Skip to main content

Table 5 The comparison between the performance of DrugRep-HeSiaGraph and DrugRep-KG based on four versions of DDKGs

From: DrugRep-HeSiaGraph: when heterogenous siamese neural network meets knowledge graphs for drug repurposing

Model

DDKG version

ACC (%)

AUC-ROC (%)

AUC-PR (%)

BS (%)

MCC (%)

F1-score (%)

DrugRep-HeSiaGraph

DDKG-V1

\(84.63 \pm 0.02\)

\(91.16 \pm 0.04\)

\(90.32 \pm 0.13\)

\(11.90 \pm 0.04\)

\(69.31 \pm 0.82\)

\(83.13 \pm 0.12\)

DDKG-V2

\(83.44 \pm 0.10\)

\(90.63 \pm 0.07\)

\(90.09 \pm 0.19\)

\(12.01 \pm 0.05\)

\(67.53 \pm 0.13\)

\(80.54 \pm 0.27\)

DDKG-V3

\(84.51 \pm 0.04\)

\(90.97 \pm 0.03\)

\(90.27 \pm 0.41\)

\(12.00 \pm 0.04\)

\(69.12 \pm 0.14\)

\(82.84 \pm 0.08\)

DDKG-V4

\(83.91 \pm 0.08\)

\(91.15 \pm 0.10\)

\(90.17 \pm 0.27\)

\(12.06 \pm 0.03\)

\(67.90 \pm 0.35\)

\(82.08 \pm 0.19\)

DrugRep-KG

DDKG-V1

\(83.9\) 3

\(91.00\)

\(90.30\)

\(11.97\)

\(67.90\)

\(82.83\)

DDKG-V2

\(83.92\)

\(90.57\)

\(89.94\)

\(12.18\)

\(67.50\)

\(83.04\)

DDKG-V3

\(82.45\)

\(91.03\)

\(90.18\)

\(12.15\)

\(64.99\)

\(82.73\)

DDKG-V4

\(82.45\)

\(90.94\)

\(89.97\)

\(12.24\)

\(64.92\)

\(81.89\)

DisDrugPred

–

–

58.73

53.45

38.91

–

–

DRP-VEM

–

52.33

55.37

55.12

29.76

32.43

52.35