Statistical study on the environmental effects on the natural variation of nutritional components in rice varieties

Abstract This study was investigated to compare the natural variation of nutrients in rice variety by different environmental factors. Fifteen kinds of rices were used, which were cultivated in two locations for 2 years. All data were analyzed by the various statistical tools to identify the nutritional variations of nutrients. The results of variable importance in the prediction analysis were found to be consistent with the % variability. The nutrient compositions most affected by variety were fatty acids, and next were vitamins, proximate nutrients, minerals, and amino acids in order. The nutrient compositions most affected by location were proximate, followed by minerals, vitamins, fatty acids, and amino acids. For cultivation year, vitamins were most affected and then minerals, fatty acids, proximate nutrients, and amino acids in order. These findings could explain that each kind of nutrients can be naturally varied by different environmental factors.


| INTRODUC TI ON
Informative big data of plants nutritional compositions are very important for the nutritional safety assessment of new biotechnology organisms. Also, these are useful to the food industry and food-based dietary guidelines for different sources of food nutrients. Furthermore, the nutritional compositions of commercial plants can play an important role by providing information on natural chemical components as well as the presence and amounts of interacting components. Specifically, for the safety assessment of new biotechnology organisms such as genetically modified organisms (GMOs), the level of nutritional compositions of commercial crops is considered as references for the identification of significant compositional changes in GMOs (EFSA, 2010). Because of the natural compositional change in commercial crops, it is necessary to study whether the component change due to other biotechnology belongs to this natural variation category or not. Natural variation of the components in crops has been reported in several articles (Nicole & Daniel, 2015;Thomas, Richard, Howard, Eugenia, & Karl, 2012). Statistical methods for compositional assessment can be used to identify natural variability differences observed from plant varieties or other variables. The natural variation of plants nutritional composition can be influenced by a number of environmental factors such as cultivation location and plant varieties, many of which can be barely controlled and induced natural variations. Also the micronutrients content and bioactive substances can be mainly affected by the seasonal variations in plant foods (Collomb et al., 2008). For this reason, the tolerant or confidence interval range of nutritional compositions by statistical tools can be used to improve credibility from obtained experimental data. Tolerance intervals are the most appropriate for evaluating whether or not a transgenic variety is within the normal range for commercial varieties of the same crop. Until now, in order to study the environmental effects of crop components, many researches have been conducted using suitable statistical models. As it was deemed reasonable to interpret integrated statistical models for natural variations of the components by environment, this study was intended to produce comprehensive results by various statistical techniques. All data were suggested by the mean value and standard deviation with the tolerance interval.
The natural variation of rice nutrition by using various statistical methods was also investigated. The effects of genotype and environmental variables on the compositional variations were identified through statistics. The environmental factors used were cultivation location, kinds of cultivar, and cultivation year. Rice grain samples (brown rice) were analyzed for their key nutritional components as recommended by OECD consensus document (OECD, 2004). All analyzed data were loaded to the database of web service, "DB for nutritional components of plant foods," to be used to the GMO safety. All data were compared statistically by using mean and standard deviation.
This study provided information on the natural variation in rice nutrient composition according to various environmental factors, and the results may offer reasonable and scientific assessment in nutritional profiles for future genetic modified of rice or other plants.  Table 1 Note. a pH value before planting. b Mean value of daily temperature. c Total volume of rainfall during 1 month. Kim, 2015). Whole grain (rough rice) samples were collected from each block and dried at a final moisture concentration of 9%-11% for compositional analysis based on the previous study conducted. The rough rice varieties were manually dehulled using a hulling machine (TR 13) to produce brown rice grain and then finely ground using a planetary mono mill (Pulverisette 6; Fritsch, Germany). The powdered grain of each sample was transferred immediately to −80°C until analysis.

| Statistical analysis
The statistical analysis used was SAS 9.

| Nutritional compositions of 15 rice varieties
To investigate the natural variations in the nutritional compositions of 15 commercial rice varieties, the rice varieties were cultivated in the test farms in Cheonan and Jeonju. The nutritional content of each rice variety was analyzed, and the results were presented as mean value and standard deviation (Supporting Information Table   S1). The units of each component were adjusted to be the same as those presented in the OECD document (OECD, 2004). The p-value for the significant differences between varieties, two locations, and two cultivation years was identified by ANOVA analysis of SAS package while the tolerance interval ranges were determined using proc capability of the SAS 9.2 software package (Supporting Information   Table S1). To identify differences among rice varieties, the significance probability associated with F statistic, labeled "Pr > F," was calculated by one-way ANOVA procedure of SAS package. To identify the difference of components between two locations and between 2 years, p-value labeled "Pr>|t|," which is calculated by paired t test procedure using SAS package, was used to examine each environmental factor separately (Supporting Information Table S1). Almost all components showed significant differences among varieties except IDF, proline, S, Cu, Zn, and B2. In terms of locations, 34 out of total 49 components showed significant differences. The components with no differences between two locations were crude fiber,  Table S1).  were affected by the year factor. In the % variability of vitamins, the natural variation was mainly contributed by the year (28.09%), next by the variety (11.70%), and location (11.40%) (Figure 1). Vit B1 was most highly affected by the variety (44.1%) while B7 was affected by the location (25.90%) (Supporting Information Table S2).

| % variability of rice grains nutrients
By including the interaction of varieties, location, and year in mixed model, the random effect was also assessed to establish equivalence limiting for the average differences over sites, cultivation year, or cultivation year. As a result, mixed variables of location × variety × year (L×V×Y) contributed to the variation of compositions more than the individual variables ( Figure 1). The variation for amino acids was highest with 83.97% by L × V × Y, and the variation for other components affected by the L × V × Y was as follows: vitamins (48.81%), proximates (48.05%), minerals (42.44%), and fatty acids (8.29%).

| VIP plot by PLS-DA
To identify the compositions which are important to the environmental factors, the VIP plot which was featured by PLS-DA in a descending order of importance was examined (Figure 2). VIP score refers to the weighted sum of squares of the PLS loadings (Mireia, Stefan, Stefan, & Romà, 2015). VIP analysis was conducted according to environmental factors such as variety, location, and year. As the value of VIP score of components is higher, it can be more im-

| Profiles for nutrition composition of 15 rice varieties
The mean values of all nutritional components analyzed in this study were contained within the range of brown rice contents as food presented in the OECD consensus documents (OECD 2004). The nutritional contents of plant foods may vary according to variety and growing conditions. In this study, since rice crops were grown in the same soil and in the same location each year, it was found out that most nutrients were significantly different between varieties more than between locations or between years. In the proximate and mineral results, it was found that the number of significant difference was higher between varieties and between locations than between years. The same result applied for amino acids and fatty acids, which there were more significantly difference between varieties than between year and between location ( In this study, ANOVA analysis was applied to complement the PCA results and determine whether they differ or not according to environmental factors.
In conclusion, the natural variations of nutrients caused by the

ACK N OWLED G M ENTS
This study was supported by the National Academy of Agricultural Science (Code PJ011752012018), Rural Development Administration, Republic of Korea. We are also thankful to the reviewers whose comments led to substantial improvements in this paper and to the Executive Editor for his consideration.