Molecular Dynamics simulations and Computational methods in Life Science

Explicit molecular modelling of biological membrane systems is computationally expensive due to the large number of solvent particles and slow membrane kinetics. Here authors present a framework for integrating coarse-grained membrane models with continuum-based hydrodynamics which facilitates efficient simulation of large biomembrane systems.

Potassium (K+) channels, such as MthK, are essentional for many biological processes, but how lipid-protein interactions regulate ion permeation of K+ channels remained unclear. Here authors conducted molecular dynamics simulations of MthK and observed different ion permeation rates of MthK in membranes with different properties.

Combining realistic coarse-grained simulations with a percolation transition theory, this study elucidates the microscopic mechanism that governs the selectivity of passive, unassisted transport through the nuclear pore complex.

TMEM16 lipid scramblases transport lipids and also operate as ion channels with highly variable ion selectivities and various physiological functions. Using computational electrophysiology simulations, the authors identify the main ion-conductive state of TMEM16 lipid scramblases and find that lipid headgroups modulate ion permeability and regulate ion selectivity of TMEM16 proteolipidic pores.

The actin homolog MreB directs cell-wall insertion and maintains cell shape in many rod-shaped bacteria. Here, Shi et al. perform molecular dynamics simulations for MreB to extract mechanical parameters for inputs into a coarse-grained biophysical polymer model that predicts MreB filament properties.

Computer simulations of large-scale changes in membrane shape are challenging since they occur across a wide range of spatiotemporal scales. Here, authors present a multiscale algorithm that backmaps a continuum membrane model represented as a dynamically triangulated surface to its corresponding molecular model based on the coarse-grained Martini force field.

Prestin, a motor protein, plays a major role in sound amplification. Using molecular dynamics simulations, the authors show that prestin causes membrane deformation patterns thereby achieving a particular lipid-mediated alignment in the membrane.

The three-dimensional organization of chromatin plays critical roles in regulating genome function. Here the authors apply a near atomistic model to study the structure and dynamics of the chromatin folding unit - the tetra-nucleosome - to provide insight into how chromatin folds.

Nucleosomes tightly wrap ~147 DNA base pairs around an octamer of histone proteins, but how nucleosome structural dynamics affect genome functioning is not completely clear. Here authors employ all-atom molecular dynamics simulations of nucleosome core particles and observe that octamer dynamics and plasticity enable DNA unwrapping and sliding.

RNA can be used as a programmable tool for detection of biological analytes. Here the authors use deep neural networks to predict toehold switch functionality in synthetic biology applications.

In quantitative genetics, it is widely assumed that mutations combine additively or epistasis can be predicted with statistical or mechanistic models. Here, the authors use the phage lambda repressor model to show how biophysical ambiguity and non-monotonic functions confound phenotypic prediction.

Gene-regulatory networks are thought to be complex, and yet perturbation of just a few transcription factors (TFs) can have major consequences. Here the authors apply DNA polymer modelling and simulations to predict how 3D genome structure and TF-DNA interactions can give rise to transcriptional regulation operating over broad genomic regions, where small perturbations can have long-reaching effects.

DNA probes used in next generation sequencing (NGS) have variable hybridisation kinetics, resulting in non-uniform coverage. Here, the authors develop a deep learning model to predict NGS depth using DNA probe sequences and apply to human and non-human sequencing panels.

The limited availability of high-resolution 3D RNA structures for model training limits RNA secondary structure prediction. Here, the authors overcome this challenge by pre-training a DNN on a large set of predicted RNA structures and using transfer learning with high-resolution structures.

Interactions between proteins and RNA are an important mechanism for post-transcriptional regulation, but predicting these interactions is difficult. Through a deep learning approach, here the authors predict RNA-binding sites and binding preference based on the local physicochemical properties of the protein surface.

The intrinsic disorder of histone tails poses challenges in their characterization. Here the authors apply extensive molecular dynamics simulations of the full nucleosome to show reversible binding to DNA with specific binding modes of different types of histone tails, where charge-altering modifications suppress tail-DNA interactions and may boost interactions between nucleosomes and nucleosome-binding proteins.

Resolving nucleosomes with chemical accuracy inside sub-Mb chromatin provides molecular insight into the modulation of chromatin structure and its liquid–liquid phase separation (LLPS). By developing a multiscale chromatin model, the authors find that DNA breathing enhances the valency, heterogeneity, and dynamics of nucleosomes, promoting disordered folding and LLPS.

Multistep nucleation phenomena are of considerable fundamental interest. Here the authors combine molecular dynamics, machine learning and molecular cluster analysis to investigate the multistep nucleation of smectic clusters from a nematic fluid that cannot be accounted for by the classical nucleation theory.

The domain growth in phase separation is usually believed to be controlled by material transport. Tateno and Tanaka find that when phase separation causes strong dynamic asymmetry between the two phases, mechanical relaxation dominates and leads to a universally valid scaling law

Simulations reveal concerted interactions between the SARS-CoV-2 spike trimers and ACE2 receptors that result in cooperative spike binding and shedding, and further suggest that variant efficacy is promoted by increased RBD opening or S1/S2 cleavage efficiency.

The authors present a machine learning approach that combines baseline multimodal data to accurately predict individualised trajectories of future pathological tau accumulation at asymptomatic and mildly impaired stages of Alzheimer’s disease.

Molecular dynamics (MD) techniques enable atomic-level observations, but simulations of “slow” biomolecular processes are challenging because of current computer speed limitations. Here, the authors develop a method to accelerate MD simulations by high-frequency ultrasound perturbation and reveal binding events between the protein CDK2 and its small-molecule inhibitors.

In this work the authors propose a multiscale computational approach, integrating atomistic and coarse-grained models simulations, to study the thermodynamic and kinetic factors playing a major role in the liquid-to-solid transition of biomolecular condensates. It is revealed how the gradual accumulation of inter-protein β-sheets increases the viscosity of functional liquid-like condensates, transforming them into gel-like pathological aggregates, and it is also shown how high concentrations of RNA can decelerate such transition.

In this work the authors report atomically detailed computer simulations revealing the binding mechanisms of small molecule drugs to an intrinsically disordered region of the androgen receptor, a castration-resistant prostate cancer drug target.

Phytochromes are photoreceptors responsible for sensing light in plants, fungi and bacteria. Here the authors use computational simulations to reveal the molecular mechanism of photoactivation and characterize the involved reaction intermediates.

Computer-aided design of protein-ligand binding is important for the development of novel drugs. Here authors present an approach to use the recently re-parametrized coarse-grained Martini model to perform unbiased millisecond sampling of protein-ligand binding interactions of small drug-like molecules.

Protein-ligand unbinding processes are out of reach for atomistic simulations due to time-scale involved. Here the authors demonstrate an approach relying on dissipation-corrected targeted molecular dynamics that enables to provide binding and unbinding rates with a speed-up of several orders of magnitude.

Neural Networks are known to perform poorly outside of their training domain. Here the authors propose an inverse sampling strategy to train neural network potentials enabling to drive atomistic systems towards high-likelihood and high-uncertainty configurations without the need for molecular dynamics simulations.

Reconstructing system dynamics on complex high-dimensional energy landscapes from static experimental snapshots remains challenging. Here, the authors introduce a framework to infer the essential dynamics of physical and biological systems without need for time-dependent measurements.

Here, the authors apply a neural relational inference model to infer dynamic networks of interacting residues in protein molecular dynamics simulations. The model can predict allosteric communication pathways and relative free energy changes upon mutations.

Cells are complex systems that make decisions biologists struggle to understand. Here, the authors use neural networks to approximate the solution of mathematical models that capture the history and randomness of biochemical processes in order to understand the principles of transcription control.

Uncertainty-aware machine learning models are used to automate the training of reactive force fields. The method is used here to simulate hydrogen turnover on a platinum surface with unprecedented accuracy.

An E(3)-equivariant deep learning interatomic potential is introduced for accelerating molecular dynamics simulations. The method obtains state-of-the-art accuracy, can faithfully describe dynamics of complex systems with remarkable sample efficiency.

Current machine-learned force fields typically ignore electronic degrees of freedom. SpookyNet is a deep neural network that explicitly treats electronic degrees of freedom, closing an important remaining gap for models in quantum chemistry.

In machine learning approaches relevant for chemical physics and material science, neural network potentials can be trained on the experimental data. The authors propose a training method applying trajectory reweighting instead of direct backpropagation for improved robustness and reduced computational cost.

High-level ab initio quantum chemical methods carry a high computational burden, thus limiting their applicability. Here, the authors employ machine learning to generate coupled-cluster energies and forces at chemical accuracy for geometry optimization and molecular dynamics from DFT densities.

Increasing the non-locality of the exchange and correlation functional in DFT theory comes at a steep increase in computational cost. Here, the authors develop NeuralXC, a supervised machine learning approach to generate density functionals close to coupled-cluster level of accuracy yet computationally efficient.

Simulating ultrafast quantum dissipation in molecular excited states is a strongly demanding computational task. Here, the authors combine tensor network simulation, entanglement renormalisation and machine learning to simulate linear vibronic models, and test the method by analysing singlet fission dynamics.