• Artificial Intelligence
• Carbonic anhydrase
• CDK5
• Chemical database
• GSK3b
• hDHODH
• Machine Learning
• MolBook UNIPI
• Selectivity
• Toxicity
• Virtual Screening

The PhD thesis is mainly focused on the application of artificial intelligence methods, especially machine learning (ML), in the context of drug discovery. In particular, the aim of the research is the development of new ML models able to predict biochemical properties of small molecules and the creation of software for the management of chemical databases. In the second chapters, ML models are developed and used to perform virtual screening (VS) studies to identify kinase inhibitors for the human targets Cdk5 and Gsk3β. In the third chapter, the selectivity of hCA inhibitors is analyzed using eXplainable Artificial Intelligence (XAI) techniques. The analysis led to the identification of structural features responsible for the selectivity, which were verified by the analysis of the X-ray complexes of the inhibitors. The fourth chapter focuses on the development of ML models to predict the toxicity of molecules. The studies led to the development of the VenomPred platform, which allows the prediction of eight toxicity endpoints. The fifth chapter describes the development of the MolBook UNIPI software. The software has been designed to facilitate the management of chemical databases and it is aimed at biochemists and synthetic chemists who need free, user-friendly tools to manage their data. The last chapter of the thesis reports on the development of a VS protocol that combines pharmacophore-based filter with docking followed by molecular dynamic simulations, which allowed the identification of new hDHODH inhibitors.