International Biometric Society
APPLICATION OF MULTIVARIATE METHODS TO EVALUATE FATTY ACIDS IN
SOYBEAN CULTIVARS
OLIVEIRA, M.C.N. de1, CARRÃO-PANIZZI, M.C2, MANDARINO, J.M.G1.
1Embrapa Soja, Caixa Postal 231, CEP 86001-970 Londrina-PR, e-mail:
mcristina.oliveira@embrapa.br 2Embrapa Trigo, Caixa Postal 451, CEP 99001-970
Abstract – Increased consumption of soyfoods and the world's health problems related to high
cholesterol levels, justify surveys to assess levels of fatty acids in soybean grains. Soybean presents
high concentration of essential unsaturated acids (oleic, linoleic, and linolenic). In this work, contents
of protein, oil and the fatty acids, palmitic (16:0), stearic (18:0), oleic acid (18:1Δ9c), linoleic acid
(18:2Δ9c,12c) and α-linolenic acid (18:3Δ9c,12c,15c) were determined in seeds of 235 Brazilian
soybean cultivars, which were analysed and classified as high, medium, low and very low content of
fatty acid. All these variables were standardized to an average equal to zero and variance equal to
one. For this study, it was use, 50 cultivars with high content of fatty acids. To evaluate the seven
responses variables, Principal Component Analysis (PCA) method was applied and a linear
combination that measures total variation of the variables, Cluster Analysis (CA), biplot graphic, and
communality, were considered. Based on a Singular Value Decomposition (SVD), biplot method
permits to present in the same graph, response variables and treatments, indicating the contribution
of both according to its components. The exploratory analysis do not indicated a normal distribution
for the oleic acid and linoleic acid. It was observed a strong negative correlation (-0.95) between oleic
and linoleic acids. A negative correlation (-0.48) between protein and oil contents were also observed.
Negative correlations among these compounds are common in soybean. The remaining correlations
ranged from 0.01 to 0.29. By cluster analysis with the method of k-means, Euclidean distance and
complete linkage, allowed the cut off level for the five linkage distance in the dendrogram (Fig. 1). The
highest variability was found for protein (SD = 1.15) and for linoleic acid (SD = 1.10). In the Principal
Component Analysis (PCA) was considered the first three eigenvalues greater than and equal to 1.0,
which totalize 71.53% of the explained variance (PCA1 = 29.06, PCA2 = 23.76 and PCA3 = 18.71%).
The highest average communality for these first three components were 0.99 for the oleic acid and
0.91 for the linoleic acid. It was observed that the response variables that most contributed to this
study were the contents of protein, oleic acid and linoleic acid. Cultivars that responded for protein
were 1 – MSBRS 169, 22- Campos Gerais , 30-IAC 1, 13-CEP 26, 26-FT Líder. For fatty acids the
cultivars were 1-MS BRS 169, 2-FT Sarayba, 3-FT Guaira, 4-Ocepar 6, 5-FT Cometa, 6- Ocepar 13
and 7-BR 38. The fatty acids with low contribution were: palmitic acid and linolenic acid (Fig. 2). By
using multivariate methods associated with the biplot, it was possible to indicate cultivars with higher
levels of protein, oil and fatty acids.
9
8
Complete Linkage
Euclidean distances
6
5
4
3
2
1
0
CAMPOS GERAIS
IAC 1
J 200
FT 10
SPS 1
IVORA
FT 9
DAVIS
FT 14
RS 5
BOSSIER
IAC 10
MSBR GUAV
BR 7
IAS 2
BR 4
FT JATOBA
BR 4
CD 203
BRS 158
IAC PL1
DM SOBERANA
BRAGG
CS 110
IAC 5
EMGOPA 309
FT 20
EMBRAPA 19
KIS 601
BR 2
BR 48
IAS 5
OCEPAR 14
OCEPAR 12
BRS 154
IAC FOSF 31
FT LIDER
UNIÃO
OCEPAR 2
CEP 26
BR 38
CD 202
OCEPAR 13
EMBRAPA 25
FT COMETA
MGBR 54
OCEPAR 6
FT GUAIRA
FT SARAYBA
MSBRS 169
Linkage Distance
7
G5
G4
G3
G2
G1
Fig. 1. Dendrogram based on the method of complete
linkage considering content of protein, oil and fatty acid of
50 soybean cultivars.
Fig. 2. Biplot considering different levels of oil,
protein, and fatty acids (PCA1 = 29.06, PCA2
=23.76 and PCA3= 18.71 %); on 50 soybean
cultivars.
International Biometric Conference, Florence, ITALY, 6 – 11 July 2014