Figure 6

PSEUDO predictions as a function of the near-optimal growth threshold. Values derived using PSEUDO (black) are compared to values derived from FBA (red) and MOMA (blue). The PSEUDO method accepts a growth threshold input parameter that has no analogy in the other methods. This threshold defines a near-optimal flux space, and mutant flux profiles are determined that minimize the distance to this space. Biologically, this growth threshold could be interpreted as a region of relaxed regulation, within which selection for increased growth is balanced by other metabolic demands or by noise. The growth threshold parameter was allowed to vary from 80% to 99% WT maximal growth. (A, B) Pearson and Spearman correlation coefficients as a function of the near-optimal growth threshold. PSEUDO predictions reached a maximum using a growth threshold of 90%, but were generally robust to parameter variation. Error bars represent one standard error of the mean, calculated with Fisher's z transformation. (C, D, E, F) Mean flux prediction errors from each of the three methods as a function of the growth threshold parameter. Errors were calculated for all 320 fluxes curated from the Tomita data set, and for subsets of reactions belonging to glycolysis, the PPP, or the TCA cycle. Flux errors were generally insensitive to the chosen threshold. TCA cycle fluxes were both the most error-prone, and the most improved by PSEUDO. Error bars represent one standard error of the mean.