LogicNet: probabilistic continuous logics in reconstructing gene regulatory networks

BMC Bioinformatics

Table 2 The LogicNet in comparison with PCA-CMI, ARACNe, Genie3, Narromi, CN, and GRNTE in reconstructing the undirected/directed E. coli network, using 10 gene expression samples and 100 repeats of the whole simulation study. Two types of logics, i.e., PC and fuzzy logics, are used separately for reconstructing the GRNs and logic functions in the LogicNet algorithm. Also, the value of c = α + β is set to 1000. The highest accuracies are indicated in boldface. Reported values for the TP, FP, TN, FN are the total of the corresponding values over 100 repeats of the whole simulation study

Method	TP	FP	TN	FN	TPR	FPR	PPV	ACC	MCC	F-measure
Undirected E. coli Network (the edge direction is not taken into account in calculating the performance)
PC-LogicNet	724	157	2843	776	0.48	0.05	0.82	0.79	0.51	0.61
Fuzzy-LogicNet	640	155	2845	860	0.43	0.05	0.81	0.77	0.46	0.56
PCA-CMI-0.1	824	1974	1026	676	0.55	0.66	0.29	0.41	−0.11	0.38
PCA-CMI-0.05	940	2214	786	560	0.63	0.74	0.30	0.38	−0.11	0.40
ARACNe	160	140	2860	1340	0.11	0.05	0.53	0.67	0.11	0.18
GENIE3-FR-sqrt	213	228	2772	1287	0.14	0.08	0.48	0.66	0.10	0.22
GENIE3-FR-all	192	235	2765	1308	0.13	0.08	0.45	0.66	0.08	0.20
Narromi	490	829	2171	1010	0.33	0.28	0.37	0.59	0.05	0.35
CN	976	2297	703	524	0.65	0.77	0.30	0.37	−0.12	0.41
GRNTE	420	750	2241	1089	0.16	0.41	0.36	0.34	− 0.28	0.22
Directed E. coli Network
PC-LogicNet	624	588	6912	876	0.42	0.08	0.51	0.84	0.37	0.46
Fuzzy-LogicNet	540	405	7095	960	0.36	0.05	0.57	0.85	0.37	0.44
ARACNe	120	180	7320	1380	0.08	0.02	0.40	0.83	0.12	0.13
GENIE3-FR-sqrt	155	445	7055	1345	0.10	0.06	0.26	0.80	0.07	0.15
GENIE3-FR-all	156	444	7056	1344	0.10	0.06	0.26	0.80	0.07	0.15
Narromi	275	1513	5987	1225	0.18	0.20	0.15	0.70	−0.02	0.17
CN	616	1369	6131	884	0.41	0.18	0.31	0.75	0.21	0.35
GRNTE	232	1210	3030	1128	0.04	0.59	0.13	0.15	−0.63	0.08

ISSN: 1471-2105