Gene network inference using continuous time Bayesian networks: a comparative study and application to Th17 cell differentiation

Table 2 Performance comparison of CTBNs, DBNs and GC on simulated data for different time granularities

Method	Time course granularity	Mean precision	Mean recall	Mean F ₁	F ₁SEM
GC	11	0.43	0.61	0.50	2.88E-02
	21	0.40	0.70	0.49	3.35E-02
	31	0.35	0.75	0.47	3.80E-02
DBNs	11	0.84	0.15	0.26	4.33E-02
	21	0.55	0.42	0.47	3.66E-02
	31	0.68	0.30	0.40	2.79E-02
CTBNs	11	0.70	0.36	0.47	2.05E-02
	21	0.72	0.48	0.57	5.54E-02
	31	0.59	0.51	0.54	3.23E-02
Method	Time course granularity	Mean precision	Mean recall	Mean F ₁	F ₁ SEM
GC	11	0.47	0.76	0.57	4.05E-02
	21	0.28	0.81	0.41	5.78E-02
	31	0.29	0.80	0.42	3.56E-02
DBNs	11	0.76	0.21	0.32	2.32E-02
	21	0.60	0.57	0.58	4.31E-02
	31	0.63	0.40	0.48	3.86E-02
CTBNs	11	0.60	0.53	0.60	3.25E-02
	21	0.72	0.70	0.70	6.03E-02
	31	0.56	0.67	0.60	3.48E-02

Tests refer to 20NETs, organism E.coli (top) and S. cerevisiae (bottom). Aggregate F ₁, precision and recall values are calculated as the arithmetic mean over the sets of 10 sampled network instances, the standard error of the F ₁ mean (SEM) is also shown. See also Figure 4.

ISSN: 1471-2105