Table 1.

BIC and ROC AUC results for joint model fits to simulated data. Compare to Supplementary Table 2.

(a) Mean (and standard error) of BIC for layered and chained fits to data generated from layered and chained topologies. Higher (less negative) values are preferred. The “difference” row shows the difference between BIC s for correct vs. incorrect fit topologies.
	Layered gen.	Chained gen.
Layered fit	–145970.6 (21.9)	–145429.8 (22.6)
Chained fit	–145995.9 (21.9)	–145414.3 (22.6)
Difference	25.3 (0.8)	15.5 (2.0)

(b) Mean (and standard error) of ROC AUC for layered and chained fits, joint fit selected by BIC, and marginal fits to data generated from layered and chained topologies. See also the first row of Figure 3.
	Layered gen.	Chained gen.
Layered fit	0.881 (0.0007)	0.854 (0.0009)
Chained fit	0.838 (0.0009)	0.877 (0.0008)
Selected fit	0.881 (0.0007)	0.875 (0.0009)
Binding only	0.727 (0.0010)	0.820 (0.0009)
Expression only	0.737 (0.0011)	0.753 (0.0010)
Conservation only	0.715 (0.0010)	0.671 (0.0009)

(c) Proportions of correct choices for layered and chained fits to data generated from the corresponding topologies (BIC, ROC AUC) and of the fit selected by BIC being best by ROC AUC (“conditional”).
	BIC	ROC AUC	Conditional
Layered fit	0.997	1.000	0.917
Chained fit	0.910	0.993	0.989