Year: 2023
Type: dissertation
Abstract: Artificial intelligence can help optimize molecules and discover new molecular structures. This work focuses on adapting state-of-the-art AI methods for various molecular representations to generate drug-like molecules. Emphasis is placed on evolutionary algorithms and transformers. Evolutionary algorithms are shown to be effective in optimizing molecules for multiple target properties. Transformers as generative neural models, on the other hand, leverage known molecular databases and enable the generation of novel molecules. This work demonstrates new ways of using transformers for spatial and graph representations. Finally, combination of transformers and evolutionary algorithms opens new possibilities for molecule generation.
Author: Cofala, Tim
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W6987300331