Fragment based drug design plays an important role in the drug discovery process as a way to reduce the complex small molecule space into a more manageable fragment space. This thesis explores mathematical techniques and deep learning methods to explore computational ways of describing and encoding proteins and drug molecules, with the goal of extracting information to predict chemical binding. The initial chapters reveal and highlight the challenges of modelling protein-ligand interactions to identify the best computational tools to use. Firstly, the the viability of experimental bio-assay data and statistical machine learning tools are explored. Secondly, different clustering algorithms are studied for their ability to retain physicochemical information of molecule encoding and thirdl...[ Read more ]

Fragment based drug design plays an important role in the drug discovery process as a way to reduce the complex small molecule space into a more manageable fragment space. This thesis explores mathematical techniques and deep learning methods to explore computational ways of describing and encoding proteins and drug molecules, with the goal of extracting information to predict chemical binding. The initial chapters reveal and highlight the challenges of modelling protein-ligand interactions to identify the best computational tools to use. Firstly, the the viability of experimental bio-assay data and statistical machine learning tools are explored. Secondly, different clustering algorithms are studied for their ability to retain physicochemical information of molecule encoding and thirdly, a preliminary off-the-shelf deep learning framework is proposed to correlate proteins and inhibitor fragments and the emerging problems are studied. The main project leverages the availability of a custom built deep learning architecture to design ChemPLAN-Net - a model that incorporates both the protein drug target and inhibitor information and learns from the thousands of protein co-crystal structures in the PDB database. Its purpose is to reliably suggest a number of inhibitor fragments for a novel query protein structure and offer corresponding binding modes for future facilitated drug design. The model is validated thoroughly from a statistical, chemical and experimental literature perspective and its applicability is demonstrated on the kinase and protease protein families.