Abstract:

Noncovalent interactions (NIs) are present in the properties and functions of all matter, from solid state to soft materials. These interactions can range from a few kcal/mol to several hundreds or even thousands of kcal/mol. Ubiquitous to all molecules is the presence of dispersion interactions, which is a force allowing geckos to stick onto walls. However, dispersion remains challenging to intuitively understand and accurately predict. Our current understanding of dispersion interactions is that these are weak and instantaneous dipole moments. Due to this “weakness,” the many-body perturbation theory (MBPT) has been expected to accurately model NIs. I reassessed this perception when alarmingly large MBPT binding errors were observed for moderately sized complexes of ~100 atoms.

Here, I present the S66, S30L, and L7 benchmark results of the second-order Møller-Plesset MBPT (MP2), spin-scaled MP2, dispersion-corrected semilocal density functional approximations (DFAs), and post-Kohn–Sham random phase approximations (RPA). The large errors were confirmed when MP2 binding errors grew by a rate of ~0.1% per valence electrons, whereas RPA and dispersion-corrected DFA errors remain constant.

To understand the errors, I present an asymptotic adiabatic connection symmetry-adapted perturbation theory (AC-SAPT). A nonperturbative “screened” expression for the dispersion energy in terms of the monomer basis is obtained. AC-SAPT expansion of the interaction energy reveals that binding energy errors come from missing or an incomplete “electrodynamic” screening of the Coulomb interaction due to induced particle-hole pairs between electrons in different monomers. MP2 and higher-ordered perturbation theories lack this property leading to a divergent series within the AC-SAPT framework, whereas RPA converges. Furthermore, extension of the AC-SAPT to the thermodynamic limit establishes the exact constraint dispersion size-consistency. MBPT is found to violate this condition and RPA does not. Taken together, nonperturbative methods such as RPA or coupled cluster methods should be used.

Lastly, I apply this knowledge to understand the novel face-on halogen-π interaction between the lissoclimide family and eukaryotic 80S ribosome. The model is based on the structure-activity relationship (SAR) of the lissoclimide family inhibition of protein synthesis obtained from the X-ray co-crystal structure of the 80S ribosome and chlorolissoclimide. I present a manageable 122 atoms model system to predict potent inhibitors. I verify the model by correlating the experimental half-maximum inhibitor concentration (IC50) and RPA predicted binding energies between the lissoclimide derivatives and the eukaryotic 80S ribosome. The relationship reveals a negative correlation consistent with the anticipation that stronger binding leads to a more potent inhibitor. Based on these results, I proposed additional inhibitors and contributed to an expanded SAR knowledge of the lissoclimide family inhibition.