Despite substantial increases in R&D spending within the pharmaceutical industry, denovo drug design has become a time-consuming endeavour. High attrition rates led to a
long period of stagnation in drug approvals. Due to the extreme costs associated with
introducing a drug to the market, locating and understanding the reasons for clinical failure
is key to future productivity. As part of this PhD, three main contributions were made in
this respect. First, the web platform, LigNFam enables users to interactively explore
similarity relationships between ‘drug like’ molecules and the proteins they bind. Secondly,
two deep-learning-based binding site comparison tools were developed, competing with
the state-of-the-art over benchmark datasets. The models have the ability to predict offtarget interactions and potential candidates for target-based drug repurposing. Finally, the
open-source ScaffoldGraph software was presented for the analysis of hierarchical scaffold
relationships and has already been used in multiple projects, including integration into a
virtual screening pipeline to increase the tractability of ultra-large screening experiments.
Together, and with existing tools, the contributions made will aid in the understanding of
drug-protein relationships, particularly in the fields of off-target prediction and drug
repurposing, helping to design better drugs faster.