Fig. 1
From: DRaW: prediction of COVID-19 antivirals by deep learning—an objection on using matrix factorization
DRaW’s Framework. (1) Instead of applying to the virus-antiviral interactions, we use the model on the similarity data of antivirals and viruses. (2) Each sample of antivirals is concatenated with a virus. The results of the concatenation are the feature inputs to a deep network. (3) The deep model consists of four consecutive Conv1D layers with Relu activation function. Each of them is followed by batch normalization and dropout 0.5. Next, we use a dense layer after a flattened layer, followed by a dropout of 0.5. Finally, a dense layer with a sigmoid activation function acts as a binary classifier and predicts the interaction between the drug and protein. We compiled our model with Adam optimizer and binary cross entropy loss function. The prediction value is the association between the corresponding virus-antivirals. (4) Molecular docking study has been conducted on top-ranked drugs