Search

Rational drug design often starts from specific scaffolds to which side chains/substituents are added or modified due to the large drug-like chemical space available to search for novel drug-like molecules. With the rapid growth of deep learning in drug discovery, a variety of effective approaches have been developed for de novo drug design. In previous work we proposed a method named DrugEx, which can be applied in polypharmacology based on multi-objective deep reinforcement learning. However, the previous version is trained under fixed objectives and does not allow users to input any prior information (i.e. a desired scaffold). In order to improve the general applicability, we updated DrugEx to design drug molecules based on scaffolds which consist of multiple fragments provided by users. Here, a Transformer model was employed to generate molecular structures. The Transformer is a multi-head self-attention deep learning model containing an encoder to receive scaffolds as input and a decoder to generate molecules as output. In order to deal with the graph representation of molecules a novel positional encoding for each atom and bond based on an adjacency matrix was proposed, extending the architecture of the Transformer. The graph Transformer model contains growing and connecting procedures for molecule generation starting from a given scaffold based on fragments. Moreover, the generator was trained under a reinforcement learning framework to increase the number of desired ligands. As a proof of concept, the method was applied to design ligands for the adenosine A2A receptor (A2AAR) and compared with SMILES-based methods. The results show that 100% of the generated molecules are valid and most of them had a high predicted affinity value towards A2AAR with given scaffolds.

Less than 1 in 10 drug candidates that enter phase 1 clinical trials actually gets approved for human use. The high failure rate is in part due to unforeseen side effects or toxicity. A better understanding of the role of selectivity and a better insight in the off-target activities of drug candidates could greatly aid in preventing candidates to fail for these reasons. This thesis has tried to address some aspects in this challenging part of drug discovery. The use of activity-based protein profiling as presented in Chapters 2 and 3 in drug discovery and hit-to-lead optimization, and in Chapter 5 and 6 for the interaction profiling of a drug candidate, highlights the versatility and importance of this chemical biology technique. Combined with knowledge derived from biochemical assays, such as that developed in Chapter 4, ABPP can greatly aid the medicinal chemist. The recent surge in popularity of machine learning algorithms, backed by exponential growth of the amount of biological data available, holds great promise for drug discovery. Chapters 7 and 8 showed the applicability of one such algorithm, which was able to quite reliably predict interaction profiles. The challenges in finding, determining and predicting selectivity are far from solved, but, by incrementally expanding our understanding of the binding of small molecules to their (off-)targets, truly selective inhibitors might at some point become a reality or their necessity might be mitigated.

The work described in this thesis focuses on the application of various NMR techniques to the study of interactions between proteind and small molecules and proteins and peptides, incluiding the well-established classical NMR approaches, and the precently developed paramagnetic NMR methods. computational tools have been used to complement and visualized the experimental data.