Extending the Generality of Molecular Dynamics Simulations on a Special-Purpose Machine

Abstract

Abstract-Special-purpose computing hardware can provide significantly better performance and power efficiency for certain applications than general-purpose processors. Even within a single application area, however, a special-purpose machine can be far more valuable if it is capable of efficiently supporting a number of different computational methods that, taken together, expand the machine's functionality and range of applicability. We have previously described a massively parallel special-purpose supercomputer, called Anton, and have shown that it executes traditional molecular dynamics simulations orders of magnitude faster than the previous state of the art. Here, we describe how we extended Anton's software to support a more diverse set of methods, allowing scientists to simulate a broader class of biological phenomena at extremely high speeds. Key elements of our approach, which exploits Anton's tightly integrated hardwired pipelines and programmable cores, are applicable to the hardware and software design of various other specialized or heterogeneous parallel computing platforms.

Gauge fields, strings, solitons, anomalies, and the speed of life

“…In particular, perestroika's prompt topological solitons such as (43) to come and go along a string. Such processes are commonplace whenever we have an energy function of the form (44). It appears that in the case of strings that model proteins, we always do.…”

“…The equations are numerically integrated with a time step around a femtosecond, which is the characteristic time scale of peptide bond vibrations. Using purpose built supercomputers like Anton [43,44] and distributed computing projects like folding@home [45], the speediest MD simulations can reach a few micro-seconds of in vivo folding time, in a day in silico [44]. This enables the modeling of relatively short and very fast folding proteins such as villin headpiece (HP35) and the λ-repressor protein (1LMB in Protein Data Bank PDB [40]), up to time scales that it takes for these proteins to fold.…”

“…For example, myoglobin which is a protein described in all biochemistry textbooks, has 154 residues and folds in about 2.5 seconds [46]. Thus, running at top speed of a few microseconds per day it should take Anton over 1.000 years to fold a myoglobin [44]. And Anton is by far the fastest special-purpose MD machine ever built.…”

Virtual substitution scan via single‐step free energy perturbation

“…With high performance clusters and special-purpose machines, simulations up to hundreds of nanoseconds are performed routinely, and reports of microsecond to millisecond simulations are no longer uncommon. 33,34 While free energy calculation can directly benefit from such an increase in sampling time, most of the reported long trajectories are the so-called "vanilla" Molecular Dynamics (MD) simulations, i.e., they do not employ any specially constructed Hamiltonian, enhanced sampling scheme, or free energy calculation technique.…”

The fast multipole method and point dipole moment polarizable force fields

“…The development of polarizable force-fields represents an important step in molecular modeling, as such fields have been shown to greatly improve the description of complex microscopic systems, such as physical interfaces and those involving charged species (Chang and Dang 2006, Cieplak et al 2009, Dang 2000, Jungwirth and Tobias 2006, Lopes, Roux, and MacKerell Jr 2009, Réal et al 2013, Salanne and Madden 2011. Polarization is also suspected to greatly improve the description of amino-acid interactions, which will lead to further promising computational techniques devoted to theoretically investigations of protein folding (Scarpazza et al 2013).…”