Identifying time-delayed gene regulatory networks via an evolvable hierarchical recurrent neural network

Table 2 The effect of network size on GRN reconstruction in case of linearity between the genes. Results are the average of the accuracy in terms of the Link criteria for GRN reconstruction of 10 different randomly generated synthetic networks

Methods	P	TP	FP	FN	Precision	Recall	F-score	Nominal p-value	Adjusted p-value
HRNN	5	5.6	4.5	4.3	0.56	0.57	0.56
	10	9.7	15.2	9.6	0.39	0.50	0.44
	20	14.8	36.5	25.3	0.28	0.37	0.32
	30	21.8	58.1	36.5	0.27	0.37	0.31
TD-ARACNE	5	4.3	7.2	5.6	0.33	0.43	0.36	0.008	0.03
	10	7.8	27.2	11.5	0.22	0.39	0.26	0.01	0.04
	20	9.1	76.9	31.0	0.10	0.23	0.13	1.6×10⁻⁶	6.4×10⁻⁶
	30	11.3	153.7	47.0	0.07	0.19	0.09	5.0×10⁻⁸	2.0×10⁻⁷
HCC-CLINDE	5	4.2	2.1	5.7	0.70	0.44	0.53	0.65	1
	10	8.4	5.5	10.9	0.62	0.43	0.51	0.35	1
	20	11.9	18.5	28.2	0.40	0.29	0.34	0.77	1
	30	17.7	19.8	40.6	0.48	0.30	0.37	0.44	1
TSNI	5	9.9	15.1	0.0	0.39	1.0	0.56	0.99	1
	10	17.9	73.1	1.4	0.19	0.93	0.32	0.04	0.16
	20	19.5	188.5	20.6	0.09	0.48	0.15	1.1×10⁻⁶	4.4×10⁻⁶
	30	14.3	189.7	44.0	0.06	0.24	0.10	1.2×10⁻⁷	4.8×10⁻⁷
ebdbNet	5	7.8	11.2	2.1	0.40	0.77	0.53	0.47	1
	10	15.8	63.2	3.5	0.19	0.80	0.31	0.03	0.12
	20	27.4	234.6	12.7	0.10	0.66	0.17	7.6×10⁻⁶	2.8×10⁻⁵
	30	31.5	430.5	26.8	0.06	0.53	0.12	9.7×10⁻⁸	3.6×10⁻⁷

The p-values are for how the Link F-scores of other methods compare with HRNN. P is the number of genes for each of the networks. The variance of the noise is equal to 1

ISSN: 1756-0381