DeepPN: a deep parallel neural network based on convolutional neural network and graph convolutional network for predicting RNA-protein binding sites

Table 3 The AUC results for each method

RBP	DeepPN	GraphProt	Deepnet-rbp	iDeepV
C17ORF85 PAR-CLIP	0.837	0.800	0.820	0.740
CAPRIN1 PAR-CLIP	0.886	0.855	0.834	0.824
C22ORF28 PAR-CLIP	0.785	0.751	0.792	0.823
ALKBH5 PAR-CLIP	0.660	0.680	0.714	0.643
ELAVL1 HITS-CLIP	0.978	0.955	0.966	0.966
HNRNPC iCLIP	0.977	0.952	0.962	0.979
SFRS1 HITS-CLIP	0.936	0.898	0.931	0.905
AGO2 HITS-CLIP	0.868	0.765	0.809	0.886
TDP43 iCLIP	0.936	0.874	0.876	0.935
AGO1-4 PAR-CLIP	0.912	0.895	0.881	0.925
TIAL1 iCLIP	0.926	0.833	0.870	0.929
TIA1 iCLIP	0.928	0.861	0.891	0.941
EWSR1 PAR-CLIP	0.954	0.935	0.966	0.962
ELAVL1 PAR-CLIP(A)	0.967	0.959	0.966	0.973
ELAVL1 PAR-CLIP(B)	0.976	0.935	0.961	0.962
FUS PAR-CLIP	0.977	0.968	0.980	0.976
PUM2 PAR-CLIP	0.952	0.954	0.971	0.965
IGF2BP1-3 PAR-CLIP	0.928	0.889	0.879	0.923
MOV10 PAR-CLIP	0.904	0.863	0.854	0.896
ELAVL1 PAR-CLIP(C)	0.994	0.991	0.994	0.990
ZC3H7B PAR-CLIP	0.898	0.820	0.796	0.883
PTB HITS-CLIP	0.938	0.937	0.983	0.936
TAF15 PAR-CLIP	0.974	0.970	0.983	0.978
QKI PAR-CLIP	0.975	0.957	0.983	0.965
Average	0.919	0.887	0.903	0.913

The best performance is marked in bold
The AUC results for GraphProt, Deepnet-RBP and iDeepV are taken from original papers

ISSN: 1471-2105