Deep learning approach is applied to detect the representative molecular behavior from the enormous and complex molecular dynamics data.

Molecular mechanism of the blueshift of water molecules.

Interionic distances are shorter in concentrated ionic solutions, thus instigating the interaction and overlap of hydration shells, as ions become separated by only one or two layers of water molecules. The simultaneous interaction of water with two oppositely charged ions has, so far, only been investigated by computer simulation studies, because the isolated vibrational spectroscopic signature of these molecules remains undetected. Our combined near-infrared spectroscopic and molecular dynamics simulation studies of alkali halide solutions present a distinct spectral feature, which is highly responsive to depletion of bulk water and merging of hydration shells. The analysis of this spectral feature demonstrates that absorption trends are in good agreement with the law of matching affinities, thus providing the first successful vibrational spectroscopic treatment of this topic. Combined with commonly observed near-infrared bands, this feature provides a spectral pattern that describes some relevant aspects of ionic hydration.

Rhodopsin is a light-driven G-protein-coupled receptor that mediates signal transduction in eyes. Internal water molecules mediate activation of the receptor in a rhodopsin cascade reaction and contribute to conformational stability of the receptor. However, it remains unclear how internal water molecules exchange between the bulk and protein inside, in particular through a putative solvent pore on the cytoplasmic. Using all-atom molecular dynamics simulations, we identified the solvent pore on cytoplasmic side in both the Meta II state and the Opsin. On the other hand, the solvent pore does not exist in the dark-adapted rhodopsin. We revealed two characteristic narrow regions located within the solvent pore in the Meta II state. The narrow regions distinguish bulk and the internal hydration sites, one of which is adjacent to the conserved structural motif “NPxxY”. Water molecules in the solvent pore diffuse by pushing or sometimes jumping a preceding water molecule due to the geometry of the solvent pore. These findings revealed a total water flux between the bulk and the protein inside in the Meta II state, and suggested that these pathways provide water molecules to the crucial sites of the activated rhodopsin.

Molecular dynamics (MD) is a powerful computational method for simulating molecular behavior. Deep neural networks provide a novel method of generating MD data efficiently, but there is no architecture that mitigates the well-known exposure bias accumulated by multi-step generations. In this paper, we propose a multi-step time series generator using a deep neural network based on Wasserstein generative adversarial nets. Instead of sparse real data, our model evolves a latent variable z that is densely distributed in a low-dimensional space. This novel framework successfully mitigates the exposure bias. Moreover, our model can evolve part of the system (Feature extraction) with any time step (Step skip), which accelerates the efficient generation of MD data. The applicability of this model is evaluated through three different systems: harmonic oscillator, bulk water, and polymer melts. The experimental results demonstrate that our model can generate time series of the MD data with sufficient accuracy to calculate the physical and important dynamical statistics.