Integrated structure-based and fragment-guided design of isonicotinic acid derivatives as potential KDM4A inhibitors:...

Lysine-specific demethylase 4 A (KDM4A) is an Fe²⁺-dependent epigenetic regulator implicated in various cancers, representing an emerging target for anticancer drug development. However, the high polarity of its catalytic pocket poses a significant challenge in designing potent small-molecule inhibitors. In this study, a structure-based virtual screening workflow was implemented to optimize the known KDM4A inhibitor QC6352 and design novel isonicotinic acid derivatives with improved predicted binding affinity and dynamic stability. Two Python-based workflows employing fragment replacement and molecular breeding strategies were developed to generate virtual libraries while retaining the isonicotinic acid core. A total of 135,000 derivatives were produced and subjected to a hierarchical screening protocol involving molecular docking, ADMET-based filtering, induced fit docking (IFD), molecular dynamics (MD) simulations, and MM-GBSA rescoring. ADMET filtering was based on standard QikProp drug-likeness and pharmacokinetic criteria to prioritize compounds with acceptable predicted physicochemical, absorption, distribution, and safety-related profiles. Among the screened derivatives, QC-L1A and QC-L2B showed the most favorable overall profiles compared with the reference compound QC6352 across multiple computational parameters. QC-L1A exhibited the most favorable binding free energy (ΔG_bind = - 30.96 kcal/mol), representing a 3.12 kcal/mol improvement over QC6352 (- 27.84 kcal/mol), along with the lowest mean RMSD (0.807 Å), indicating marked conformational stability. QC-L2B achieved the most favorable IFD score (- 751.02 kcal/mol), surpassing QC6352 by 7.64 kcal/mol, and yielded a ΔG_bind of - 29.62 kcal/mol. MD simulations revealed stable coordination between the conserved isonicotinic acid core and the catalytic Fe²⁺ ion, together with persistent interactions with key active-site residues. Selectivity analysis, performed by IFD-based profiling against KDM4 isoforms and other related JmjC demethylases, indicated a preferential binding profile toward KDM4 family members, particularly KDM4A and KDM4D. Retrosynthetic feasibility was assessed using the Spaya CASP platform, with both lead compounds showing favorable RScore values and practical synthetic routes. Accordingly, these findings highlight the effectiveness of a fragment-guided computational workflow for rational inhibitor optimization and introduce promising scaffolds for KDM4A-targeted drug design.