FURY -0,16 0,02 -0,05 0,18 0,17 -0,06 0,07 -0,17 2,83

GCA - M -2,30 1,67 4,25 1,24 2,04 -2,63 -3,40 -0,87

TM Figure 4: Genera�on of new crosses in LABKEY (picture on the bo�om right) from the crosses

matrix (background picture). The GCA of the parents is always displayed whereas the predicted SCA for non tested hybrids (tested hybrids are in grey) are displayed on the tool�p. Hybrids for which the SCA predic�on is larger than 0.5 or smaller than -0.5 are highlighted by colored squares: red colors for nega�ve predic�ons and green colors for posi�ve predic�ons.

www.doriane.com

-8 -6 -4 -2 0 2 4 6 8 10

-8-6

-4-2

02

46

8

R2 = 0.677

Pre

dic

ted

inte

ract

ion

va

lue

s w

ith t

he

firs

t tw

o p

rin

cip

al c

om

po

ne

nts

Observed interaction values