Machine learning-assisted discovery of growth decision elements by relating bacterial population dynamics to environmental diversity

Abstract

Microorganisms growing in their habitat constitute a complex system. How the individual constituents of the environment contribute to microbial growth remains largely unknown. The present study focused on the contribution of environmental constituents to population dynamics via a high-throughput assay and data-driven analysis of a wild-type Escherichia coli strain. A large dataset constituting a total of 12,828 bacterial growth curves with 966 medium combinations, which were composed of 44 pure chemical compounds, was acquired. Machine learning analysis of the big data relating the growth parameters to the medium combinations revealed that the decision-making components for bacterial growth were distinct among various growth phases, e.g., glucose, sulfate and serine for maximum growth, growth rate and growth delay, respectively. Further analyses and simulations indicated that branched-chain amino acids functioned as global coordinators for population dynamics, as well as, a survival strategy of risk diversification to prevent the bacterial population from undergoing extinction.

Contribution of the genomic and nutritional differentiation to the spatial distribution of bacterial colonies

“…The enrichment of the single resource of glucose and the overall nutritional richness (LB) in the media resulted in the reverse directional changes in the colony size. This result strongly suggested that the balance of the medium composition largely participated in colony growth, as similar as to population growth in liquid media (Ashino et al, 2019;Aida et al, 2022). In addition, the coefficient of variation of colony size (V k ), kurtosis (U k ), and skewness (Z k ) of the size distribution of colonies all remained roughly comparable (p > 0.05), independent of either genome reduction or nutritional variation (Figure 3B).…”

Employing active learning in the optimization of culture medium for mammalian cells

“…The remaining 28 components were changed from zero to 10~100 folds of their concentrations in EMEM. In brief, four to five different concentrations were used for each component, and the changes were roughly on a logarithmic scale, as described previously 30 . The medium combinations were prepared by mixing the stock solutions of individual components, prepared at high concentrations in advance, as described previously 30 .…”

“…In brief, four to five different concentrations were used for each component, and the changes were roughly on a logarithmic scale, as described previously 30 . The medium combinations were prepared by mixing the stock solutions of individual components, prepared at high concentrations in advance, as described previously 30 . The stock solutions were sterilized by the sterile syringe filters (Merck) with hydrophilic PVDF membranes of 0.22 µm pore size and were dispensed in a small volume and stored at -20°C for future use.…”

“…Absorbance reads of the chemical assay were exported from the plate reader and processed with Python, as described previously 30 . The mean A450 of biological replications was calculated using "mean" in the "numpy" library.…”

“…The algorithm of gradient-boosting decision tree (GBDT) was used in machine learning (ML), which was performed with Python, as described previously 30 . The medium components and A450 of cell culture were employed as the explanatory and objective variables, respectively.…”