For years, there has been ongoing interest in experimentally and theoretically understanding inversion and racemization processes. However, to the best of our knowledge, there has been no computational study that systematically investigated how well low-level quantum-chemical methods predict inversion barriers. Herein, we provide an answer to this question and we present the INV24 benchmark set of 24 high-level, ab initio inversion barriers. INV24 covers inversion in triatomics, in pyramidal molecules, in one cyclic system, and in various helical and bowl-shaped compounds. Our results indicate that previously applied DFT approximations combined with small basis sets are not reliable enough and that at least a triple-ζ basis is needed for meaningful results. Moreover, we show that intramolecular London dispersion influences the barriers by 2 kcal/mol or more and that dispersion corrections should always be applied to DFT results. With our analysis of 34 DFT approximations, we can reproduce the well-known Jacob’s Ladder scheme with (meta-)generalized-gradient-approximation methods underestimating barriers and global-hybrid DFT functionals performing better. Range-separated hybrids or Minnesota-type hybrids are not particularly superior to more conventional methods, such as B3LYP-D3. The by far best results are achieved with dispersion-corrected double hybrids, which give results below the chemical accuracy target of 1 kcal/mol. They also outperform wave-function second-order perturbation theory approaches and we recommend using them whenever possible. Given that our systematic study of INV24 is the first of its kind, our findings have the potential to change common practice in this field and they will guide future investigations of inversion processes.

First-principles quantum calculations for anharmonic vibrational spectroscopy of three protected dipeptides are carried out and compared with experimental data.

The sequence-dependent statistical mechanics of double-stranded nucleic acid, or dsNA, is believed to be essential in its biological functions. In turn, the equilibrium statistical mechanics behaviour of dsNA depends strongly both on sequence-dependent perturbations in its ground state shape away from an idealised, uniform, double-helical configuration, and on its fluctuations as governed by its sequence-dependent stiffness. We here describe the cgNA+web browser-based interactive tool for visualising the sequence-dependent ground states of dsNA fragments of arbitrary sequences, as predicted by the underlying cgNA+ coarse-grain model. Parameter sets are provided to model dsDNA, including the possibility of epigenetically modified CpG dinucleotide steps, dsRNA, and DNA:RNA Hybrid double helical fragments. The cgNA+web interface is specifically designed to compare ground state shapes of different sequences of the same dsNA, or analogous sequences of different dsNAs. The cgNA+web server is freely available at cgDNAweb.epfl.ch without any login requirement.

Marker-based Optical Motion Capture (OMC) systems and associated musculoskeletal (MSK) modelling predictions offer non-invasively obtainable insights into muscle and joint loading at an in vivo level, aiding clinical decision-making. However, an OMC system is lab-based, expensive, and requires a line of sight. Inertial Motion Capture (IMC) techniques are widely-used alternatives, which are portable, user-friendly, and relatively low-cost, although with lesser accuracy. Irrespective of the choice of motion capture technique, one typically uses an MSK model to obtain the kinematic and kinetic outputs, which is a computationally expensive tool increasingly well approximated by machine learning (ML) methods. Here, an ML approach is presented that maps experimentally recorded IMC input data to the human upper-extremity MSK model outputs computed from (‘gold standard’) OMC input data. Essentially, this proof-of-concept study aims to predict higher-quality MSK outputs from the much easier-to-obtain IMC data. We use OMC and IMC data simultaneously collected for the same subjects to train different ML architectures that predict OMC-driven MSK outputs from IMC measurements. In particular, we employed various neural network (NN) architectures, such as Feed-Forward Neural Networks (FFNNs) and Recurrent Neural Networks (RNNs) (vanilla, Long Short-Term Memory, and Gated Recurrent Unit) and a comprehensive search for the best-fit model in the hyperparameters space in both subject-exposed (SE) as well as subject-naive (SN) settings. We observed a comparable performance for both FFNN and RNN models, which have a high degree of agreement (r${}_{\mathrm{avg},\mathrm{SE},\mathrm{FFNN}}={0.90}_{\pm 0.19}$, r${}_{\mathrm{avg},\mathrm{SE},\mathrm{RNN}}={0.89}_{\pm 0.17}$, r${}_{\mathrm{avg},\mathrm{SN},\mathrm{FFNN}}={0.84}_{\pm 0.23}$, and r${}_{\mathrm{avg},\mathrm{SN},\mathrm{RNN}}={0.78}_{\pm 0.23}$) with the desired OMC-driven MSK estimates for held-out test data. The findings demonstrate that mapping IMC inputs to OMC-driven MSK outputs using ML models could be instrumental in transitioning MSK modelling from ‘lab to field’.