De Novo Design and Molecular Dynamics Evaluation of AZ67-Derived Allosteric PFKFB3 Modulators for Anticancer Metabolic Drug Discovery

Abstract

PFKFB3 is a druggable regulator of fructose-2,6-bisphosphate production, glycolytic flux, angiogenic metabolism, and tumour-cell survival signalling. This study redesigned the AZ67 allosteric scaffold to prioritise computational leads for anticancer metabolic drug discovery. ChemoDOTS enumeration produced 575,760 virtual products and 509,921 standardised unique structures. A reconstructed two-branch workflow applied sequential molecular-weight, lipophilicity, and relaxed drug-like criteria in one branch and PAINS filtering in parallel; intersection with Brenk-alert exclusion retained 10,996 candidates. Five-objective Pareto ranking and diversity filtering selected 50 compounds. A separate six-descriptor AZ67-referenced triage rule identified 27 comparator-qualified candidates but was not interpreted as proof of improved overall drug-likeness. Three nonredundant representatives—MOL_0052111, MOL_0051973, and MOL_0000552—were selected to test distinct substitution hypotheses, followed by docking in the PFKFB3 allosteric pocket (PDB 5AJV), 50-ns molecular dynamics, and MM/GBSA estimation. GNINA outputs were reported according to their distinct conventions: the empirical docking score was interpreted as more favourable when more negative, whereas CNNaffinity was reported on a positive pK scale, with higher values indicating stronger predicted affinity. The selected derivatives had higher molecular weight and topological polar surface area and lower QED than AZ67, although several had lower logP and higher Fsp3. MOL_0000552 gave an empirical docking score of -7.55 kcal/mol and CNNaffinity of 8.18 pK, compared with -4.33 kcal/mol and 4.21 pK for AZ67. It showed the least variable ligand trajectory among the derivatives and an MM/GBSA profile near -65 kcal/mol versus approximately -40 kcal/mol for AZ67. MOL_0051973 was a secondary lead, whereas MOL_0052111 underwent a marked ligand-RMSD shift after approximately 25 ns. MOL_0000552 therefore represents a prioritised computational hypothesis, not a validated PFKFB3 inhibitor. Biochemical activity, isoform selectivity, cellular effects, ADMET properties, and reproducibility in longer replicate simulations require confirmation.