Simulating Protein-DNA recognition: distributed molecular dynamics (Application)
Conducted byBenjamin Bouvier
DescriptionAdvancing our knowledge of the function of living organisms is essential, for instance in the field of medical research. It requires a thorough understanding of the mechanisms of the living cell and the properties of the macromolecules (proteins, nucleic acids) that are responsible for them. These molecules interact according to very complex and self-regulated networks, forming large edifices (called complexes) in which each individual partner plays its own role in the collective function, very much like individual parts in an engine. The recognition, assembly, and collective function of macromolecular complexes is based (in a non-trivial fashion) on the properties of their constituent macromolecules (mechanical, electrostatic, thermodynamic), whose u nderstanding requires the study of these systems at the atomic level. This is by no means a simple task: large multi-partner edifices such as the ribosome or the nucleosome can contain several million atoms; the necessit y to consider the cellular environment (water, ions) which often acts as an additional partner adds further complexity to the problem. The last two decades have seen the rise of molecular dynamics, a very powerful computational technique which allows the simulation of biological macromolecules and has proved an invaluable complement to experimental stud ies, especially in areas where experimental techniques experience difficulties. Molecular dynamics uses a force field to describe the interaction between the individual atoms of a macromolecule and compute the resulting forces; it allows the study of the evolution of the system over time, via the integration of the corresponding equations of motion. It yields very accurate insights into the function of complex biological systems, over t imescales which depend on the computational resources available and the size of the system under study, typically ranging from a few nanoseconds to a microsecond. Today, molecular dynamics methods and force fields have reached a good level of maturity. The challenge is now to be able to simulate systems of growing sizes (up to several million atoms) over timescales that are suffic iently long to observe relevant biological processes - and these tend to become longer as the system size grows. Considering that molecular dynamics is, in essence, highly parallelizable, the use of large-scale distribut ed computing environments such as Grid5000 is very appealing to this effect. We will deploy two major academic open-source molecular dynamics packages: GROMACS (http://www.gromacs.org) and NAMD (http://www.ks.uiuc.edu/Re search/namd), and test the performance and scaling of each under different conditions: varying number of nodes, varying repartition of atoms on the different nodes, different message-passing frameworks (MPI, Charm++)... We hope to provide guidelines for the rational use of distributed computer resources for molecular dynamics simulation, whose importance is expected to increase in years to come. In addition, we will test on the grid novel simulation methodologies which we have implemented on the basis of the GROMACS and NAMD packages. These include enhanced sampling and/or directed molecular dynamics, that give access to timescales normally not accessible to standard molecular dynamics (provided certain assumptions are made), and free energy calculation techniques. The biological systems that we will study are instances of impo rtant recognition processes between proteins and DNA; they range from small complexes (SRY/DNA, involved in sexual differenciation) to much larger edifices (the nucleosome, involved in the packaging of DNA inside the cell nucleus, and the target of numerous enzymes). It is our hope that the computing power provided by the Grid5000 testbed will help advance our understanding in this very active field of research.
Tools usedNAMD GROMACS
Shared by: Benjamin Bouvier
Last update: 2008-07-11 18:01:49