(a) A virus (HIV seen schematically here) is a hierarchical free energy-minimizing structure wherein macromolecules with their internal structure are organized into a composite. (b) Three-dimensional structure of Poliovirus capsid showing 713,580 atoms in CPK coloring predicted using VirusX.

VirusX ^®: a multi-level virus simulator

A quantitative model of the structure and dynamics of a virus, or other macromolecular complex, is being developed to meet the drug discovery and treatment optimization challenges for risk assessment in light of emerging strains and for fundamental studies in virology. All-atom, molecular simulations have played an important role in drug discovery by giving insight into the selective binding of prospective drug compounds to target sites. The large number of atoms in a virus and its immediate surroundings has rendered these approaches computationally unfeasible for assessing the effect of compounds on the overall functioning and self-assembly of a virus. The objective of our VirusX project is to develop new computational methods for addressing this challenge.

The magnitude of the virus modeling challenge is suggested in the figure. Viruses contain millions to hundreds of millions of atoms. An approach must be used that adequately accounts for the effects that local atomic-scale organization has on the overall structure of the virus and, conversely, how overall structure affects local phenomena such as the binding of a drug molecule to a particular site. We have developed and implemented a variety of advanced molecular modeling techniques that simultaneously account for atomic-scale processes and overall structural changes to simulate these many-atom systems. VirusX is being used to study the assembly and stability of partial and whole viruses. We also integrate this type of detailed model with our field theory of the host medium in which the virus resides. We are also developing mixed mesoscopic models wherein the capsid or other viral features can be described by continuous variables or wherein capsid protomers are treated as lumped objects.

VirusX addresses the following:

• the self-assembly of a mature virus from its constituent RNA or DNA and associated enzymes and proteins or lipids;
• the penetration of a virus through the host cell outer membrane;
• the influence of mutated proteins on viral self-assembly or other behavior;
• the effect of a change in the host medium on the success of the virus; and
• the role of drugs in the above.

VirusX integrates the following techniques to determine energy-minimizing structure and the dynamics of fluctuations or self-assembly. The virus is described by the location of its N atoms . The Born-Oppenhiemer N-body potential U is minimized via a gradient method wherein a Langevin equation of the form

      (1)

The force is computed from the CHARMM22 formulas. To accelerate the computation of the interatomic forces, a tree code is implemented. Long range contributions to the forces (i.e. Lenard-Jones and Coulomb) are computed via a multi-pole method adapted specially for charged systems. The option to use the GRAPE hardware for accelerating force computations on an IBM SP is also built into VirusX.

This space warping method is based on the notion that large scale deformations (e.g. translation, bending, twisting and compression) do not change nearest-neighbor distances appreciably. Thus collective variables are introduced; as the numbers of these overall variables is much less than 3N, minimization can be accomplished quickly. As this constitutes a constrained minimization, direct simulation of (1) is alternated with the collective coordinate reduction. This method also greatly enhances the simulation of viral self-assembly as each constituent macromolecule is given its own set of collective variables.

The multi-pole method reduces the force calculations to be N ln N and similar accelerations can be achieved with the space-warping method. A two-time method is also introduced wherein long-range effects are not computed every timestep. Finally, the entire algorithm is parallelized in VirusX to make supra-million atom simulation feasible.

For a modeling methodology to be practical, it should be integrated with experimental approaches (e.g. X-ray crystallography, NMR, quantum dot, mass spectroscopy of tryptically digested fragments, SEM or cross-section measurements). In this way, an approach is realized for integrating viral modeling with this data to both guide the virus simulations and advance the interpretation and resolution of these laboratory measurements. In effect, this approach results in a methodology that makes laboratory and computer modeling a single activity. As there is uncertainty in the experimental data and the models, the natural framework to carry out this data/model integration is statistical in character. This allows for the assessment of uncertainty in all predictions. Such an approach is being developed by the us using information theory.

The all-atom features of VirusX can be used through this website. Other features will be made available over the next several months.