A Five-Control Specificity Checklist Overturns an Apparent Phytochemical Disruptor of the VDAC1-Hexokinase-I Interface

Abstract

Hexokinase-I anchors to voltage-dependent anion channel 1 (VDAC1) at a membrane-buried glutamate, E73, and disrupting this contact is an established anticancer objective. The site-directed docking of natural product libraries into such hotspots routinely identifies disruptors, yet these candidates are rarely tested for either site specificity or ligand specificity. We screened a verified 43-compound library from Camptotheca acuminata against the VDAC1 E73 groove and then applied five orthogonal controls, calibrated against an empirical uncertainty budget. The conventional workflow was persuasive; ursolic acid ranked first at −7.25 kcal/mol, redocked reproducibly, remained associated with the groove through 100 ns of POPC-embedded dynamics, and returned −42.7 kcal/mol by Molecular Mechanics/Generalized Born Surface Area (MM-GBSA). Replicate calculations then showed that score uncertainty was dominated not by stochastic search, which contributes a standard deviation of 0.007 kcal/mol, but by the arbitrary definition of the site, 0.379 kcal/mol, and by receptor conformation, 0.550 kcal/mol. Against this scale, the controls rule out the candidates. Blind docking identified alternative sites with scores equal to or better than E73. A scan of twenty-four surface sites placed E73 within one box placement standard deviation of several unrelated positions. Neutralizing the E73 carboxylate changed affinity by −0.30 to +0.04 kcal/mol across fifteen compounds. Cholesterol, a membrane lipid, scored within the same uncertainty, and outscored the lead ligand under a second scoring function and in four of six relaxed receptor conformations. Contact occupancies reached 29% for the lead ligand compared to 100% for the reference ligand, NADH. We propose using the controls as a checklist, as illustrated by this case study.