Application of Virtual Screening Tools to Aid in the Discovery of Novel PI3Kβ Inhibitors

Abstract

The phosphoinositide 3-kinase (PI3K) are a family of lipid kinases involved in many intracellular processes that influence cell growth, survival, autophagy, and intracellular trafficking. There is growing evidence for a role of the class Ia PI3K, PI3Kβ, in phosphatase and tensin homolog deficient cancers. PI3Kβ also has a non-catalytic function that occurs through a protein – protein interaction with Rab5. The recent elucidation of residues which form the Rab5 – PI3Kβ interface provides opportunities to identify small molecule inhibitors of this interaction. Chapter 3 used the available crystal structures of PI3Kβ to perform a retrospective study to characterise the ability of the protein structure and the docking algorithm of GOLD to distinguish active compounds from a library of inactive and decoy compounds. A prospective screen of a large commercially available lead-like library was then conducted to identify novel compounds that potentially modulate the ATP binding site which identified a diverse set of compounds predicted to form favourable interactions with the ATP binding site. However they were unable to inhibit catalytic activity in vitro. An analysis of these compounds identified factors such as pocket occupancy and less favourable geometry of protein-ligand interactions may contribute to their failure to work as enzyme inhibitors. A scalable and intuitive workflow was created that can generate 3D structures from 2D structure representations was created to assist in the generation of docking-ready structures. This process also identified limitations of the application of open source software to generate 3D structures and was resolved with the addition of the Ligand Preparation component from the CCDC. This multi-step workflow can enumerate isomers and generate robust 3D structures comparable to other available programs and was applied in the generation of structures for Chapter 5. Chapter 5 used a combination of covalent docking methods and in silico mapping of an identified pocket near the identified residues to characterised the PI3Kβ – Rab5 interaction interface. This information was used prospectively to screen a small library of sulfonyl fluoride containing compounds, and identified a class of purine sulfonyl fluorides that form favourable interactions in this pocket. The sulfonyl fluoride library was also screened in the ATP binding site to identify compounds which could potentially form interactions in this site when tethered to the catalytic lysine (LYS799). Furthermore, probing of this site with an in-house sulfonyl fluoride compound showed it was unable to label the hypothesised target residue, but did identify two new residues, TYR833 and LYS923 that can form covalent adducts with SN39031.