“…Computational de novo drug design involves the use of techniques such as genetic algorithms 3,39,47,51,54,73 , reinforcement learn-ing, including deep reinforcement learning 57,61,63,74,86 , generative deep learning models 5,6,9,35,41,43,55 , or other deep learning methods, e.g., graph transformers 46,71 , models that blend deep learning and evolutionary algorithms 1,26,54 , and string-based trans-formers (i.e. operating on a SMILES string representation of molecules) 27,33 .…”