Figure 5. Model Architecture and construction of two-dimensional convolutional neural network model, SSeq CNN.

A two-dimensional convolutional neural network (2D-CNN) model was constructed and implemented with 5:1 class weight applied as a balancing strategy and hyperparameter optimization method and was compared to our SSeq1.0 MLP drug sensitivity model. The architecture of each of the two identically constructed 2D-CNN subnetworks is shown on the right. The final version of the 2D-CNN is more complex, balanced, and displays higher recall compared to the SSeq MLP model, as shown in Table S2.