In Silico Identification of 2,4-Diaminopyrimidine-Based Compounds as Potential CK1ε Inhibitors

Abstract

Background: Casein kinase 1 epsilon (CK1ε) plays a critical role in cancer progression by activating oncogenic signaling pathways, making it a target for cancer therapy. However, no inhibitors are currently available for clinical use, highlighting the need for novel therapeutic candidates. Methods: This study aimed to identify potential CK1ε inhibitors. To achieve this, a modified version of a previously reported pharmacophore model was applied to an ultra-large database of over 100 million compounds for virtual screening. Hits were filtered based on drug-likeness and pH-dependent pharmacophore compliance and then grouped according to their structural core. A representative compound from each structural group underwent molecular dynamic (MD) simulations and binding free energy calculations to predict its stability and affinity, allowing extrapolation of the results to the entire set of candidates. Results: Pharmacophore matching initially identified 290 compounds. After energy minimization, and an assessment of drug-likeness and pharmacophore compliance, we selected 29 structurally related candidates. MD simulations showed that most of the compounds representative of structural groups had stable binding modes, favorable intermolecular interactions, and free energies comparable to those of previously reported CK1ε inhibitors. An analysis of additional members of the most promising structural group showed that two 2,4-diaminopyrimidine-based compounds likely inhibit CK1ε. Conclusions: These findings provide structural insights into the design of CK1ε inhibitors, supporting compound optimization and the eventual development of targeted cancer therapeutics.

Keywords: CK1ε; cancer; diaminopyrimidine; molecular dynamic simulations; pharmacophore model; virtual screening.

Figure 1

Schematic representation of CK1ε’s role in the Wnt/β-catenin signaling pathway. Proteins are represented as ellipses, with CK1ε highlighted by a distinct shape. Target genes are represented as green rectangles. ATP and a competitive inhibitor are shown as yellow and red circles, respectively. Solid arrows indicate downstream signaling effects. Dashed arrows represent cellular outcomes resulting from the regulation of target genes.

Figure 2

Compound identification and RMSD analysis after energy minimization. (a) Identification of pharmacophore-matching compounds and their energy-minimized conformers within the ATP-binding site. (b) Pharmacophore matching (gray) vs. minimized conformers (colored). Compounds with an RMSD >2.0 Å (red chart) were discarded. Pharmacophore features: purple (aromatic), yellow (hydrogen bond acceptor: HBA), white (hydrogen bond donor: HBD), and green (hydrophobic).

Figure 3

Clustering of compounds based on Bemis & Murcko (B&M) frameworks. The compounds of interest were divided into eight groups based on shared structural frameworks. The numbers in the pie graph show the number of compounds per group. Each B&M framework is labeled with its corresponding group number and color-coded to match the pie graph. A randomly selected candidate (Cand) for MD studies from each cluster is displayed as a 2D structure.

Figure 4

Molecular dynamic simulations of CK1ε with reference (REF) or candidate (Cand) compounds. (a) Heavy-atom RMSD of ligands across two independent replicates (R1 and R2). (b) RMSD matrices showing pairwise comparisons of ligand conformations between two independent replicates (R1 and R2) of CK1ε/ligand MD simulations. Axes represent simulation progression based on each replicate with ligand conformations sampled every 10 ns.

Figure 5

Principal-component-based free energy landscapes for CK1ε in complex with reference compounds and candidates. Binding free energy landscape (FEL) contour for the simulated CK1ε/REF1–3 and Cand 1–8 complexes.

Figure 6

Binding energies of CK1ε/ligand complexes. (a) Free energy change (ΔG). The replicate with the lowest average ΔG is labeled accordingly. Error bars indicate the SD. (b) Per-residue decomposition of the enthalpic contribution. Two columns per compound are shown, each representing an independent MD replicate. The N-terminal lobe, hinge, and C-terminal lobe are indicated.

Figure 7

Probability density function plots for CK1ε/ligand complexes based on radius of gyration and RMSD. The sampled frame for each ligand representative binding mode is labeled accordingly, with red dashed lines indicating the associated Rg and RMSD values.

Figure 8

Representative binding modes of reference compounds and candidates. CK1ε is shown as a grey cartoon, with the hinge region colored orange. Ligands are depicted in grey. Interacting residues are shown as sticks: orange for residues involved in hydrogen or halogen bonds, green for those participating in hydrophobic interactions, and yellow for K38 and E52. Blue and green dashed lines indicate hydrogen and halogen bonds, respectively.

Figure 9

MD assessment of additional members of group 5. (a) Two-dimensional structures of Cand 5.1 and 5.2. (b) RMSD of ligands across two independent replicates (R1 and R2). Dashed lines indicate the subinterval used for ΔG calculations. (c) RMSD matrices comparing ligand conformations between the two independent replicates. (d) Principal-component-based free energy landscapes for CK1ε bound to Cand 5.1 and 5.2. (e) Probability density function plots for CK1ε complexes with Cand 5.1 and 5.2. (f) ΔG for CK1ε/Cand 5.1 and 5.2 complexes, along with per-residue decomposition of the enthalpic contribution. (g) Representative ligand conformations from the lowest-energy replicate. CK1ε is shown as a grey cartoon, with the hinge region colored orange and hydrophobic-interacting residues green. Ligands are depicted in grey. E83, L85, and D132 are shown as orange sticks, while K38 and E52 are yellow. Blue and green dashed lines indicate hydrogen and halogen bonds, respectively.