Virtual screening of GSK-3β and CDK5 multi-target inhibitors targeting tauopathies

Tauopathies represent a heterogeneous group of neurodegenerative diseases characterized by abnormal aggregation and intracellular accumulation of the microtubule-associated protein tau in brain cells. Under physiological conditions, tau is mainly involved in the assembly and stabilization of microtubules, and in the regulation of axonal transport and neuronal structure. The activity of tau is regulated by several post-translational modifications, among which phosphorylation represents a key regulatory mechanism. The phosphorylation state of tau is controlled by a complex balance between the activity of kinases and phosphatases. Among the kinases involved, GSK-3β and CDK5 play central and interconnected roles in tau regulation. CDK5 is normally activated through its association with the neuronal cofactors p35 and p39, which ensure control of its kinase activity. However, under pathological conditions p35 can be cleaved, resulting in formation of the CDK5/p25 complex, which leads to aberrant phosphorylation of multiple neuronal substrates, including tau. GSK-3β is another relevant tau phosphorylating kinase that is constitutively active under basal conditions and it is mainly regulated through inhibitory phosphorylation. When this regulatory mechanism is impaired, GSK-3β overactivates promoting tau hyperphosphorylation. Importantly, CDK5 and GSK-3β are functionally linked in tau pathology and Alzheimer's Disease (AD), their dysregulated interplay representing a critical molecular mechanism underlying tau hyperphosphorylation and aggregation. Given their central roles in driving tau pathology, both kinases represent attractive target for the development of polypharmacological therapeutic agents against AD. Polypharmacologycal approaches have recently demonstrated to provide improved therapeutic effects in complex therapeutic contexts as AD. On these premises, the aim of this project is the identification of novel tau-centric multi-target compounds with GSK-3β and CDK5 inhibitory activity. To this aim, I first curated datasets of publicly available compounds with known inhibitory activity against GSK-3β and CDK5. Then, around 2000 molecular descriptors and different types of fingerprints were calculated for each compound by means of RDKit and Mordred; this data was then used to develop of a series of machine learning classifiers with the aim of discriminating active (activity values below 1 μM) from inactive (activity values above 1 μM) ligands against these two kinases. Subsequently, the models with the best predictive performances were applied to screen large chemical libraries of commercial compounds in order to identify putative ligands able to bind both targets. Finally, docking calculations were performed on selected crystallographic conformations of GSK-3β and CDK5, to evaluate whether the most promising compounds resulting from the ML-based screenings were able to establish favourable interactions with hot-spot residues of the two enzymes. A visual inspection of the predicted binding modes allowed to identify a set of putative tau-centric multi target agents with GSK-3β and CDK5 inhibitory activity, to be experimentally assessed for their activity on recombinant proteins.