Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently—those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors. The algorithms are tested on a standard Lennard-Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers--the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y-MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventional vector supercomputers even for small problems. For large problems, the spatial algorithm achieves parallel efficiencies of 90% and a 1840-node Intel Paragon performs up to 165 faster than a single Cray C9O processor. Trade-offs between the three algorithms and guidelines for adapting them to more complex molecular dynamics simulations are also discussed.

Fast parallel algorithms for short-range molecular dynamics

http://kth.diva-portal.org/smash/get/diva2:1303357/FULLTEXT01

GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers

Motivation: Molecular simulation has historically been a low-throughput technique, but faster computers and increasing amounts of genomic and structural data are changing this by enabling large-scale automated simulation of, for instance, many conformers or mutants of biomolecules with or without a range of ligands. At the same time, advances in performance and scaling now make it possible to model complex biomolecular interaction and function in a manner directly testable by experiment. These applications share a need for fast and efficient software that can be deployed on massive scale in clusters, web servers, distributed computing or cloud resources.

Results: Here, we present a range of new simulation algorithms and features developed during the past 4 years, leading up to the GROMACS 4.5 software package. The software now automatically handles wide classes of biomolecules, such as proteins, nucleic acids and lipids, and comes with all commonly used force fields for these molecules built-in. GROMACS supports several implicit solvent models, as well as new free-energy algorithms, and the software now uses multithreading for efficient parallelization even on low-end systems, including windows-based workstations. Together with hand-tuned assembly kernels and state-of-the-art parallelization, this provides extremely high performance and cost efficiency for high-throughput as well as massively parallel simulations.

Availability: GROMACS is an open source and free software available from http://www.gromacs.org.

Contact: erik.lindahl@scilifelab.se

Supplementary information:Supplementary data are available at Bioinformatics online.

Gromacs 4.5

Many biologically interesting phenomena occur on a time scale that is too long to be studied by atomistic simulations. These phenomena include the dynamics of large proteins and self-assembly of biological materials. Coarse-grained (CG) molecular modeling allows computer simulations to be run on length and time scales that are 2–3 orders of magnitude larger compared to atomistic simulations, providing a bridge between the atomistic and the mesoscopic scale. We developed a new CG model for proteins as an extension of the MARTINI force field. Here, we validate the model for its use in peptide-bilayer systems. In order to validate the model, we calculated the potential of mean force for each amino acid as a function of its distance from the center of a dioleoylphosphatidylcholine (DOPC) lipid bilayer. We then compared amino acid association constants, the partitioning of a series of model pentapeptides, the partitioning and orientation of WALP23 in DOPC lipid bilayers and a series of KALP peptides in dimyris...

The MARTINI Coarse-Grained Force Field: Extension to Proteins

The widespread use of molecular machines in biology has long suggested that great rewards could come from bridging the gap between synthetic molecular systems and the machines of the macroscopic world. In the last two decades, it has proved possible to design synthetic molecular systems with architectures where triggered large amplitude positional changes of submolecular components occur. Perhaps the best way to appreciate the technological potential of controlled molecular-level motion is to recognize that nanomotors and molecular-level machines lie at the heart of every significant biological process. Over billions of years of evolution, nature has not repeatedly chosen this solution for performing complex tasks without good reason. When mankind learns how to build artificial structures that can control and exploit molecular level motion and interface their effects directly with other molecular-level substructures and the outside world, it will potentially impact on every aspect of functional molecule and materials design. An improved understanding of physics and biology will surely follow.

The first steps on the long path to the invention of artificial molecular machines were arguably taken in 1827 when the Scottish botanist Robert Brown observed the haphazard motion of tiny particles under his microscope.1,2 The explanation for Brownian motion, that it is caused by bombardment of the particles by molecules as a consequence of the kinetic theory of matter, was later provided by Einstein, followed by experimental verification by Perrin.3,4 The random thermal motion of molecules and its implications for the laws of thermodynamics in turn inspired Gedankenexperiments (“thought experiments”) that explored the interplay (and apparent paradoxes) of Brownian motion and the Second Law of Thermodynamics. Richard Feynman’s famous 1959 lecture “There’s plenty of room at the bottom” outlined some of the promise that manmade molecular machines might hold.5,6 However, Feynman’s talk came at a time before chemists had the necessary synthetic and analytical tools to make molecular machines. While interest among synthetic chemists began to grow in the 1970s and 1980s, progress accelerated in the 1990s, particularly with the invention of methods to make mechanically interlocked molecular systems (catenanes and rotaxanes) and control and switch the relative positions of their components.7−24

Here, we review triggered large-amplitude motions in molecular structures and the changes in properties these can produce. We concentrate on conformational and configurational changes in wholly covalently bonded molecules and on catenanes and rotaxanes in which switching is brought about by various stimuli (light, electrochemistry, pH, heat, solvent polarity, cation or anion binding, allosteric effects, temperature, reversible covalent bond formation, etc.). Finally, we discuss the latest generations of sophisticated synthetic molecular machine systems in which the controlled motion of subcomponents is used to perform complex tasks, paving the way to applications and the realization of a new era of “molecular nanotechnology”.

1.1. The Language Used To Describe Molecular Machines
Terminology needs to be properly and appropriately defined and these meanings used consistently to effectively convey scientific concepts. Nowhere is the need for accurate scientific language more apparent than in the field of molecular machines. Much of the terminology used to describe molecular-level machines has its origins in observations made by biologists and physicists, and their findings and descriptions have often been misinterpreted and misunderstood by chemists. In 2007 we formalized definitions of some common terms used in the field (e.g., “machine”, “switch”, “motor”, “ratchet”, etc.) so that chemists could use them in a manner consistent with the meanings understood by biologists and physicists who study molecular-level machines.14

The word “machine” implies a mechanical movement that accomplishes a useful task. This Review concentrates on systems where a stimulus triggers the controlled, relatively large amplitude (or directional) motion of one molecular or submolecular component relative to another that can potentially result in a net task being performed. Molecular machines can be further categorized into various classes such as “motors” and “switches” whose behavior differs significantly.14 For example, in a rotaxane-based “switch”, the change in position of a macrocycle on the thread of the rotaxane influences the system only as a function of state. Returning the components of a molecular switch to their original position undoes any work done, and so a switch cannot be used repetitively and progressively to do work. A “motor”, on the other hand, influences a system as a function of trajectory, meaning that when the components of a molecular motor return to their original positions, for example, after a 360° directional rotation, any work that has been done is not undone unless the motor is subsequently rotated by 360° in the reverse direction. This difference in behavior is significant; no “switch-based” molecular machine can be used to progressively perform work in the way that biological motors can, such as those from the kinesin, myosin, and dynein superfamilies, unless the switch is part of a larger ratchet mechanism.14

Artificial Molecular Machines

In biological systems, proteins can be attracted to curved or stretched regions of lipid bilayers by sensing hydrophobic defects in the lipid packing on the membrane surface. Here, we present an efficient end-state free energy calculation method to quantify such sensing in molecular dynamics simulations. We illustrate that lipid packing defect sensing can be defined as the difference in mechanical work required to stretch a membrane with and without a peptide bound to the surface. We also demonstrate that a peptide’s ability to concurrently induce excess leaflet area (tension) and elastic softening – a property we call the ‘characteristic area of sensing’ (CHAOS) – and lipid packing sensing behavior are in fact two sides of the same coin. In essence, defect sensing displays a peptide’s propensity to generate tension. The here-proposed mechanical pathway is equally accurate yet, computationally, about 40 times less costly than the commonly used alchemical pathway (thermodynamic integration), allowing for more feasible free energy calculations in atomistic simulations. This enabled us to directly compare the Martini 2 and 3 coarse-grained and the CHARMM36 atomistic force-fields in terms of relative binding free energies for six representative peptides including the curvature sensor ALPS and two antiviral amphipathic helices (AH). We observed that Martini 3 qualitatively reproduces experimental trends, whilst producing substantially lower (relative) binding free energies and shallower membrane insertion depths compared to atomistic simulations. In contrast, Martini 2 tends to overestimate (relative) binding free energies. Finally, we offer a glimpse into how our end-state based free energy method can enable the inverse design of optimal lipid packing defect sensing peptides when used in conjunction with our recently developed Evolutionary Molecular Dynamics (Evo-MD) method. We argue that these optimized defect sensors – aside from their biomedical and biophysical relevance – can provide valuable targets for the development of lipid force-fields. TOC Graphic

https://pubs.acs.org/doi/pdf/10.1021/acs.jctc.2c00222

Efficient Quantification of Lipid Packing Defect Sensing by Amphipathic Peptides: Comparing Martini 2 and 3 with CHARMM36

Souza et al. present the latest release of the popular and widely used force field for near-atomistic coarse-grained biomolecular simulations.

Martini 3: a coarse-grained force field with an eye for atomic detail.

Proteins can specifically bind to curved membranes through curvature-induced hydrophobic lipid packing defects. The chemical diversity among such curvature ‘sensors’ challenges our understanding of how they differ from general membrane ‘binders’, that bind without curvature selectivity. Here, we combine an evolutionary algorithm with coarse-grained molecular dynamics simulations (Evo-MD) to resolve the peptide sequences that optimally recognize the curvature of lipid membranes. We subsequently demonstrate how a synergy between Evo-MD and a neural network (NN) can enhance the identification and discovery of curvature sensing peptides and proteins. To this aim, we benchmark a physics-trained NN model against experimental data and show that we can correctly identify known ‘sensors’ and ‘binders’. We illustrate that sensing and binding are in fact phenomena that lie on the same thermodynamic continuum, with only subtle but explainable differences in membrane binding free energy, consistent with the serendipitous discovery of sensors. Teaser AI-based design helps explain curvature-selective membrane binding behavior.

/pdf/physics-based-generative-model-of-curvature-sensing-peptides-1veuedv4.pdf

Physics-based generative model of curvature sensing peptides; distinguishing sensors from binders

The mitochondrial inner membrane is an integral part of the cellular lipid biosynthesis network. Intramitochondrial lipid transfer shuttles specific lipid species between the two mitochondrial membranes. This pathway is facilitated by designated protein complexes in the intermembrane space. A hetero-dimeric complex of Ups1 and Mdm35 has been identified in yeast to transfer phosphatidic acid from the outer to the inner membrane. Here, we define pH and lipid charge dependency of Ups1 membrane interaction. We show that Ups1 interacts with membranes in a membrane curvature dependent manner. Ups1 predominantly binds to membrane domains of positive curvature. Desorption of phosphatidic acid from positively curved membranes is energetically favorable over flat or negatively curved membranes. Hence, our results demonstrate that Ups1 specifically binds to membrane regions where extraction and insertion of lipids is enhanced.

Membrane interactions of mitochondrial lipid transfer proteins

https://doi.org/10.1016/j.dib.2022.108598

H. Jelger Risselada

Papers

Efficient Quantification of Lipid Packing Defect Sensing by Amphipathic Peptides: Comparing Martini 2 and 3 with CHARMM36

Martini 3: a coarse-grained force field with an eye for atomic detail.

Physics-based generative model of curvature sensing peptides; distinguishing sensors from binders

Membrane interactions of mitochondrial lipid transfer proteins

Curvature model for nanoparticle size effects on peptide fibril stability and molecular dynamics simulation data