FDiscover



2       Application--The Insight Environment

This section presents the basics for using FDiscover in the Insight environment. It includes:

The Discover module in Insight provides an interface to the FDiscover program, which is invoked when you select Discover from the Module pulldown (by clicking the MSI logo). Several new pulldowns appear on the lower menu bar: Constraint, Parameters, Run, and Pseudo_Atom, which contain the FDiscover commands.

The Constraint, Parameters, Run, and Pseudo_Atom pulldowns contain commands that you use to set up and run an FDiscover job. After you execute these commands, the results can be viewed using commands available in Insight.

As with all commands in Insight, each Discover command has equivalent typed and pulldown menu forms. Typed commands are issued from the keyboard at the Command: prompt located near the bottom of the Insight window. Typed commands can also be placed in a file that can be executed using the Source_File command in the File pulldown. Commands are typed at the Command: prompt in the format Command Pulldown or Command, as indicated by the title at the top of the corresponding parameter block. (However, for documentation purposes, the format is Pulldown/Command, since that is the order in which items are selected when using the mouse.)


Command summary

For complete descriptions of the functionality of FDiscover pulldowns and commands, please refer to Insight's online help facility. (For descriptions of the standalone commands, please refer to Commands and DSL--Standalone Mode.)


Outline of using FDiscover within Insight

When you use FDiscover in the Insight environment:

  1. To set up and submit a Discover job, use commands that are contained in these pulldowns: Constraint, Parameters, Run, and Pseudo_Atom.

  2. To define special conditions under which to perform a simulation, such as tethering, template forcing, or defined torsion angles, use commands in the Constraint pulldown.

  3. To define pseudoatoms--atom-like entities whose positions are the average positions of sets of real atoms--for use with distance constraints, use commands in the Pseudo_Atom pulldown

  4. To set values such as temperature or type of minimizer for a minimization, dynamics run, free energy calculation, or dielectrics, use commands in the Parameters pulldown.

  5. To submit, resume, or stop a job or to control which files are saved from a job or what type of information is displayed during a job, use commands in the Run pulldown.

  6. To view the results of a Discover job, use commands in the Viewer module of Insight (on the upper menu bar and icon bar) and in the Analysis or DeCipher module (see the Module pulldown, i.e., the MSI logo).


Methodology

This section outlines the basic steps for using commands in the Discover module, in the order in which they are typically used. For more detailed discussions of each command, refer to Insight's online help.

Most Discover calculations consist of three main steps:

  1. Setting up how the energy is to be calculated.

  2. Running the calculation.

  3. Analyzing the results.

Setting up the calculation

Discover calculations can generally be run on any model that has been created by or can be read into Insight . The model, however, must first be prepared for simulation by properly assigning potential types and charges to each atom. This is done with the Forcefield/Potentials command, which is accessed from the Biopolymer or Builder modules of Insight .

The Potentials command is used to check and modify the potential types. If an error message appears stating, for instance, that open or undefined valences exist, commands in the Modify pulldown (which is accessed from the Biopolymer or Builder module) are used to change bond orders or add hydrogens or to respond to any other error messages before trying again to fix potentials.

The Potentials command is also used for setting the partial charges. Insight assigns partial charges to each atom in the model based on bond increments, which are found in the forcefield file.

Types of calculations

Discover can be used for two fundamentally different types of calculations: minimization and molecular dynamics. The type of calculation is specified by toggling the Run_Minimization and Run_Dynamics parameters of the Run/Run command of the Discover module. If both options are turned on, minimization is performed before dynamics. The various forms of minimization provide a static description of the model, and molecular dynamics provides a simulation of the motion of atoms at a given temperature, etc.

Minimization

Minimization of a model is done in two steps. First, an equation describing the energy of the model as a function of its coordinates must be defined and evaluated for a given conformation. Next, the conformation is adjusted to the lower value of the target function. A minimum may be found after one adjustment or after many thousands of iterations, depending on the nature of the algorithm, the form of the target function, and the size of the model. The efficiency of minimization is therefore affected by both the time to evaluate the target function and the number of structural adjustments (iterations) needed to converge on the minimum.

Minimization is specified with the Run_Minimization option of the Run/Run command, which is the default calculation type. The minimization algorithm and other parameters are controlled with the Parameters/Minimize command. This command provides a choice of several types of minimizers, each suited for somewhat different situations.

Which algorithm to use depends on two factors--the size of the model and its current state of optimization. Until the derivatives are well below 100 kcal mol-1 Å-1, it is likely that the structure is outside the quadratic region of the potential energy surface. Algorithms that assume the energy surface to be quadratic (Newton-Raphson, VA09A [BFGS], and conjugate gradients) can be unstable in this situation. Therefore, as a general rule, steepest descents is often the best minimizer to use for the first 10-100 steps, after which conjugate gradients or a Newton-Raphson minimizer can be used to complete the minimization to convergence.

Other options in the Parameters/Minimize command include specification of the convergence criteria, such as the number of Iterations or the maximum Derivative desired. Note that the default operation is to perform a steepest-descents minimization for 100 iterations or until the rms derivative falls below 0.05 kcal mol-1 Å-1, whichever occurs first.

The Charges option can be included in the calculation if charges have been properly assigned. If Charges is turned off, Coulombic interactions are not included in the potential energy expression.

The Cross option means to include cross terms in the calculation when the CVFF forcefields are used and has no effect when other forcefields are used. The cross terms represent couplings between deformations of internal coordinates. For example, the CVFF forcefield includes a term representing the coupling between adjacent bonds--a bond-bond term--as well as a term describing the interaction between an angle and the bond forming one side of the angle--a bond-angle term. Such cross terms are important for accurately reproducing experimental properties such as vibrational frequencies, but can cause problems during the initial refinement of highly distorted structures. Therefore, when starting with a poor initial structure (derivatives > 100 mol-1 Å-1), you should use 10-100 steps of steepest-descents minimization without cross terms or Morse bond terms, to correct the major distortions. Once the structure is reasonable, you may then include cross terms, use Morse bond terms, and switch to other, more rapidly convergent minimizers.

The Morse option is used to specify the use of a Morse potential for bond stretching rather than the default harmonic potential, when CVFF is used. For other forcefields this option has no effect. The Morse potential is a better approximation of the true bond-stretching potential, but it can lead to computational problems for unoptimized structures having very short (< 0.5 Å) or very long (> 3 Å) bonds. Therefore, as with cross terms, the Morse potential should be used only with reasonably optimized structures.

Dynamics

Molecular dynamics simulations using Discover in the Insight environment are also done in a two-step process. The first stage is to describe the job to be performed and the second is to run the calculation.

A dynamics run has two distinct phases: the initialization or equilibration phase and the resume or data-collection phase. During initialization, the model is raised to and equilibrated at the target temperature. During the resume phase, the temperature is constant and data are collected for analysis.

Dynamics is set up using the Parameters/Dynamics command. Options in this command let you specify Equilibration (to define how many iterations are used to reach and equilibrate at the target temperature) and Steps (to define how many iterations are performed once the temperature is achieved). Increasing the default Time Step of 1 fs (the time interval between each iteration) is not recommended, although in very-high-temperature dynamics, it may be necessary to shorten the Time Step to prevent breakdown of the numerical integration, which would cause the run to stop due to a sudden large increase in the energy of the model.

The Charges, Cross, and Morse options (described under Minimization) can be turned on or off for a dynamics run.

Using constraints and restraints

The Constraint pulldown contains commands for tailoring the energy calculation to fit specific needs. The Constraint/Fix command freezes specified atoms at their initial location for the duration of a calculation. Not only is this useful for the obvious purpose of keeping some atoms fixed, but fixing atoms can often cut computational costs significantly. The calculated absolute energy is arbitrary, depending on the forcefield and the model topology. Only relative energies of models with the same number of bonds, etc. are meaningful. Because the terms for energies between fixed atoms sum to a constant, which cancel in any relative-energy calculation, Discover ignores all terms between fixed atoms. If you are working with a large model but are interested in only a small part of it, you might fix all the atoms or residues not involved with the interesting part. The moving atoms still feel the effect of the fixed atoms--the bonds, angles, etc., as well as the van der Waals and Coulombic interactions, are still accounted for. However, if 80% of the atoms are fixed the calculation is about 5 times faster.

Several other commands in the Constraint pulldown add extra terms to the energy expression to force or restrain particular coordinates to given values. The GenericDis command applies a flat-bottomed potential to restrain a pair of atoms to a minimum and maximum separation distance. Particularly designed for incorporating experimental distance information extracted from nuclear Overhauser effect (NOE) NMR experiments, NOE distance restraints can be applied to pseudoatoms created with the Pseudo_Atom pulldown. Although originally designed for NOE distance information, these distance restraints are useful for a multitude of other situations where distance information can be inferred or where you want to ensure that certain distances are maintained in your model.

Two more commands, TemplateForce and Tether, are related to each other. Constraint/TemplateForce is used to force the best rms match between the current model and a template. You set up a one-to-one correspondence between atoms in the template and the moving model and add an energy term proportional to the rms fit between the two lists. By running dynamics or minimization with this forcing term added, you can search for conformations that closely match the template with only a minimum of excess steric or strain energy. The Constraint/Tether command sets up a similar calculation, except that the template is the initial structure of the same model, rather than perhaps a different template model. Tethering is useful, for instance, to restrain the backbone of a model when modifying the sidechains, in order to maintain the gross structure of the model while allowing the sidechains to rearrange.

The last command that adds an explicit term to the energy expression is Constraint/TorsionForce, which adds a harmonic forcing function to one or more dihedral angles. This added term tends to force the dihedral angle to the given value and is used, for example, to induce a conformational change or explore a rotational barrier.

A related command, Constraint/Rotor, causes Discover to systematically increment a torsion angle and evaluate the energy at each point. This command is useful for systematically searching conformations to construct an energy map.

The final command in the Constraint pulldown is Warp, which is used to set up relative free energy calculations. Relative free energies involve two chemically distinct systems. The Warp command is used to map atomic properties, including forcefield atom types, charges, and masses, from one system to the other.

Other parameters

Details of the calculation of nonbond interactions--the van der Waals and Coulombic interactions--are controlled by the Parameters/Set and Parameters/Variables commands. The Set command allows you to control the value of the dielectric constant used for the Coulombic interactions, as well as to switch the form to a distance-dependent dielectric "constant" if desired. The Variables command gives access to the variables controlling the nonbond list. For large systems it becomes impractical to include all pairwise nonbond interactions--there are simply too many. Discover allows you to truncate the nonbond interactions at a cutoff distance specified in the Variables command. The Variables command controls the temperature and pressure baths used for constant-temperature and constant-pressure dynamics.

The Parameters/Scale_Terms command can be used to individually scale various contributions to the energy. One important use of these scale factors is for various types of simulated annealing. You can reduce selected terms, such as the torsions, to allow the model to efficiently explore various sections of conformational space. The Scale_Terms command contains one special scale factor for the 1-4 nonbond interactions. Some forcefields, notably AMBER, scale these 1-4 interactions separately from the other nonbond interactions.

Analyzing the results

Insight provides many commands that are useful for analyzing structural aspects of a model, such as distances and angles, for comparing models using rms matching, etc. whether the models are the results of a Discover calculation or created in some other fashion. Although it is beyond the scope of this section to describe these commands, some comments about Insight features particularly related to Discover are presented here. Please refer to the Insight documentation for more details.

The primary result of most Discover calculations is a new structure. As long as the calculation was started from the current Insight session, Insight automatically reads the final structure of a Discover calculation when it finishes. When you start the Discover job from the Run/Run command, you may set the Computation Mode to Interactive, which means the model, along with an energy summary, is updated continuously on the screen as the calculation proceeds; or Batch, which means Insight updates the structure only once, at completion of the Discover job.

Running the job interactively allows you to view the model's trajectory once as the calculation proceeds. During dynamics runs, Discover periodically writes a snapshot of the current structure, energy, temperature, and pressure to the history file. Insight can then be instructed to play back the resulting trajectory using the Trajectory/Animate command in the Analysis module. Insight also provides extensive capabilities to create graphs from the information in the history file, via the commands in the Graph pulldown (available as an icon).




Last updated September 27, 1997 at 11:14AM PDT.
Copyright © 1997, Molecular Simulations, Inc. All rights reserved.