Abstract
Our objective of this study is to elucidate a whole response of plants to sulfur starvation by integrating metabolomics with transcriptomics. To elucidate gene-to-metabolite and gene-to-gene networks, the time-course data of metabolome and transcriptome of sulfur-starved Arabidopsis were combined into a single data set and analyzed by batch-learning-self-organizing mapping (BL-SOM), a sophisticated form of cluster analyses. BL-SOM can cluster the metabolites and genes showing similar accumulation and expression patterns. By this analysis, a known gene-to-metabolite network, that is, correlation between the sulfur assimilation genes and O-acetylserine, a positive regulator of them, was shown. Moreover, an unknown gene-to-gene network, that is, correlation of the glucosinolate-biosynthesis genes was shown. Based on this finding, three previously unknown sulfotransferase genes involved in glucosinolate biosynthesis were identified. This strategy is useful for functional identification of unknown genes, especially when the genes of interest are involved in metabolism and the knock-out mutants don't show apparent phenotypes.