Important:

We have now reached the capacity for attendance in this workshop.

We will begin 8.30 AM on 12th March and finish by 1PM on 14th March.

See you in Lugano!

Molecular Dynamics (MD) simulation has seen extensive use across a variety of disciplines over the past few decades, including Materials Science, Chemistry, Biology and Earth & Planetary Sciences. With enormous improvements in computer power and in the quality of classical interaction forces, much progress has been made in our capability to study larger and more complicated systems. From the first MD simulation of 864 Argon atoms by Rahman [1] in 1964, to the first MD study of a protein with the simulation of the bovine pancreatic trypsin inhibitor (BPTI) by McCammon et al. in 1977 [2], one now routinely sees much larger, complicated structures being treated by MD simulations. For instance it is quite common now to find in the literature MD simulations of explicitly solvated proteins, protein-DNA complexes and other complicated systems [3] addressing a variety of issues concerning the thermodynamics of such systems.

THE TIME SCALE PROBLEM

In spite of its tremendous popularity, MD still has a major limitation that curtails its actual predictive power: the inability to describe dynamics at experimentally relevant time scales. Most of the interesting dynamics occurs as the system moves from one free energy basin to another through infrequent rare events, while the system remains stuck in some basin for extended periods of time. This, combined with the small time step needed for total energy conservation, severely restricts timescales accessible in MD simulations. Most MD simulations have thus tended to be in conditions that are far from realistic, making their usefulness limited. Developing algorithms that can preserve atomistic resolution yet give system dynamics for milliseconds, seconds or even longer is thus a problem of great current relevance.

Recently this problem has seen significant progress towards its solution, emanating from a variety of advanced sampling algorithms, distributed computing platforms and specialized hardware development. While there has been significant and very impressive progress in the last few decades towards recovering the free energy surfaces for such rare event systems [4-6], much more remains to be achieved in the analogous problem of recovering their long time dynamics.

The sequential nature of the problem of simulating long time dynamics makes it less amenable to raw computer power, though there have been excellent recent efforts using specialized hardware for this. As such several novel ideas have been proposed to solve the time-scale problem, all successful to their respective extents.

The focus of the proposed workshop will be on methods that can simulate long time dynamics in a limited amount of computer time and with limited (i.e. commercially viable) computer resources, while preserving atomic scale resolution. This workshop will bring together experts at a spectrum of long time-scale techniques, and facilitate discussion on where we stand and how should we proceed next.

The techniques that are now being used to simulate long time dynamics include but are not limited to:
1. Transition state theory (TST) inspired methods that involve biasing the energy landscape a priori to accelerate rare events sampling. These methods include conformation flooding [7], Hyperdynamics [8-9], Temperature Accelerated Dynamics (TAD) [10], SISYPHUS [11-12], kappa-dynamics [13] and others.

2. Accelerated MD (aMD) methods that have been used for a variety of biomolecular systems [14-15].

3. Transition Path Sampling, Transition Interface Sampling and related methods that go beyond the limitations of TST [16-19].

4. Reaction coordinate based methods such as Milestoning [20]

5. Methods centered on identifying saddle points in potential energy surface and then performing kinetic Monte Carlo scheme with the list of possible transition pathways explicitly updated on-the-fly [21-23].

6. Distributed computing methods such as Folding@home [24] and other methods exploiting parallelism such as parallel replica dynamics [25].

7. Dedicated hardware approaches such as the recent ANTON computer developed at DE Shaw Research [26].

OBJECTIVES OF THE WORKSHOP

At this workshop, we will be hosting talks and discussions by pioneers of all of the above mentioned and other approaches. We believe it is for the first time that such a diverse variety of experts at the long time scale problem will be collecting together under one roof. The workshop should thus be truly state of the art. Through interactions between these experts over 3 days, we expect the workshop to aid in synthesis of novel viewpoints, and to foster the kind of new collaborations that are needed to solve the longstanding time scale problem in MD.

A number of PhD students and young researchers will also be present at the workshop, and this should give them a unique opportunity to hear from the experts of the field. They are also welcome to present a poster.