Analysis of electric moments of RNA-binding proteins: implications for mechanism and prediction

BMC Structural Biology

Table 3 Neural network performance to discriminate between proteins binding to different types of RNA based on charge, dipole and quadrupole moments*.

Positive class binding to	Negative class binding to	Number of proteins in + ve class	Number of proteins in -ve class	AUC	F1	Precision	Recall	Accuracy
RNA	NB	160	2441	0.78	0.37	0.31	0.45	0.91
rRNA	NB	84	2441	0.79	0.26	0.23	0.30	0.94
tRNA	NB	20	2441	0.42	0.02	0.01	1.00	0.03
vRNA	NB	17	2441	0.75	0.24	0.24	0.24	0.99
mRNA	NB	13	2441	0.10	0.01	0.01	1.00	0.02
tRNA	rRNA	20	84	0.70	0.45	0.32	0.75	0.64
mRNA	rRNA	13	84	0.56	0.30	0.18	1.00	0.37
vRNA	rRNA	17	84	0.44	0.32	0.19	1.00	0.28
mRNA	tRNA	13	2441	0.07	0.57	0.39	1.00	0.39
mRNA	vRNA	13	2441	0.02	0.60	0.43	1.00	0.43
tRNA	vRNA	20	17	0.19	0.63	0.46	1.00	0.46
DNA	NB	143	2441	0.72	0.22	0.20	0.26	0.90
RNA	DNA	160	143	0.58	0.69	0.53	1.00	0.53
rRNA	DNA	84	143	0.74	0.64	0.52	0.83	0.65
tRNA	DNA	20	143	0.33	0.24	0.13	1.00	0.20
mRNA	DNA	13	143	0.07	0.16	0.09	1.00	0.14

* AUC is area under the ROC curve, F-measure (F1) is the highest geometric mean of precision and recall and accuracy is number of correct predictions relative to all predictions at peak F-measure. In all cases, neural network with three units in the hidden layer was used for training in a leave-one-out procedure and the training was performed for a fixed number of epochs without using information from left-out protein.

ISSN: 1472-6807