Table 2 Hyperparameter setting of NAGTLDA

NAFS

Threshold of lncRNA network ɑ

[0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]

Threshold of disease network β

[0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]

Maximum smoothing steps k

[2, 3, 4, 5, 6, 7, 8, 9]

List of r value

[{0,0.1,0.2,0.3,0.4,0.5}, {0.3,0.4,0.5}]

SDNE

First-order loss parameter alpha

1e-6

Coding dimension nhid1

[32, 64, 128, 256]

Regularization term parameters nu1

1e-5

Regularization term parameters nu2

1e-4

NAGTLDA

Learning rate lr

0.001

Random seed

50

Dropout

0.4

Adam optimizer weight-decay

5e-3

Number of layers of global-level embedding L1

[1, 5, 10, 15]

Number of layers of global-level embedding fusion L2

[15, 10, 20, 25]

Number of heads of global-level embedding H1

[4, 8, 16, 32]

Number of heads of global-level embedding fusion H2

[16, 32, 64, 128]

Feature embedding size out-dim

[32, 64, 128, 256, 512]

Epoch

150