Bioprocess stability depends on the variety of yeast extract, which varies from lot-to-lots and between brands, thereby leading to variable bacterial growth and productivity in manufacturing processes. As a model experiment for good stability of bioprocesses, Escherichia coli growth in media containing different brands of yeast extract was evaluated and predicted using component-profiling and multivariate data analysis (metabolomics approach). The components of yeast extract were extracted from media containing varying concentrations of yeast extract and analyzed using gas chromatography-mass spectrometer. The yeast extract was categorized into three clades by principal component analysis (PCA). The E. coli growth using yeast extract showed approximately 30% difference at equivalent amount of supplementation. The bacterial growth in the media was estimated for the component profiles by partial least squares regression analysis (PLS-R). A predictive model was developed from the relationship between bacterial growth (as subjective attributes) and component profiles (as objective attributes), and correlation coefficients were calculated. Most of the amino acids in the media stimulated growth; however, methionine had negative effect on growth. In a culture validation, Asp, Val, Glu, and Try stimulated the bacterial growth, but Met inhibited. The other amino acids tested, Ser, Ile, Asp, Lys, Phe, Leu, Thr, and Gly did not show significant effects on the growth. The results indicate that the metabolomics approach can provide useful feedback information to improve the cultivation. (C) 2019, The Society for Biotechnology, Japan. All rights reserved.