Tackling the GDCC hosts with ATM
by Solmaz Azimi
(We have submitted our ATM predictions for the SAMPL8 challenge.)
SAMPL, Statistical Assessment of the Modeling of Proteins and Ligands, is a series of blind predictive challenges aimed at testing computational models of protein-ligand binding for improving drug discovery. The project not only includes protein-ligand modeling, but also applies smaller, less complex molecular systems, such as hydration free energy of solutes and binding of host-guest systems.
The SAMPL8 challenge involves the Gibb deep cavity cavitand (GDCC) hosts binding to a series of five guests. GDCC is similar to the “octa acid” hosts that were contended in previous SAMPL challenges, namely TEMOA from SAMPL6. GDCC hosts present a considerable interest to the chemistry community because of their ease of synthesis and their propensity to dimerize into supramolecular nano-capsules (1, 2). Although monomeric, contact with an appropriate guest will induce dimerization for these hosts (2). Each GDCC host is a bowl-shaped amphiphile with a concave pocket and an elaborate rim, both of which are hydrophobic. What makes the host water-soluble, however, is its outer surface, which is coated with hydrophilic carboxylic acid groups.
The two GDCC hosts of SAMPL8 include tetraethyl octal acid (TEMOA) and tetraethyl octa acid (TEETOA), both of which differ in the substituents around the cavity rim. Of the five guests, four are negatively charged and one is expected to be neutral in the context of 10 mM sodium phosphate buffer and pH 11.5. The structures for the host and guests are shown below, as provided by the SAMPL8 organizers. The guest containing the halogen, 4-bromophenol, is particularly tricky. We expect multiple species, namely the deprotonated and protonated forms of the guest, to contribute to binding and therefore, treated complexes of this guest with particular care.
We employed our novel Alchemical Transfer Method (ATM) to compute standard binding free energies of the TEETOA and TEMOA complexes. ATM utilizes an alchemical intermediate state, similar to the notorious Double Decoupling Method (DTM), to link the unbound and bound states of molecular association in the thermodynamic path. At the intermediate state in ATM, the guest interacts simultaneously with the solvent bulk and the binding region of the host at half strength. The intermediate state in DTM on the other hand is defined as the guest in vacuum. In contrast to DTM, ATM is performed in the same solvent box of the thermodynamic cycle, in which the guest does not leave the solvated system.
The excess binding free energy is expressed as the difference of the free energies between the unbound and bound states. In the context of ATM, the unbound state is defined as a configuration in which the guest is in the solvent bulk and the bound state as a configuration in which the guest is in the binding site region of the host. In order to rigidly translate the guest from the binding site of the host and into the solvent environment, a constant translation vector is defined such that a configuration of the bound state maps to a unique configuration in the unbound state. As a result, all the other degrees of freedom of the system remain unchanged.
The alchemical path in ATM is split in two “legs” that represent the bound and unbound end states and their respective hybrid potential energy functions. If lambda = 0 represents the bound state, in which the guest interacts only with the binding site of the host, the alchemical pathway is terminated at lambda = 1/2. If lambda = 1 represents the unbound state, leg 2 connects the unbound state to the intermediate state similarly to leg 1, in which the alchemical pathway is extended up to lambda = 1/2. The binding free energy is expressed as the difference between leg 2 and leg 1 of the alchemical transformation.
Lambda = 1/2 represents the intermediate state of ATM. In this alchemical transformation, the intermediate state is unique such that it represents an ensemble in which the guest has simultaneous and symmetric interactions with not only the binding site of the host, but also the solvent environment. At this intermediate state, the guest interacts with its relative components at half strength and with minimal steric clashes, which preserves the original and soft-core perturbation potentials.
The SAMPL8 GDCC challenge is an excellent context to test ATM, as it allows us to assess not only conformational sampling with regards to sterics nad flexibility, but also protonations states that might be of interest in addressing for binding free energy protocols.
(1) Liu, Simin et al. An Improved Synthesis of ‘Octa-Acid’ Deep-Cavity Cavitand. Supramolecular chemistry. 2017, 23.6, 480–485.
(2) Saltzman, Alexander et al. Emergence of Non-Monotonic Deep Cavity Cavitand Assembly with Increasing Portal Methylation. Molecular systems design & engineering. 2020, 5.3, 656–665.