Use of QM/MM scheme to reproduce macromolecule–small molecule noncovalent binding energy

Reliable and fast computation of macromolecule–small molecule noncovalent binding energy is crucial for structure-based molecular design, lead modification and molecular docking, and quantum mechanics/molecular mechanics (QM/MM) has recently emerged as a new and powerful tool to fulfill this purpose. In the present study, based on 20 high-quality, elaborately selected macromolecule–small molecule complex crystal structures we systematically compare the performance of QM/MM methods at different levels of theory and found that: (a) The calculation precision of semiempirical QM/MM, especially the AM1/AMBER, is comparable with or even better than that of first-principle-based scheme, but the computational cost of former is significant less than latter. (b) The convenient all-atom force field UFF appears to be a good candidate for replacing the nontrivial macromolecular force field AMBER in semiempirical QM/MM analysis of macromolecule–small molecule systems. (c) Although absolute error between semiempirical QM/MM-derived and experimentally measured binding energies is significant, they exhibit a good linear correlation to each other. (d) Direct interaction energy of macromolecule with small molecule is very favorable, but a large proportion of this favorable effect is used to pay off unfavorable desolvation penalty due to the burial of polar or charged molecular groups in the low dielectric complex interface during binding. (e) Many additional factors such as crystallographic resolution, protonized state, molecular flexibility and packing depth could affect the calculation precision of semiempirical QM/MM in predicting macromolecule–small molecule binding energy.