We present a supercomputer-driven pipeline for in silico drug discovery using enhanced sampling molecular dynamics (MD) and ensemble docking. Ensemble docking makes use of MD results by docking compound databases into representative protein binding-site conformations, thus taking into account the dynamic properties of the binding sites. We also describe preliminary results obtained for 24 systems involving eight proteins of the proteome of SARS-CoV-2. The MD involves temperature replica exchange enhanced sampling, making use of massively parallel supercomputing to quickly sample the configurational space of protein drug targets. Using the Summit supercomputer at the Oak Ridge National Laboratory, more than 1 ms of enhanced sampling MD can be generated per day. We have ensemble docked repurposing databases to 10 configurations of each of the 24 SARS-CoV-2 systems using AutoDock Vina. Comparison to experiment demonstrates remarkably high hit rates for the top scoring tranches of compounds identified by our ensemble approach. We also demonstrate that, using Autodock-GPU on Summit, it is possible to perform exhaustive docking of one billion compounds in under 24 h. Finally, we discuss preliminary results and planned improvements to the pipeline, including the use of quantum mechanical (QM), machine learning, and artificial intelligence (AI) methods to cluster MD trajectories and rescore docking poses.

Some years ago we developed a theoretical-computational hybrid quantum/classical methodology, the Perturbed Matrix Method (PMM), to be used in conjunction with molecular dynamics simulations for the investigation of chemical processes in complex systems, that proved to be a valuable tool for the simulation of relevant experimental observables, e.g., spectroscopic signals, reduction potentials, kinetic constants. In typical PMM calculations the quantum sub-part of the system, the quantum centre, is embedded into an external perturbing field providing a perturbation operator explicitly calculated up to the dipolar terms. In this paper we further develop the PMM approach, beyond the dipolar terms in the perturbation operator expansion, by including explicitly the quadrupolar terms and/or by expanding the perturbation operator on each atom of the quantum centre. These different levels of the perturbation operator expansion, providing different levels of theory, have been tested by calculating three different spectroscopic observables: the spectral signal of liquid water and aqueous benzene due to the lowest energy electronic excitation and the infrared amide I band of aqueous trans-N-methylacetamide. All the systems tested show that, even though the previous PMM level of theory is already capable of reproducing the main features of the spectral signal, the higher levels of theory improve the quantitative reproduction of the spectral details.

The wide range of variability of the reduction potential (E(0)) of blue-copper proteins has been the subject of a large number of studies in the past several years. In particular, a series of azurin mutants have been recently rationally designed tuning E(0) over a very broad range (700 mV) without significantly altering the redox-active site [Marshall et al., Nature, 2009, 462, 113]. This clearly suggests that interactions outside the primary coordination sphere are relevant to determine E(0) in cupredoxins. However, the molecular determinants of the redox potential variability are still undisclosed. Here, by means of atomistic molecular dynamics simulations and hybrid quantum/classical calculations, the mechanisms that determine the E(0) shift of two azurin mutants with high potential shifts are unravelled. The reduction potentials of native azurin and of the mutants are calculated obtaining results in good agreement with the experiments. The analysis of the simulations reveals that only a small number of residues (including non-mutated ones) are relevant in determining the experimentally observed E(0) variation via site-specific, but diverse, mechanisms. These findings open the path to the rational design of new azurin mutants with different E(0).

The essential features of the experimental Soret bands of two covalent cages, consisting of two zincporphyrins connected by four flexible spacers, are for the first time interpreted and characterized at a molecular level by means of a mixed quantum/classical procedure based on molecular dynamics (MD) simulation and the perturbed matrix method (PMM). The same method allows also for a comprehensive interpretation of the changes in the UV− visible absorbance of the cages upon silver(I) complexation to the peripheral binding sites. Although the zinc-to-zincdistance is found to be similar in both cages, the MD-PMM calculations show that the conformation adopted by the cage with longer linkers corresponds to more slipped porphyrins, giving rise to a red-shifted (7−8 nm), broader, and slightly split Soret peak with respect to the cage with shorter linkers. The process of silver(I) complexation separates the two porphyrins in a face-to-face conformation in both cages, resulting in narrower (and more similar) Soret bands due to a reduced excitonic coupling. Despite the similar features of the spectra of the two silver(I)complexed cages, a slight difference in the peak maxima of about 2 nm is observed, arising from a slightly shorter zinc-to-zinc distance in the cage with longer linkers. These results show that the MD-PMM methodology is a reliable method to obtain information on the relative disposition and exciton coupling interaction of porphyrins in flexible systems in solution, from the analysis of their absorption spectra.

Calcitonin is a polypeptidic hormone involved in calcium metabolism in the bone. It belongs to the amyloid protein family, which is characterized by the common propensity to aggregate acquiring a beta-sheet conformation and include proteins associated with important neurodegenerative diseases. Here we show for the first time, to our knowledge, by transmission electron microscopy (TEM) that salmon-calcitonin (sCT) forms annular oligomers similar to those observed for beta-amyloid and alpha-sinuclein (Alzheimer's and Parkinson's diseases). We also investigated the interaction between sCT and model membranes, such as liposomes, with particular attention to the effect induced by lipid "rafts" made of cholesterol and G(M1). We observed, by TEM immunogold labeling of sCT, that protein binding is favored by the presence of rafts. In addition, we found by TEM that sCT oligomers inserted in the membrane have the characteristic pore-like morphology of the amyloid proteins. Circular dichroism experiments revealed an increase in beta-content in sCT secondary structure when the protein was reconstituted in rafts mimicking liposomes. Finally, we showed, by spectrofluorimetry experiments, that the presence of sCT allowed Ca(2+) entry in rafts mimicking liposomes loaded with the Ca(2+)-specific fluorophore Fluo-4. This demonstrates that sCT oligomers have ion-channel activity. Our results are in good agreement with recent electrophysiological studies reporting that sCT forms Ca(2+)-permeable ion channels in planar model membranes. It has been proposed that, beyond the well-known interaction of the monomer with the specific receptor, the formation of Ca(2+) channels due to sCT oligomers could represent an extra source of Ca(2+) entry in osteoblasts. Structural and functional data reported here support this hypothesis.