Improved sampling methodologies, more accurate force fields, and access to longer simulation time scales have led to an increased application of Relative Binding Free Energy (RBFE) calculations in drug discovery projects. In order to assess the strengths and limitations of such tools, adequate benchmark sets are required that challenge the methodology in certain well-defined aspects. We applied Free Energy Perturbation (FEP) calculations to six congeneric ligand pairs taken from the literature, in which addition of a functional group resulted in the displacement of buried binding site water molecules and compared the calculated relative binding free energies with the experimental ones. We started the perturbations from different initial solvation states and registered large inconsistencies (large hysteresis) between the calculated values. We furthermore applied a Grand Canonical Monte Carlo (GCMC) solvent sampling step prior to the FEP calculation that led to a smaller hysteresis for the simulations. By applying a hydration site analysis to the trajectories of the end-states of the perturbation, we could point out that the low accuracy of the predictions as well as the high dependence of the prediction on the chosen initial state is likely caused by the trapping of binding site water molecules and/or insufficient solvation of buried cavities that are formed upon completion of the perturbation. This work highlights that RBFE calculations can suffer from slow solvent exchange of buried parts of the binding sites with the bulk.

Mitochondria play an important role in sensing both acute and chronic hypoxia in the pulmonary vasculature, but their primary oxygen-sensing mechanism and contribution to stabilization of the hypoxia-inducible factor (HIF) remains elusive. Alteration of the mitochondrial electron flux and increased superoxide release from complex III has been proposed as an essential trigger for hypoxic pulmonary vasoconstriction (HPV). We used mice expressing a tunicate alternative oxidase, AOX, which maintains electron flux when respiratory complexes III and/or IV are inhibited. Respiratory restoration by AOX prevented acute HPV and hypoxic responses of pulmonary arterial smooth muscle cells (PASMC), acute hypoxia-induced redox changes of NADH and cytochrome c, and superoxide production. In contrast, AOX did not affect the development of chronic hypoxia-induced pulmonary hypertension and HIF-1α stabilization. These results indicate that distal inhibition of the mitochondrial electron transport chain in PASMC is an essential initial step for acute but not chronic oxygen sensing.

Preprocessing of Raman spectra is generally done in three separate steps: (1) cosmic ray removal, (2) signal smoothing, and (3) baseline subtraction. We show that a convolutional neural network (CNN) can be trained using simulated data to handle all steps in one operation. First, synthetic spectra are created by randomly adding peaks, baseline, mixing of peaks and baseline with background noise, and cosmic rays. Second, a CNN is trained on synthetic spectra and known peaks. The results from preprocessing were generally of higher quality than what was achieved using a reference based on standardized methods (second-difference, asymmetric least squares, cross-validation). From 105 simulated observations, 91.4% predictions had smaller absolute error (RMSE), 90.3% had improved quality (SSIM), and 94.5% had reduced signal-to-noise (SNR) power. The CNN preprocessing generated reliable results on measured Raman spectra from polyethylene, paraffin and ethanol with background contamination from polystyrene. The result shows a promising proof of concept for the automated preprocessing of Raman spectra.

The Platinum dataset of protein-bound ligand conformations was used to benchmark the ability of the MMFF94s force field to generate bioactive conformations by minimization of randomly generated conformers. Torsion angle parameters that generally caused wrong geometries were reparameterized by conducting dihedral scans using ab initio calculations at the MP2 level. This reparameterization resulted in a systematic improvement of generated conformations.

Computational methods, namely molecular dynamics (MD) simulations in combination with inhomogeneous fluid solvation theory (IFST) were used to retrospectively investigate various cases of ligand structure modifications that led to the displacement of binding site water molecules. Our findings are that water displacement per se is energetically unfavorable in the discussed examples, and that it is merely the fine balance between change in protein-ligand interaction energy, ligand solvation free energies, and binding site solvation free energies that determine if water displacement is favorable or not. We furthermore evaluated if we can reproduce experimental binding affinities by a computational approach combining changes in solvation free energies with changes in protein-ligand interaction energies and entropies. In two of the seven cases, this estimation led to large errors, implying that accurate predictions of relative binding free energies based on solvent thermodynamics is challenging. Nevertheless, MD simulations can provide insight regarding which water molecules can be targeted for displacement.