Introduction.

This is a guide to sander, the AMBER module which carries out energy minimization, molecular dynamics, and NMR refinements. The acronym stands for Simulated Annealing with NMR-Derived Energy Restraints, but this module is used for a variety of simulations that have nothing to do with NMR refinement.

Sander provides standard protocols for minimization and molecular dynamics, and we use it for just about everything except free energy calculations. Some general features are outlined in the following paragraphs:

1. Sander provides direct support for the AMBER and AMBER/OPLS force fields for proteins and nucleic acids, and for the TIP3 and TIP4 models for water. Other types of restraints can be applied, and the code allows some variation in functional form as well as in parameters. These variations include alternate functions for "improper" torsions and Urey-Bradley interactions, so that force fields like that of version 22 of CHARMM can be supported; full implmentation of the CHARMM 22 force field is under development, and is not yet functional. Right now, "non-additive" force fields based on atom-centered dipole polarizabilities can be used, but general parameters for macromolecules are still under development. Biomoleular simulations with appropriate parameters will accurately conserve energy over multi-nanosecond runs without modification of the equations of motion; weak-coupling algorithms for pressure and temperature regulation are also available.
2. The particle-mesh Ewald (PME) procedure (or, optionally, a "true" Ewald sum) is used to handle long-range electrostatic interactions. Long-range van der Waals interactions are estimated by a continuum model; a version of PME for van der Waals terms is under development.
3. Two periodic imaging geometries are included: rectangular parallelopiped and truncated octahedron (box with corners chopped off). (Sander itself can handle many other periodically-replicating boxes, but input and output support (in LEaP and ptraj is only available right now for these two.) The size of the repeating unit can be coupled to a given external pressure, and velocities can be coupled to a given external temperature by several schemes. The external conditions and coupling constants can be varied over time, so various simulated annealing protocols can be specified in a simple and flexible manner.
4. It is also possible to carry out non-periodic simulations, either using "vacuum" potentials or with a generalized Born/ surface area model for aqueous solvation. This preserves much of the functionality of earlier versions of sander, with the following principal changes: (a) atom-based cutoffs are used, rather than residue-based cutoffs; (b) polarizable force fields are not yet supported; (c) compuational efficiency and parallel scaling is not as good for some types of simulations as in earlier versions. For those who require this earlier functionality, a slightly-modified version from Amber 5 is included; this is called sander_classic.
5. Users can define internal restraints on bonds, valence angles, and torsions, and the force constants and target values for the restraints can vary during the simulation. The penalty function can consist of as many as three types of region: it can be flat between an "inner" set of upper and lower bounds (called and ); then rise parabolically when the internal coordinate violates these bounds; and finally, since large violations may lead to excessive parabolic penalties, these parabolas can smoothly turn into linear penalties outside even wider upper and lower bounds (called and ). The imposition of restraints can be made dependent upon the distance that residues are apart in the amino-acid sequence, so that much of the functionality of programs like DISMAN or DIANA is available. The relative weights of various terms in the force field can be varied over time, allowing allowing one to implement a variety of simulated annealing protocols in a single run.
6. Internal restraints can be defined to be "time-averaged", that is, restraint forces are applied based on the averaged value of an internal coordinate over the course of the dynamics trajectory, not only on its current value. Alternatively, restraints can be "ensemble-averaged" using the locally-enhanced-sampling (LES) option.
7. Restraints can be directly defined in terms of NOESY intensities (calculated with a relaxation matrix technique), residual dipolar couplings, scalar coupling constants and proton chemical shifts. There are provisions for handling overlapping peaks or ambiguous assignments. In conjunction with distance and angle constraints, this provides a powerful and flexible approach to NMR structural refinements.

We have divided this manual into the six sections listed below.

+--------------------------------------------------------------------+
|		     Purpose			   Sections involved |
+--------------------------------------------------------------------+
|
								     |
|Simple min/md						   1	     |
|
								     |
|varying parameters over time				  1,2	     |
|(simulated annealing)						     |
|
								     |
|using internal restraints				  3,4	     |
|(including NMR distance & angle constraints)			     |
|
								     |
|nmr refinement using NOESY volume restraints		   5	     |
|
								     |
|nmr refinement using chemical shift restraints		   6	     |
|
								     |
|nmr refinement using direct dipolar splittings		   7	     |
|
								     |
|group input						   8	     |
|
								     |
+--------------------------------------------------------------------+

If you are just doing "standard" minimization or dynamics, read section one, and ignore the rest. If you want to carry out simulated annealing, consult section two. Those who wish to carry out simulations while imposing internal coordinate restraints should also read sections three and four. Sections five and six allow you to add sophisticated penalty functions during NMR refinement.