Abstract
The search for novel low-molecular regulators using molecular docking continues to be crucial for addressing challenges in modern biomedical science. However, the current literature lacks examples of modeling covalent interactions between the ligands being docked and those already present within the proteins, such as enzyme cofactors. This study aims to improve the existing algorithms for modeling such interactions, exemplified by those in thiamine diphosphate (ThDP)-dependent enzymes. Structures containing adducts of ThDP with enzyme substrates or inhibitors are used as protein templates; the putative ligand models are prepared as (R)- or (S)-hydroxy derivatives. The Gnina framework with AD4 or Vinardo favors ligand conformations resembling those found in the protein templates and consistent with their relative inhibitory potentials in experiments in vitro. For example, local hydrophobic regions within pyruvate and branched-chain 2-oxo acid dehydrogenase structures favor the binding of esterified substrate analogs compared to their de-esterified counterparts. The preferred binding of esterified vs. de-esterified ligands is absent or even reversed for 2-oxoglutarate dehydrogenase. As a result, covalent docking of 2-oxo acid analogs to enzyme structures containing ThDP coenzyme offers a predictive capability for protein-ligand complex formation and should be used when inhibitors mimic transition states in enzymatic reactions, as observed with ThDP-dependent catalysis.