Earth Sciences & Env.
Engineering & Tech.
Information & Comm. Tech.
Life Sciences & Biotech
Use this search facility to find out more about the profile of our HPC-Europa2 visitors, the type of work they have been doing, and their project achievements.
The aims of this project were to reintroduce the Mauri self interaction correction into the Quantum ESPRESSO codebase, in order to continue to study anionic systems. Secondary aims were to complete production runs of HCl on ice surfaces both neutral and anionic in order to investigate the dissociation of the molecule at finite temperatures relevant to the atmosphere.
Additional study of the ESPRESSO codebase had shown that the correction had only been partially removed, and so was easy to utilise. Secondary achievements were that initial studies into the finite temperature dynamics of HCl on the neutral crystalline ice surface were completed. Two separate simulations at 50K both returned the information that the HCl desorbed from the ice interface and moved into the vacuum region, contrary to previous theoretical calculations. Calculations in the anionic framework were not initially undertaken.
Recent work show that algorithms originating from robotics can be the basis for the development of efficient conformational sampling and exploration methods in structural biology and materials science. Substantial efforts have been made in the field of robotics to efficiently compute, plan, simulate or control the motions of complex articulated systems. Based on this background, we have developed computationally efficient methods for sampling and exploring the conformational space of proteins. These methods are based on the ML-RRT algorithm, a variant of the Rapidly-exploring Random Tree (RRT) algorithm from the robotics field, which we have specifically developed to study computationally demanding protein-ligand interaction problems. Within this HPC-Europa2 project, we plan to improve the performance of the ML-RRT algorithm by parallelizing it on a distributed-memory architecture, and to evaluate several parallel schemes. Then, we plan to integrate these parallelized algorithms within the Protein Energy Landscape Exploration (PELE) tool developed at the Barcelona Supercomputing Center (BSC). Our objective is to conceive a new method combining the strengths of both approaches: the great exploration power of ML-RRT, and the efficient conformational energy computation of PELE.
We have proposed three parallel versions of the Rapidly-exploring Random Tree (RRT) algorithm, based on classical parallelization schemes: OR parallel RRT, Distributed RRT and Manager-worker RRT. For scalability reasons, these parallel algorithms have been designed for distributed-memory architectures, following the message passing paradigm. The OR parallel RRT is a simple algorithm based on the concurrent construction of several solutions. In the Distributed RRT and the Manager-worker RRT, a single solution is built collaboratively by all the involved processes, the difference between both schemes being in the nature of their task scheduling strategies: decentralized vs. centralized.We have evaluated these three algorithms on several molecular simulation (and more specifically protein-ligand interaction) problems. This evaluation has been performed on the MareNostrum supercomputer at the Barcelona Supercomputing Center (BSC). It has revealed that none of the algorithms can be held as an obvious best parallelization scheme for RRT: it really depends on the characteristics of the studied problem. For example, the OR parallel RRT performs well only on problems showing a great variability in runtime. The Distributed RRT globally shows good results, except when the cost of an RRT expansion is too small compared with the communication costs. The Manager-worker RRT performs well only when the computation costs are far greater than the communication costs, but in that case it outperforms the Distributed RRT.Finally, we have integrated in our application the energy computation functionality of the Protein Energy Landscape Exploration (PELE) tool developed at BSC. Our objective is now to produce more realistic molecular trajectories, not limited to geometrically feasible paths, by exploiting the efficient conformational energy computation of PELE. Besides, this integration will enable us to fully benefit from the potential of the parallel versions of RRT, since energy computation considerably increases the cost of an RRT expansion. On the other hand, exploiting the speedup from parallel computation will allow us to run more complex molecular simulations in a reasonable amount of time.
The main objectives from this study are:
The objective of our project was to study the electrostatic properties of a model charged latex particle. Electrostatic properties of the ionisable interface was investigate by performing MD simulations in systems with different conditions of ionic strength and using explicit molecules of water to model the solvent.
In our study we made simulations of different models of methane spheres at different bulk conditions. Our systems includes charged latex particle surrounded by several layers of ions and explicitly including the water solvent molecules.Different amounts of monovalent (Na+) and divalent (Ca++) counter ions. The Cl- ion was used as a coion with different concentration of these salts.
In presence work we can separate our study in two parts. First part was to constructed latex particle and second step was molecular dynamic simulation performed with GROMACS.
A model of charged latex particle was constructed using a spherical distribution of two kinds of site particles: van der Waals and electrically charged site particles.
The program GENLATEX has been written in the host group to design spherical distributions of different radius with different charge densities of the latex particle .The surface charges representing ionizable functional groups on the surface will be distributed randomly.
The system we have studied consist 942 methane molecules separated in six layers, 130954 water molecules and different amount of Na+ and Cl- ions depends of concentration of solutions (0.1M , 0.4M) and number of charges added to the system.
The amount of Na+ ions added like charge are: 2,4,16,50,100,200,300.
The protocol for simulations can be described in few steps.In this study, the molecular dynamics simulations were performed using GROMACS with ffG43a1 force field to calculate the trajectories of the system. The intermolecular potentials used in all simulations were the four-site TIP4P model of water and the fitted Lennard-Jones 12-6 potentials for CH4-H2O.For the calculation of long-range electrostatic forces,perticale-mesh Ewald (PME) method .All bond lengths were constrained usig Shake algorithm with a geometric tolerance 0.0001.The temperature was fixed by use Berendsen thermostat at 300K and the Berendsen pressure coupling algorithm was used to keep the pressure constant.Periodic boundary conditions were applied in three directions. The initializations of each MD run were done in two steps. The first step is a step- descent method perform energy minimization in order to reduce the thermal noise in the structures and potential energies. The second step consisted of a 2-ns or 5-ns time steps, depends of the system simulation at the same temperature to reach equilibrium.
MD simulations have been performed with GROMACS studying the density profiles and integrated charge distribution functions near nanoparticles bearing mobile unit charges at a variety of solution conditions. In particular, we analyzed the ability to generate particle charge inversion.
We started our study with monovalent (Na+) counterions , Cl- ion will be used as a coion. The ionic strengths was 0.4M and 0.1M. The number of ions in the simulation box was chosen to simulate the appropriate ionic ionic strength