Table 2. Comparison of models extracted by the four evaluated contending methods: Mean values across contexts.

Global characteristics of the models are derived by applying RegrEx (with automated determination of λ), RegrEx-λ₀ (i.e., RegrEx without regularization), Lee2012, iMAT and FastCORE. The abbreviations stand for the following: $\stackrel{-}{Card.}$ denotes mean cardinality, $\stackrel{-}{O}$ _R, mean data-orphan ratio,$\stackrel{-}{\rho }$ _(V,D), mean correlation between data and predicted flux values, $\stackrel{-}{R}$ _(V,D), mean residual value between fluxes and data $, \stackrel{-}{I}$ _J, mean Jaccard index to any other context, Shared, number of shared reactions across all contexts, and Total Exclusive represents total number of exclusively context-specific reactions across all contexts. Values in round brackets correspond to the standard deviation.

	$\overline{Card.}$	$\overline{O}$ R	$\overline{\rho }$(V,D)	$\overline{R}$(V,D)	$\overline{I}$ J	Shared	Total Exclusive	Total Card
RegrEx	842,91(55.14)	0,34(0.05)	0,42(0.07)	0,29(0.08)	0,56(0.04)	299	332	1618
RegrExλ0	1030,30(76.32)	0,47(0.04)	0,38(0.08)	0,28(0.08)	0,65(0.03)	490	239	1711
Lee2012	784,6 (26.53)	0,28(0.03)	0,13 (0.05)	-	0,77 (0.01)	509	140	1092
FastCORE	1357,9(39.03)	0,50(0.04)	-	-	0,61(0.05)	503	230	2232
iMAT	1411(41.62)	0,42(0.04)	-0.17(0.03)	-	0,65(0.04)	611	210	2205