In recent years, there has been a surge of interest in the potential of artificial intelligence (AI) and machine learning (ML) tools, particularly deep learning models, to significantly reduce the time and cost associated with traditional drug discovery methods. By leveraging vast databases of chemical and biological information, these tools hold the promise of expediting the discovery of compounds with desired pharmacological profiles. To further enhance the speed and efficiency of existing drug discovery efforts for central nervous system disorders, we are exploring generative deep learning architectures for the de novo design of customized small molecules that employ transfer learning for pre-training and hybrid ligand-based and structure-based schemes for reinforcement learning. Additionally, we are developing transfer learning-enabled strategies for efficient in silico screening of ultra-large chemical libraries to identify ligands with desired pharmacological profiles using dense or graph convolutional neural network models.
Representative Publications
Salas-Estrada L, Provasi D, Qui X, Kaniskan HÜ, Huang XP, DiBerto JF, Ribeiro JML, Jin J, Roth BL, Filizola M. De Novo Design of κ-Opioid Receptor Antagonists Using a Generative Deep Learning Framework BiorXiv 2023 Apr 26;2023.04.25.537995. doi: 10.1101/2023.04.25.537995. [PMID: 37162828]; Journal of Chemical Information and Modeling (2023) accepted.