Development of new computational methods for fragment-based drug discovery by NMR
thesisposted on 03.12.2020, 12:53 by Luca G. Mureddu
My project covers the three fundamental steps of fragment-based drug discovery by NMR, (NMR-FBDD): hit identification, binding site identification and hit optimisation. This division enabled me to cover the full NMR-FBDD approach while simultaneously to experience a broad spectrum of techniques as well as to foster collaborations with different labs and levels of expertise. I first focused on the development of a new software package, called CcpNmr AnalysisScreen, designed for the automated analysis of early stage 1D NMR screening data. AnalysisScreen integrates all the necessary algorithms and routines for supporting the analysis of these data. Furthermore, it provides a novel modular platform for combining tasks in bespoke workflows, and it includes a straightforward mechanism for adding new custom algorithms to the main program. Using a series of simulated spectral datasets with known answers, the performance of the software was assessed. Following this proof of principle, routines were tested using actual experimental data recorded by several collaborators. These analyses prompted the development of novel routines for scoring the outcomes to classify results accurately. Successively, I focused on the integration of tools for analysing and identifying ligand binding sites. These tools were tested using experimental data recorded by collaborators and successfully used in several postgraduate teaching classes, including national and international workshops. Lastly, I focused on the optimisation of ligand-binding properties of hits obtained from the initial virtual screening steps. During my internship at the Dana Farber Cancer Institute (Harvard Medical School, Boston, USA), I was involved in a study of the protein-protein complex eIF4G-Mnk. Starting from previous works, which identified transient binding pockets, I conducted a series of studies using a combination of computational tools to design a novel virtual FBDD workflow. This procedure prompted to the generation of potential drug candidates for the inhibition of the eIF4G-Mnk complex.