“…For the remainder of this work (unless otherwise noted) we use the term “growth condition” to refer to the four-dimensional vector of categorical variables defining: i) growth phase (exponential, stationary, late stationary), ii) carbon source (glucose, glycerol, gluconate, lactate), iii) Mg 2+ concentration (low, base, high), and iv) Na + concentration (base, high). While we note that growth phase is not strictly an environmental feature, we suspected that this indicator of cellular state would be an important feature to consider since prior research has shown that the macromolecular composition of cells varies substantially between exponentially growing and stationary phase cells [35,36]. With these data and features, the question we set out to answer is: to what extent are machine learning models capable of discriminating between the known growth parameters given only knowledge of gene expression levels?…”