Thiazolidinedione-based PI3Kα inhibitors: an analysis of biochemical and virtual screening methods

A series of synthesized and commercially available compounds were assessed against PI3Kα for in vitro inhibitory activity and the results compared to binding calculated in silico. Using published crystal structures of PI3Kγ and PI3Kδ co-crystallized with inhibitors as a template, docking was able to identify the majority of potent inhibitors from a decoy set of 1000 compounds. On the other hand, PI3Kα in the apo-form, modeled by induced fit docking, or built as a homology model gave only poor results. A PI3Kα homology model derived from a ligand-bound PI3Kδ crystal structure was developed that has a good ability to identify active compounds. The docking results identified binding poses for active compounds that differ from those identified to date and can contribute to our understanding of structure-activity relationships for PI3K inhibitors.