Table 2 Model hyperparameters

Overall model

Number of layers

14

Loss function

Binary cross-entropy

Learning rate

0.005

Optimizer

Adam

Training epochs

80

Callbacks

Validation loss

Batch size

32

Genotyping input layers

Architecture [Layer name (output shape)]

Input layer (5188)

Pathways layer (186)

Reshape layer (186, 1)

1D Convolutional layer (186, 12)

Flatten layer (2232)

Dense layer (150)

Gene expression input layers

Architecture [Layer name (output shape)]

Input layer (19403)

Dense layer (150)

Concatenation layer

Output nodes

300

Batch normalization layer

Momentum

0.99

Epsilon

0.001

Hidden layers

Number of layers

3

Architecture [Layer name (output shape)]

Dense layer 1 (180)

Dense layer 2 (30)

Dense layer 3 (15)

Output layer

Output nodes

2

Activation

Softmax