| DS Description
|
A total of 513 metabolites, including two … A total of 513 metabolites, including two internal standards contained in the extraction solvent, were detected based on optimized SRM conditions and retention time of LC-QqQ-MS. PubChem ID (Kim et al. 2016) and KEGG ID (Kanehisa et al. 2014) were assigned to each metabolite as well as our internal ID (serial number) (for detailed information, see Supplementary Tables S2–S4).
The procedure for data pre-processing is summarized in Supplementary Fig. S1. Metabolomic data matrix of 513 metabolite intensities, which was obtained by LC-QqQ-MS analysis (Data matrix 1 in Supplementary Table S5), was comprised of 48 samples: four experimental groups (2 ploidy types × 2 sample preparation methods) × 12 replicates derived from the 12 individual plants (Supplementary Table S1). The metabolomic data were analyzed using the statistical software R (version 3.1.2, http://www.R-project.org/). After the missing values were set to 10, the signal intensities of each experimental group were averaged in individual metabolites. The metabolites with signal-to-noise ratio (defined as ratio of averaged signal intensity to those of extraction solvent control) <3 in all four experimental groups were removed. In addition, the metabolites with relative standard deviation greater than 50% in all experimental groups were removed, leaving 125 metabolites for further analysis to compare single-grain-based and weight-based analyses. Using these criteria for data pre-processing, more than 100 metabolites can be detected in A. thaliana seeds and leaves (Tsukaya et al. 2015). The intensities of the 125 metabolites were divided by those of the internal standards (Data matrix 2 in Supplementary Table S6). Following this, the data distribution of each sample was checked using boxplots (Supplementary Fig. S1). We found that the data trend of sample no. 14 differed from that of the others, and thus omitted the data of this sample as an outlier. The resulting data matrix (Data matrix 3 in Supplementary Table S7) was used for comparative analysis. In the case of single-grain-based analysis, metabolite intensities were further normalized by various measurements of seed size; namely, volume, lengths of major and minor axes, and 2-dimensional projected area. We performed multivariate analyses to evaluate the global trend in the metabolomic data. After transformation into log2, the data obtained from single-grain-based analysis (samples no. 1–13 and 15–24) and weight-based analysis (samples no. 25–48) were respectively transformed into z-score and used for boxplot, volcano plot, and Welch's t test using R. volcano plot, and Welch's t test using R.
|
