project picture

Spring 2021 challenge: winner

Design of Indirect CDK5-Mediated Tau-Hyperphosphorylation Inhibitors

Ahmed Elkerdawy, Faculty of Pharmacy - Cairo Uni., Cairo, Egypt

During the last 12 months we have been working on designing small molecules to interfere with calpain binding to p35 (CDK5’s activator) to prevent its conversion to the pathogenic p25. It is a new hypothesis with no previous reference compounds, so we mainly relied on the computational de novo design techniques in both the hit to lead and lead optimization stages of our project. Inspirator mode in SeeSAR came in handy in each of the aforementioned stages through its powerful and user-friendly core-replacement and fragment-growing options. Hits were first generated by virtual screening of the ZINC database using a pharmacophore query generated for the most frequent seed fragments. However, most of the retrieved hits were not available for purchasing and/or with low synthetic feasibility. That was overcome through analysis of the scaffolds by DataWarrior software and sending the most frequent scaffold to Inspirator and applying interactive manual modifications assessed by Hyde- estimated affinity. Using molecule editor mode the molecules were modified to be more synthetically feasible. Three main scaffolds with few derivatives were generated and predicted to be Blood-Brain Barrier permeant and not a substrate of P-glycoprotein. A representative compound for each scaffold was then subjected to a 100 ns molecular dynamics simulation using GROMACS package and the binding free energy along with the binding mode were then studied and the representative compound with the highest binding energy was further subjected to a 500 ns molecular dynamics simulation run. Again, the binding free energy along with the binding mode and their effect on secondary structure of the calpain binding site were then studied in detail. The binding poses of the ligand were clustered and it was found that the most dominant pose agreed with the best docking pose resulted from FlexX. For calpain enzyme to bind to its substrate protein, the substrate protein should be unstructured. If there were structured domains like a helix or sheet, it unfolds to allow for the calpain binding. Our extended molecular dynamics study showed that our designed compound relatively stabilized a certain helix that tends to unfold to accommodate the calpain binding. Accordingly, we decided to increase our chemical space by generating more derivatives by SeeSAR's Inspirator mode for those three main scaffolds. All designed derivatives were subjected to molecular docking and those fulfilling the criteria of interfering with calpain binding were cherry-picked for synthesis. Synthesis of a series of derivatives of each of those promising scaffolds were finally initiated to be tested biologically.
After 1 year, Ahmed has achieved the following goals:
  1. De novo design of indirect small molecule calpain-p35 binding inhibitors. Initially we docked a fragment library and ranked the results based on their binding energies and analyzed the frequency of fragments binding with certain residues by R software to pinpoint the “Key residues”. A pharmacophore was then generated for the most frequent fragments, and used to virtually screen ZINC database. This resulted in several hits which were mostly not purchasable and/or with low synthetic feasibility. DataWarrior software was used to analyze the scaffolds of those hits, and the most frequent scaffold was used as a starting point for further extensive optimizations using both SeeSAR's Inspirator mode and applying manual modifications interactively assessed by Hyde-estimated affinity using SeeSAR's Molecule Editor mode. These modifications resulted in three synthetically feasible, Blood-Brain Barrier permeant and non-P-glycoprotein substrates scaffolds.
  2. Testing our hypothesis through molecular dynamics simulation: Small molecules showed satisfying binding modes that is claimed to interfere with p35 unfolding in molecular docking simulation by FlexX. The molecules docking poses were further studied through molecular dynamics simulations to gain deeper insight about the binding modes of these designed compounds. To this end, GROMACS package was used and initial 100 ns simulations were run for a representative compound of each of the promising scaffolds. Then, the most promising derivatives in terms of binding energy and binding mode was subjected further to a 500 ns molecular dynamics simulation run, again the binding free energy along with the binding mode and their effect on stabilizing the secondary structure of the calpain binding site were studied in detail. The binding poses of the ligand were then clustered and it was found that the most dominant pose is in agreement with the best docking pose resulted from FlexX.
  3. Transformation from the dry lab phase to the wet lab phase: Encouraged by the results achieved by the molecular docking, molecular dynamics, and pharmacokinetic properties predictions of the three main designed scaffolds and their series of derivatives, their synthetic feasibility was predicted by SwissADME web tool, Spaya AI score, a metric related to the probability of a disconnection to happen, in addition to synthetic complexity score (SCScore) generated by MIT, a learned synthetic complexity metric trained on reactions from Reaxys database, which is a database built on expert-curated chemistry information. Eventually, after a lot of optimization trials guided mainly by SeeSAR's Inspirator mode, reasonable synthetic accessibility scores were reached and synthetic schemes were devised relying on suggestions by the Reaxys database. Starting materials were purchased, and the actual synthesis process has been initiated.