Page 258 - PC2019 Program & Proceedings

PLANT CANADA 2019

P89. Genetic diversity of grain fatty acid composition in 295 accessions of Korean Rice Core Set
3
2
1
1
1
1
2
2
Yang, J. ; B. Ha ; S. Noh ; S Eom ; S. Chu ; K. Kim ; Y. Park ; Lee, Y.
1 Soochunhyang University
2
Kongju National University
3
Soonchunhyang University

Fatty acid is an important phytonutrient as a component of lipids in rice grains. To understand genetic
diversity in fatty acid compositions, 295 accessions of Korean Rice Core Set developed from 25,604
germplasm were cultivated in 3 separate locations in Korea, and the composition of fatty acids in
harvested brown rice were evaluated according to one-step methylation/extraction method coupled with a
GC-FID. Among 9 identified, linoleic, oleic, and palmitic acids were the 3 major fatty acids consisting
36.6%, 35.0%, and 23.4% of total fatty acids, respectively. Average compositions of other fatty acids
such as stearic, linolenic, myristic, arachidic, eicosenoic, and behenic acids were 1.8%, 1.1%, 0.9%,
0.5%, 0.3%, and 0.3%, respectively. Throughout all tested accessions, oleic acid showed correlations
negative with palmitic (r=-0.675) and linoleic (r=-0.657) acids, but positive with eicosenoic acid
(r=0.639). Ecotype of rice also affected fatty acid compositions in that Aus-type rice accessions showed
higher saturated fatty acids (31.8%), while japonica-type accessions exhibited higher unsaturated fatty
acid compositions. All these results suggested wide genetic variations of fatty acid composition in tested
Korean Rice Core Set accessions, which could be utilized in a breeding programs to develop a new rice
variety of higher nutritional value.

Young-Sang Lee (mariolee@sch.ac.kr)

P90. Regression data driven models on canopy hyperspectral reflectance for soybean yield
prediction
*
Yoosefzadeh Najafabadi, M. ; M. Eskandari
University of Guelph

Direct measurements of important agronomic traits, including morpho-physiological yield-related traits,
are prone to human error, resource- and labour-intensive. Current advances in high-throughput
phenotyping such as hyperspectral reflectance have provided breeders with new opportunities to measure
these traits, indirectly, and screen large number of genotypes at early-generation. Although the
interpretation of individual hyperspectral reflectance is challenging, it can be facilitated through the
development of robust regression-data-driven models. Using 250 soybean lines grown in two
environments in southwestern Ontario in 2018, we have collected data for yield and 188 discrete
reflectance wavebands, ranged from 391 to 1010 nm, at three reproductive growth stages (R3, R4, and
R5). These data were used to create 24 regression models based on ordinal least square regression
(OLSR) and principal components with partial least square regression (PLSR). Furthermore, nine
conventional vegetation indices (VIs) were constructed for predicting the soybean yield. The best model
for yield prediction was the PLSR of 10 nm average binning wavelengths without normalization at R5
(PLSR-BWWN5) that explained up to 51% of the yield. In PLSR-BWWN5 model, 93% of the predicted
yield was explained by only 25 binning wavebands with a heritability of 0.4. This information seems to
be useful for soybean yield prediction at the early stage (i.e., R5), which in turn can facilitate the
development of soybean cultivars with increased seed yield and the genetic gain.

Mohsen Yoosefzadeh Najafabadi (myoosefz@uoguelph.ca)

Page 256 of 339