In Silico Optimization of Dihydropteridine Derivatives Targeting PLK1 for Glioblastoma: An Integrated QSAR, Docking, and Molecular Dynamics Study

Abstract

A 2D quantitative structure–activity relationship (QSAR) model was developed for a series of 34 dihydropteridine derivatives to predict their antitumor activity against glioblastoma. The multiple linear regression model (MLR), based on four descriptors (ATS7s, AATS8e, AATS3p, and AATS5p), demonstrated satisfactory statistical performance (R² = 0.714, R²_adj = 0.660, Q² = 0.513, RMSE = 0.096, F = 13.134, p < 0.0001), with strong external validation (R²_test = 0645). Model-guided optimization led to the design and screening of new structural analogs, which were subsequently docked against Polo-like kinase 1 (PLK1, PDB ID: 3BD6), a key mitotic regulator overexpressed in glioblastoma. Among the compounds evaluated, X14 exhibited the most favorable docking affinity (-10.5 kcal / mol), compared to -6.6 kcal/mol for Temozolomide (TMZ) and 8.1 kcal/mol for the reference compound N27. Although generally favorable ADMET profiles were observed, hepatotoxicity alerts were predicted for all compounds, which represents an important limitation supporting the prioritization of X14. In general, this study provides. Molecular dynamics simulations over 100 ns supported stable complex formation, with RMSD backbone values stabilizing around 2.5 to 3.0 Å. The ligands remained bound within the binding pocket throughout the simulation, exhibiting RMSD values below 1.5 Å for X14 and temozolomide and below 2.5 Å for N27, while the PLK1–TMZ complex showed higher structural fluctuations. A plausible synthetic route was proposed to assess the experimental feasibility of X14. Overall, these computational findings support the prioritization of X14 for further experimental validation in glioblastoma therapy.