Linear Discriminant Analysis for the in Silico Discovery of Mechanism-Based Reversible Covalent Inhibitors of a Serine Protease: Application of Hydration Thermodynamics Analysis and Semi-empirical Molecular Orbital Calculation.

We recently reported that the Gibbs free energy of hydrolytic water molecules (ΔGwat) in acyl-trypsin intermediates calculated by hydration thermodynamics analysis could be a useful metric for estimating the catalytic rate constants (kcat) of mechanism-based reversible covalent inhibitors. For thorough evaluation, the proposed method was tested with an increased number of covalent ligands that have no corresponding crystal structures. After modeling acyl-trypsin intermediate structures using flexible molecular superposition, ΔGwat values were calculated according to the proposed method. The orbital energies of antibonding π* molecular orbitals (MOs) of carbonyl C=O in covalently modified catalytic serine (Eorb) were also calculated by semi-empirical MO calculations. Then, linear discriminant analysis (LDA) was performed to build a model that can discriminate covalent inhibitor candidates from substrate-like ligands using ΔGwat and Eorb. The model was built using a training set (10 compounds) and then validated by a test set (4 compounds). As a result, the training set and test set ligands were perfectly discriminated by the model. Hydrolysis was slower when (1) the hydrolytic water molecule has lower ΔGwat; (2) the covalent ligand presents higher Eorb (higher reaction barrier). Results also showed that the entropic term of hydrolytic water molecule (-TΔSwat) could be used for estimating kcat and for covalent inhibitor optimization; when the rotational freedom of the hydrolytic water molecule is limited, the chance for favorable interaction with the electrophilic acyl group would also be limited. The method proposed in this study would be useful for screening and optimizing the mechanism-based reversible covalent inhibitors.