GPUGRID

Experiments

Our experiments aim at understanding the function of proteins and other related molecular events.

We recommend going through the Background section first if you aren't familiar yet with most basic concepts in molecular biology and molecular dynamics simulations.

Molecular simulations of the D2 Dopamine receptor under physiological ionic strength conditions

Sodium ions have been shown to play an important role in the binding of antipsychotic drugs to the D2 Dopamine receptor. Understanding the sodium-induced mechanism is of great interest for future drug design in the treatment of schizophrenia.

By means of molecular dynamics we simulate the mobility of sodium ions and its effect on the dynamic properties of the D2 receptor under physiological ionic strength conditions.

GPUGRID technology allows us to handle an all-atom system in which D2 receptor is embedded in a membrane bilayer with a total of 61,000 atoms and accelerates the computational performance up to microseconds.

Jana Selent (UPF)

Molecular simulations of the SH2 and ligand peptide binding affinity

The SH2 is a protein domain involved in protein-protein interactions. This particular domain plays a major role in cell communication on the sigalling processes for cell growth and development. However, the end goal for running such simulations is not to expand the knowldege on this particular system, but to use it as a model for developing methods to calculate protein-protein binding affinities.

Such methods will be very useful, for example, in the study of why certain wrong forms of proteins stop interacting with other partner proteins, as a way to give explanation to diseases in which these sort of mechanisms occur.

We are now gradually giving workunits that compute these binding affinities only for GPU clients.

Molecular simulations of Triose Phospate Isomerase (TPI) enzymes

This workunit investigates the conformational differences that arise when TPI enzymes undergo tyrosine nitration as a consequence of inflammation, defective mithocondrial respiration and oxydative stress. Workunits last around one day and must completed before 4 days.

Forward-Reverse Steered Molecular Dynamics

Potassium ion permeation in Gramicidin A We are giving workunits comprising full-atom simulations of gramidicin A for ion transport, a total of 30,000 atoms. Each workunit lasts less than one day and you have to complete it before 4 days.

Background

This section contains basic concepts and definitions on Molecular Biology and Molecular Dynamics simulations.

Basics of Molecular Biology

To help contextualize a bit our work within the molecular biology field, we have gathered some videos that show, at different scales, the functioning of the minimal living organism: the Cell. Unlike a cellular biologist for example, who studies the functioning of the cell at the microscopic level by culturing living cells, we are focused on the atomic scale of the cellular function.
Atoms are the building blocks of molecules and proteins are a kind of molecules -biomolecules we could say-.
By means of understanding what means 'functioning' in a cellular context, we can say that a cell is very much like a factory whose workers are proteins. Proteins have their specific tasks and they interact with other workers to perform them. In this way, proteins are involved in all kind of functions in the cell such as catalyzing reactions, giving structure, transmitting signals, providing adhesion, etc.
But what happens if they go wrong? When these proteins misbehave or simply stop functioning, the processes in which they are involved are affected. Whenever this happens we talk of 'disease', and all diseases will have a 'molecular explanation', a cause in one or several wrong molecular mechanisms.

Therefore, the study of protein function at the atomic scale can help providing explanations about the molecular mechanisms of proteins and other biomolecules in health and disease. This is crucial to design strategies that might lead to treatments for diseases for instance.

Nowadays, the only why to understand how proteins function at the atomic detail is simulatinig them on computers. And we do so using Molecular Dynamics Simulations techniques.

What is a Cell?

Central Dogma: "From DNA to Proteins"

General Cell Biology

Basics of Molecular Dynamics

Molecular dynamics (MD) is a form of computer simulation in which atoms and molecules are allowed to interact for a period of time by approximations of known physics, giving a view of the motion of the atoms. Because molecular systems generally consist of a vast number of particles, it is impossible to find the properties of such complex systems analytically; MD simulation circumvents this problem by using numerical methods. It represents an interface between laboratory experiments and theory, and can be understood as a "virtual experiment". MD probes the relationship between molecular structure, movement and function. Molecular dynamics is a multidisciplinary method. Its laws and theories stem from mathematics, physics, and chemistry, and it employs algorithms from computer science and information theory. It was originally conceived within theoretical physics in the late 1950s, but is applied today mostly in materials science and biomolecules. More

Source: Wikipedia.

Performance

Features

The main feature of GPUGRID is the brute force that it allows in terms of computational power by using accelerator processors, not only as an aggregate, but also as individual volunteered machines. This level of granularity is a fundamental requirement for all-atom molecular simulations as it permits to use a wide range of protocols on a very volatile grid. For instance, running MD on a single graphics device allows to execute trajectories of up to 4 ns/day on molecular system of 60,000 atoms. The practical way of use is therefore equivalent from a scientific perspective to that of a large personal supercomputer for molecular systems up to 100,000 atoms.

Goals

We aim at performing all-atom high-throughput molecular dynamics simulations. In particular, using the large computational power, we are setting up adapted thermodynamic protocols which work on the underlying distributed infrastructure to compute free energies for protein-ligand and protein-protein interactions and conformational sampling. These quantities offer the possibility to support experimentation by a better understanding of the molecular mechanisms of interaction and to help biomedical research by providing a way for accurate virtual screening.

Computing performance

Updated on March, 2009

GPUGRID is currently experiencing a tremendous growth in computational performance. In molecular simulations experiments, computational performance can be measured as the total amount of simulated time per day. In our case it is currently between 4-5 microseconds [figure above] on systems of the order of 50,000 atoms.

While the volunteered computing power increases, we are also working to optimize the simulation protocols in order to make the most of it.

On accelerator processors

The computational power of graphics hardware is much higher than standard processor due to the large amount of calculations which are required to display realistic three-dimensional images.

This is achieved by a design and programming style change (see CUDA). A conventional processor dedicates a relatively large fraction of its transistors to complex control logic, to maximise performance of a serial code. The Cell processor contains 8 fast Synergetic Processing Elements (SPEs) designed to maximise arithmetic throughput. Graphical processing units (GPUs) devices have a very large number of slower cores maximizing parallel throughput.

For example, just ten thousand volunteers provides the computational power of the largest supercomputer on the planet for certain specific applications.