Supramolecular chemistry from small molecules to cavitands: predictive and experimental approaches

Abstract

We used the hydrogen-bond propensity (HBP) protocol for predicting if a co-crystal would form or not for six different target molecules in combination with 25 potential co-formers each. The correct outcome was successfully predicated 92-95% of the time, which indicates that for a series of small molecules, HBP is a very reliable indicator for determining if a co-crystal will form. In order to examine if hydrogen and halogen-bonded systems can mimic each other, we conducted co-crystallization experiments using eight targets in combination with 25 co-formers. Molecular electrostatic potential surface (MEPS) calculations were used to guide the experimental space. The results suggested that in all the cases, hydrogen and halogen-bonds mimic each other. Validation studies of MEPS gave a prediction accuracy range of 64-84%. Three different factors, hydrogen-bond propensity (HBP), hydrogen-bond coordination (HBC), and hydrogen-bond energies (HBE), were evaluated to predict the experimental outcomes of attempted co-crystallizations between two known drug molecules, Nevirapine and Diclofenac, and a series of potential co-formers. HBP gave the correct result in 26 out of 30 cases, whereas the HBC method predicted the correct outcome in 22 out of 30 cases. Finally, HBE gave the correct result in 23 out of 30 experiments. In those cases, where the crystal structure of a co-crystal of either Nevirapine or Diclofenac was known, we also examined how well the three methods predicted which primary hydrogen-bond interactions were present in the crystal structure. HBP correctly predicted 6 out of 6 cases, HBC could not predict any of the synthon formations correctly, and HBE successfully predicted 1 out of 6 cases. Three different factors, hydrogen-bond propensity (HBP), hydrogen-bond energy (HBE), and molecular complementarity (MC), were used for predicting co-crystallization outcomes of seven active pharmaceutical ingredients (APIs) in combination with 42 potential co-formers. Validation studies indicate that individually the methods did not offer high accuracy. Hence, we implemented a combination of methods using Venn diagrams. The results suggested that the combination of MC and HBP method yielded the highest accuracy of 80%. In order to ease the complexity of the predictive approaches, we designed CoForm, an automatic app for predicting co-crystallization outcomes. CoForm was used to predict co-crystallization outcomes for Loratadine and Desloratadine in combination with 42 generally regarded as safe (GRAS) list co-formers. The predictive abilities of CoForm were compared to commercially available tools such as HBP and MC. The results indicate that CoForm delivered a success rate of 80% for both Loratadine and Desloratadine in comparison to HBP 76% and 54%, respectively, and MC 39% and 22%, respectively. Six cavitands were synthesized to explore the guest encapsulation via cavity inclusion, using five solvents (xylene, dichloromethane, ethyl acetate, acetonitrile, and dimethyl sulfoxide) with varying dipole moment as a potential guest. Cavity inclusion studies indicated that the cavitands could encapsulate a guest in five different orientations, with stoichiometries 1:1 to 1:2. A selectivity study showed that the cavitands were more selective towards DMSO, which also had the highest dipole moment. Finally, cavitands were used as molecular containers for fragrant compounds and heterocyclic-N- oxides. In both cases, the host interacted with guests in 1:2 stoichiometry, as indicated by NMR titrations and TGA analysis. The host-guest interaction of cavitands with fragrant compounds lowered the volatility of the fragrant compounds from 5-10 mins to over a month.