Our fourth principal component, plotted as metallicity (12+log(O/H)) vs. the optimum linear combination of stellar mass, and molecular gas mass. We did not include the contribution from SFR because it is essentially zero: the optimum projection of the data requires only the stellar mass and molecular gas mass. The quadratic fit, defined in Eq. (1), is overplotted as a dashed line. Galaxies with only upper limits on their H₂ mass are not shown (to avoid crowding the plot), but are consistent with the derived relation.