This work introduces a novel application of generative adversarial networks (GANs) for the prediction of starting geometries in transition state (TS) searches based on the geometries of reactants and products. The multi-dimensional potential energy space of a chemical reaction often complicates the location of a starting TS geometry, leading to the correct TS combining reactants and products in question. The proposed TS-GAN efficiently maps the space between reactants and products and generates reliable TS guess geometries, and it can be easily combined with any quantum chemical software package performing geometry optimizations. The TS-GAN was trained and applied to generate TS guess structures for typical chemical reactions, such as hydrogen migration, isomerization, and transition metal-catalyzed reactions. The performance of the TS-GAN was directly compared to that of classical approaches, proving its high accuracy and efficiency. The current TS-GAN can be extended to any dataset that contains sufficient chemical reactions for training. The software is freely available for training, experimentation, and prediction at https://github.com/ekraka/TS-GAN.

In this work hydrogen bonding in a diverse set of 36 unnatural and the three natural Watson Crick base pairs adenine (A)–thymine (T), adenine (A)–uracil (U) and guanine (G)–cytosine (C) was assessed utilizing local vibrational force constants derived from the local mode analysis, originally introduced by Konkoli and Cremer as a unique bond strength measure based on vibrational spectroscopy. The local mode analysis was complemented by the topological analysis of the electronic density and the natural bond orbital analysis. The most interesting findings of our study are that (i) hydrogen bonding in Watson Crick base pairs is not exceptionally strong and (ii) the N–H⋯N is the most favorable hydrogen bond in both unnatural and natural base pairs while O–H⋯N/O bonds are the less favorable in unnatural base pairs and not found at all in natural base pairs. In addition, the important role of non-classical C–H⋯N/O bonds for the stabilization of base pairs was revealed, especially the role of C–H⋯O bonds in Watson Crick base pairs. Hydrogen bonding in Watson Crick base pairs modeled in the DNA via a QM/MM approach showed that the DNA environment increases the strength of the central N–H⋯N bond and the C–H⋯O bonds, and at the same time decreases the strength of the N–H⋯O bond. However, the general trends observed in the gas phase calculations remain unchanged. The new methodology presented and tested in this work provides the bioengineering community with an efficient design tool to assess and predict the type and strength of hydrogen bonding in artificial base pairs.

For decades one has strived to synthesize a compound with the longest covalent C−C bond applying predominantly steric hindrance and/or strain to achieve this goal. On the other hand electronic effects have been added to the repertoire, such as realized in the electron deficient ethane radical cation in its D${}_{3d}$ form. Recently, negative hyperconjugation effects occurring in diamino-o-carborane analogs such as di-N,N-dimethylamino-o-carborane have been held responsible for their long C−C bonds. In this work we systematically analyzed CC bonding in a diverse set of 53 molecules including clamped bonds, highly sterically strained complexes such as diamondoid dimers, electron deficient species, and di-N,N-dimethylamino-o-carborane to cover the whole spectrum of possibilities for elongating a covalent C−C bond to the limit. As a quantitative intrinsic bond strength measure, we utilized local vibrational CC stretching force constants $k^a$(CC) and related bond strength orders BSO n(CC), computed at the $\omega$B97X-D/aug-cc-pVTZ level of theory. Our systematic study quantifies for the first time that whereas steric hindrance and/or strain definitely elongate a C−C bond, electronic effects can lead to even longer and weaker C−C bonds. Within our set of molecules the electron deficient ethane radical cation, in D${}_{3d}$ symmetry, acquires the longest C−C bond with a length of 1.935 Å followed by di-N,N-dimethylamino-o-carborane with a bond length of 1.930 Å. However, the C−C bond in di-N,N-dimethylamino-o-carborane is the weakest with a BSO n value of 0.209 compared to 0.286 for the ethane radical cation; another example that the longer bond is not always the weaker bond. Based on our findings we provide new guidelines for the general characterization of CC bonds based on local vibrational CC stretching force constants and for future design of compounds with long C−C bonds.

Recent biochemical studies have identified a cell surface receptor for thyroid and steroid hormones that bind near the arginine-glycine-aspartate (RGD) recognition site on the heterodimeric αvβ3 integrin. To further characterize the intermolecular interactions for a series of hormone analogues, combined quantum mechanical and molecular mechanical (QM/MM) methods were used to calculate their interaction energies. All calculations were performed in the presence of either calcium (Ca²⁺) or magnesium (Mg²⁺) ions. These data reveal that 3,5^′-triiodothyronine (T₃) and 3,5,3^′,5^′-tetraiodothyroacetic acid (T₄ac) bound in two different modes, occupying two alternate sites, one of which is along the Arg side chain of the RGD cyclic peptide site. These orientations differ from those of the other ligands whose alternate binding modes placed the ligands deeper within the RGD binding pocket. These observations are consistent with biological data that indicate the presence of two discrete binding sites that control distinct downstream signal transduction pathways for T₃.