Supplementary Materials A Protein Targets Chosen for Generation

Figure A.1 shows the amino acid sequences corresponding to the three SARS-CoV -2 targets. We used a bidirectional Gated Recurrent Unit (GRU) with a linear output layer as an encoder. Figure B.1: The novelty of the scaffold of each generated molecule compared to the most similar scaffold in the training set. Similarity of the fingerprints, is shown next to the scaffold of each generated molecule. We show a representative set of molecules generated for each target in Figure D.1 Figure D.1: Representative molecules generated for (top to bottom): NSP9 Replicase, Receptor-Binding Domain (RBD) of S protein, and Main Protease of SARS-CoV -2 RBD has maximum subgraph similarity to a commercially available drug Telavancin (See Figure E.3).