perturb (c46b2)


Perturbation: Thermodynamic Perturbation Calculations.



* Syntax | Syntax of the set up of the perturbation command.
* Description | Description of the keywords and options for
setting up the perturbation calculation. Includes
an explanation of the reset command TSM CLEAr.
* Post-processing | How to process the perturbation output of the
dynamics run.

* pdetail: How to run perturbation
calculations.
* pimplem: How it is implemented.
Programming details.

* cfti: Conformational Energy/Free Energy
Calculation (Krzysztof Kuczera)


Top

Syntax for the Perturbation Command

[SYNTAX TSM]

TSM

Chemical Perturbation Parameters:

1. REACtant atom_selection_list | NONE

2. PRODuct atom_selection_list | NONE

3. LAMBda <real> [ POWEr <int> ]

4. SLOW TEMP <real> LFROm <real> LTO <real> [ POWEr <int> ]


5. DONT {REACtant} {internal_energy_spec} [SUBTract]
{PRODuct} {internal_energy_spec}

6. GLUE {CM FORCe <real> MIN <real>} [SUBR] [SUBP]
{ATOMs FORCE <real> MIN <real> atom_spec atom_spec

7. NOKE {REAC}
{PROD}

8. SAVE UNIT <integer> [FREQ <integer>]

9. COLO atom_spec PCHArge <real> [RCHArge <real>]
atom_spec ::= segid resnum type

10. PIGGyback PIGGy atom_spec BACK atom_spec
atom_spec ::= segid resnum type

11. UMBRella 4x( atom_spec) VACTual <real>
atom_spec: segid resnum type


Internal Coordinate (IC) Perturbation Parameters:

12. FIX {ic-spec} [TOLI <real>]

13. MAXI <integer>

14. MOVE {ic-spec} 2x{atom-selection} BY <real>

15. SAVIc [ICUNit <integer>] [ICFReq <integer>] [NWINdows <integer>]

16. END

internal_energy_spec ::== BOND THETa|ANGLe PHI|DIHEd IMPHi|IMPR

ic-spec ::= {[DISTance] 2x{atom-spec} }
{[BOND] 2x{atom-spec} }
{[ANGLe] 3x{atom-spec} }
{[THETa] 3x{atom-spec} }
{[DIHEdral] 4x{atom-spec} }
{[PHI] 4x{atom-spec} }

atom_spec ::= segid resid type

atom-selection ::= see (» select)


***** Note: must have non-bonded exclusions between reactant and product
atoms in rtf.

-----------------------------------------------------------------------

TSM CLEAr
Clears heap data structures used in perturbation setup, cancels
constraints and perturbations, and resets logical flags.


Top

Explanation of the Perturbation Setup


Currently the perturbation setup is initiated by invoking the
command TSM with nothing else on the command line. This is followed by
a number of other commands, listed below, and terminated with an END
command. Two types of thermodynamics perturbations are available: chemi-
cal perturbation and internal coordinate perturbation. Each is discussed
separately below.


* ChemPert | Chemical Perturbation
* ICPert | Internal Coordinate Perturbation


Top
Chemical Perturbation

For chemical perturbations, a minimum of three commands are necessary
besides TSM and END: REAC - to specify the reactant atom list; PROD - to
specify the product atom list; LAMBda or SLOW to specify lambda for win-
dowing or the slow growth technique.

1. REACtant atom_selection_list | NONE

Specifies the reactant atom list (» pdetail).
The atom selection list uses the standard CHARMM selection command syntax
select) Subsequent invocations of this command
clears the selections of any earlier invocation.

2. PRODuct atom_selection_list | NONE

Specifies the product list (see above).

3. LAMBda <real> [ POWEr <int> ]

The hybrid Hamiltonian is defined, in this implementation, as

H(lambda) = ( (1 - lambda)**N )V(reac) + (lambda**N)V(prod).

This command specifies lambda and N. It also indicates that the window
method is to be used (» pdetail).

4. SLOW TEMP <real> LFROm <real> LTO <real> POWEr <int>

This command specifies that the "slow growth" (see
» pdetail). method be used. LFROm and LTO indicates the
limits of integration. POWEr has the same meaning in the previous
command.

5. DONT {REACtant} {internal_energy_spec} [SUBTract]
{PRODuct} {internal_energy_spec}
internal_energy_spec :== BOND THETa|ANGLe PHI|DIHEd IMPHi|IMPR

This command indicates that the specified internal energy term(s)
for the reactant or product atoms is (are) to be ignored as perturbation
interactions. That means that the specified interactions are not
factored by lambda**N or (1 -lambda)**N and do not contribute to the
value of V(reac) or V(product). The interaction is, however, computed in
full and treated as part of H(env) (» pdetail). More
than one internal energy type can be specified at a time but reactant and
product must be specified in separate commands. The optional sub-command
SUBTract causes specified perturbation forces on the environment atoms
pdetail). to be subtracted. This is very
non-Newtonian and was included as an early (and largely unsuccessful)
attempt to generate configurations at lambda = 0 and 1 for the
non-existent group. See the discussion of the endpoint problem, (in
» pdetail). also known as the lambda goes to zero
catastrophe (Beveridge, 1987). Anyway, what this does is remove the
forces due to terms specified in the DONT option from the environment
atoms involved. In addition, the terms are not lambda factored or
included in H(lambda) or in V(reac) or V(pert). The forces are left on
the perturbed atoms with the hope that this would produce usable
configurations. It does not. Rather, the atoms drag along and bad
non-bond contacts result when lambda is 0 or 1 (the original intended
use).

We generally use the DONT option for both reactant and product
for the bond stretching and bending terms. These terms are generally
uncoupled from the interactions of interest and it appears that their
exclusion, even with the resulting non-physical Hamiltonian, does not
significantly affect the relative free energies of solvation or
drug/enzyme binding. The same cannot be said for torsions which other
implementations leave out.

The educated advice is use the DONT REAC (and PROD) BOND THETA
and donot!!! use the SUBTract option. Note that we have had problems
with the fraying of bonds to hydrogen near the lambda endpoints when
the DONT BONDs options were NOT used.

Note that the REAC and PROD must be issued before the respective
DONT command. Subsequent invocations of a DONT REAC/PROD command clears
the applicable flags first.

6. GLUE {CM FORCe <real> MIN <real>} [SUBR] [SUBP]} |
{ATOMs FORCE <real> MIN <real> atom_spec atom_spec }
atom_spec ::= segid resnum type

Here we have another failed attempt. The GLUE ATOM command sets
an harmonic force between a reactant and a product atom. One of each
must be given. The GLUE CM indicates that the centers of mass of the
reactant and product atoms are to be connected by the harmonic force.
FORCe is the force constant in the same units as the bond stretching
force constant kcal/mol/A**2. MIN is the minimum distance in angstroms.
Unfortunately, using MIN set to zero causes problems with SHAKE (how does
a floating point zero divide check error grab you, Buckaroo?). We
intended to use this on systems where there were no environment atoms to
keep the groups together. Shake would be used since the "GLUE" force is
unphysical. However, the aforementioned error made use of this option
undesirable. As it turns out, in solvated systems our concerns about the
two groups flopping around, with attendant problems with sampling
convergence, was unfounded. Best not to use this option.

7. NOKE {REAC}
{PROD}

This specifies that the kinetic energy should not include either
contributions form the REACtant or PRODuct atoms. REACtant and PRODuct
must be selected in separate commands. When the number of degrees of
freedom are calculated in the subroutine DCNTRL, the ones due to REACtant
or PRODuct (depending on which command(s) is/are issued) are not counted.
When the kinetic energy is calculated in the subroutine DYNAMC, these
degrees of freedom are ignored. Same for the temperature calculation.
This option should be used for the non-existent atoms at lambda = 0
(product atoms) or 1 (reactant atoms) since the atoms donot exist (hence
they are termed non-existent buckaroo) they should not be expected to
contribute their (3/2)kT per degree of freedom to the kinetic energy.

Normally, the kinetic energy contributions from the reactant and
product atoms are factored by (1 - lambda)**N and lambda**N respectively.

8. SAVE UNIT <integer> [FREQ <integer>]

This command determines where the output generated in DYNAMC is
to be sent (UNIT command) and the frequency of output (FREQ command). We
generally use a frequency of one (output on every step). Note that
before dynamics are run the file must be opened for formatted writing
with the CHARMM OPEN command. Binary output is not currently supported.

9. COLO atom_spec PCHArge <real> [RCHArge <real>]
atom_spec ::= segid resnum type

Sometimes the van der Waal characteristics and the "identity" of
of an atom remains the same while the charge interactions with this atom
atom is perturbed from a reactant value to a product value. An example
is the methanol -> ethane mutation where the methanol OH atoms are
full-fledged reactant atoms and one ethane methyl group is a
full-fledged product atom. The common methyl group is treated as a COLO
atom. Interactions appropriately factored by lambda are calculated using
a product charge, specified by the PCHArge command and a reactant charge,
either the charge in the residue topology file or that specified with the
RCHArge command. The COLO command is issued once for each COLO atom.
The colo atoms cannot be in either the reactant or product lists. If any
of this is unclear » pdetail An explanation about how
this whole thing works is given there.

10. PIGGyback {PIGGy} atom_spec {BACK} atom_spec
{REACtant} {PRODuct}
atom_spec ::= segid resnum type

This command allows you to convert isolated atom into another.
It is intended for mutations such as Br- -> Cl- and Ar -> Ne. The atoms
cannot be bonded to anything else (a wrndie error will be issued).
Currently, only one reactant/product atom pair can be used. The PIGGy
atom must be a REACtant atom and the BACK atom must be a PRODuct atom,
otherwise an error will be flagged. The REAC and PROD commands must
already have been issued. Note the synonyms for PIGGy and BACK.

When this option is in effect the forces on the back atom are
added to those on the PIGGy atom at each step of the dynamics. The
coordinates of the BACK atom are made equal to those of the PIGGy atom
at each time step.

11. UMBRella 4x( atom_spec) VACTual <real>
atom_spec ::= segid resnum type

This command specifies an umbrella sampling correction to all of
the averages in post-processing. The four atom specifications define the
the dihedral angle involved. The command is repeated for each dihedral.
If there are multiple dihedral angles through the axis of two atoms, all
all of them should be specified. It is assumed that the surrogate
potential term is in the parameter file for the particular type of
torsion. VACTual is the coefficient for the real potential. Currently,
only the three-fold term is supported. A further limitation is that
although you can specify particular dihedral angles for this treatment
all torsions with that type will use the modified potential in the
parameter file. This part of the program is slated for modification
as soon as possible. For an explanation of the terms and how
the umbrella correction works, » pdetail



Top
Internal Coordinate (IC) Perturbation

To setup an ic perturbation, you need to 1) specify the internal
coordinate(s) to be constrained during the perturbation (FIX), 2) specify
which atoms will move during the perturbation and which atoms will remain
fixed (MOVE), and 3) indicate how and where the perturbation data will be
saved (SAVI).


12. FIX {ic-spec} [TOLI <real>]

ic-spec ::= {[DISTance] 2x{atom-spec} }
{[BOND] 2x{atom-spec} }
{[ANGLe] 3x{atom-spec} }
{[THETa] 3x{atom-spec} }
{[DIHEdral] 4x{atom-spec} }
{[PHI] 4x{atom-spec} }

The FIX command defines an internal coordinate to be constrained:
DIST or BOND specify distance constraints, ANGL or THET bond angle
constraints, and DIHE or PHI dihedral angle constraints. TOLI sets the
tolerance (the allowed deviation in Angstroms (distance constraints) or
degrees (angle constraints)) to be used for the specified constraint in
the constraint resetting procedure (» pdetail). The
default is 10E-10 (Angstroms or degrees). It is very important to note
that the reference value of the constraint is set to the value of the
internal coordinate at the instant the command is issued. One or more
i.c. constraints can be specified per simulation.

13. MAXI <integer>

MAXI sets the maximum number of iterations to be used in the
iterative i.c. constraint resetting procedure ( details:
(pdetail).). The default is 500.

14. MOVE {ic-spec} BY <real> INTE {atom-selection}

The MOVE commands specify the internal coordinates to be perturbed
and define the atoms to be moved by the perturbation. Several MOVE
commands may be used to set up a perturbation consisting of changes in
several internal coordinates. Since i.c. perturbations are really only
useful in conjunction with i.c. constraints, for each MOVE command there
should be a corresponding FIX command with the same ic-spec. Following
the BY keyword is a real number which is the amount that the internal
coordinate will be changed by the perturbation (in Angstroms for distances
and degrees for angles).

The INTE selection part of the MOVE command defines the solute
partition, that is, the atoms to be moved by the perturbation. The
perturbation is applied to all of the atoms specified in the selection
using displacements determined from moving the internal coordinate BY
value. However, some atoms may not be moved even though they are included
in the solute partition with the INTE selection because zero displacements
will be computed for them (e.g. if they lie on the rotation axis, like the
central atom in a perturbed angle, or either of the two central atoms in a
perturbed dihedral angle). If a double selection is given (e.g. INTE SELE
atom-selection END SELE atom-selection END), then the two selected groups
of atoms are considered separate sections of the solute partition. In
that case, to accomplish the overall perturbation, each section is moved
half of the BY value.

The INTE selection also specifies which contributions are to be
included in the perturbation interaction energies. The calculation of the
perturbation interaction energies is based on the interaction energy
calculation which is done when the CHARMM INTEre command is issued (see
» energy for more details). Thus, the perturbation
interaction energies may contain the following energy contributions: bond,
bond angle, dihedral, improper dihedral, van der Waals, electrostatic,
hydrogen bond, harmonic positional constraint, and harmonic dihedral
constraint. In addition to these contributions, which are the usual
also contain the following image contributions: van der Waals,
electrostatic, and hydrogen bond. (Note that the CHARMM INTEre command is
parsed by the main CHARMM command parser. It should not be confused with
the INTE part of the MOVE cammand which is parsed by the TSM command
parser. We apologize for any confusion which may result from the use of
the INTE keyword in the TSM command. It seemed appropriate since it
indicates a similar interaction energy calculation.)

To explain how the interaction energy calculation works, we define
two "selection groups". The first selection group contains all of the
atoms in the system. The second selection group contains all of the atoms
included in the INTE selection. The rules which are used to determine
which contributions are included in the interaction energies are as
follows: a bond is included if the two atoms defining the bond are in
different selection groups; a bond angle if the central atom is in both
selection groups; a dihedral angle (intrinsic torsion or harmonic
constraint) if the two central atoms are in different selection groups; an
improper dihedral angle if the first atom is in both selection groups; a
nonbonded interaction (van der Waals and electrostatic) if both atoms are
in different selection groups and the interaction is in the nonbonded
list; a hydrogen bond if the donor and acceptor are in different selection
groups; and finally, a harmonic positional constraint is included if the
atom is in both selection groups.

The user should decide carefully which interaction energy contribu-
tions she wants to have included before running the perturbation simula-
tion. Then she must appropriately design the INTE selection. For
example, suppose she wants to compute the free energy as a function of the
dihedral angle in an extended-atom (four atom) model for butane. A change
in the dihedral angle only changes the position of the methyl group. The
user might therefore select only a terminal methyl group (e.g. segment
BUTA, residue 1, atom C4) using the INTE command:

INTE SELE (ATOM BUTA 1 C4) END

With this selection, the intrinsic torsional potential would not be
included in the interaction energies since the CHARMM interaction energy
routine only computes torsional terms if the central two atoms of the
torsion are in different selection groups. Of course, this is not
generally a problem, since the missing term could be simply added to the
thermodynamics after processing the interaction energies. If the user
preferred to have the intrinsic torsional contribution included in the
interaction energies, she would add the methylene group (atom C3) to the
INTE selection, e.g. she could use the following selection in place of the
one above:

INTE SELE ((ATOM BUTA 1 C3) .or. (ATOM BUTA 1 C4)) END

With this specification, the C3 atom is included in the solute partition.
However, its position is not changed by the perturbation since it lies on
the axis about which the solute atoms are rotated in the dihedral angle
perturbation. Now the two central atoms, C2 and C3, are included in
different selection groups, so the intrinsic torsional contribution is
included in the interaction energies.
There is a subtle point that must be considered when the perturba-
tion consists of moving more than one internal coordinate. As an example,
suppose we want to perturb both the dihedral angles, which we call phi and
psi, in an extended atom (five-atom) model for pentane. Further suppose
that, in the double-wide sampling, we want the perturbation in one
direction to increase both phi and psi, and the perturbation in the other
direction to decrease them. We might try the following MOVE commands
(e.g. +/- 5 degree perturbations of each dihedral angle; C2 and C4 are
selected so the intrinsic torsion terms are included in the interaction
energies):

MOVE DIHE PENT 1 C1 PENT 1 C2 PENT 1 C3 PENT 1 C4 BY 5.0 -
INTE SELE ((ATOM PENT 1 C1) .OR. (ATOM PENT 1 C2)) END
MOVE DIHE PENT 1 C2 PENT 1 C3 PENT 1 C4 PENT 1 C5 BY 5.0 -
INTE SELE ((ATOM PENT 1 C4) .OR. (ATOM PENT 1 C5)) END

However, in the algorithm which changes bond and dihedral angles, a
perturbation in the forward direction corresponds to a counterclockwise
rotation of the atoms to be moved around the bond vector (e.g. C2ÐC3 or
C3ÐC4 in pentane). Thus, with the above MOVE commands, the forward
perturbation decreases phi while it increases psi. That is not what we
wanted. To fix the problem, we simply reverse the sign of the BY value in
one of the MOVE commands:

MOVE DIHE PENT 1 C1 PENT 1 C2 PENT 1 C3 PENT 1 C4 BY 5.0 -
INTE SELE ((ATOM PENT 1 C1) .OR. (ATOM PENT 1 C2)) END
MOVE DIHE PENT 1 C2 PENT 1 C3 PENT 1 C4 PENT 1 C5 BY Ð5.0 -
INTE SELE ((ATOM PENT 1 C4) .OR. (ATOM PENT 1 C5)) END

Now the forward perturbation decreases both phi and psi and the reverse
perturbation decreases them. The user should carefully consider how the
atoms will be moved when choosing the signs of the BY value when more than
one internal coordinate is perturbed.


15. SAVIc [ICUNit <integer>] [ICFReq <integer>] [NWINdows <integer>]
[SUPP]
(for "on-the-fly" free energy and average energy calculations:)
[RUNA] [PEVEry <integer>] [RUNIt <integer>] [RPRInt <integer>]
[TEMP <real>]
The SAVI command specifies how and where the perturbation data
(which consists primarily of internal coordinate values and interaction
energies) will be saved during the simulation. The integer following the
ICUN keyword is the number of the fortran unit to which the perturbation
data is written. The perturbation file should be opened for formatted
writing on this unit using the CHARMM OPEN command before the dynamics
command is issued. The integer following ICFR is the frequency (in
dynamics steps) with which the data is written to the file. The default
ICFR value is 10. If the frequency is zero (e.g. if the SAVI command is
not issued) or ICFR 0 is indicated, then a level 0 warning is issued since
there is no need to do perturbations if the data is not going to be saved.
The integer, m, following the NWIN keyword indicates the number of
"double-wide" subintervals that the BY value, dx, will be divided into.
Thus, the 2m perturbations, dxi = i*dxm, where dxm = dx/m and i = -m,-
m+1,..., -1, 1,...m-1,m, are all carried out during the simulation,
yielding 2m free energy differences. For example, if the BY value is 1.0
and the NWIN value is 2, perturbations which change the internal
coordinate by -1.0,-0.5,0.5, 1.0 are carried out. The default NWIN value
is 1. The SUPP keyword suppresses printing of the internal coordinate
values to the output file.
Running or "on-the-fly" free energy changes and average energy
changes can be calculated for internal coordinate perturbations through
the invocation of the "RUNAverage" keyword. The routine will include all
data points (i.e. every ICFR'th step in the trajectory) in these cal-
culations. The results will be written to the specified file
in RUNIt every RPRInt sampled data points (default is no writing out of
averages). Hence for ICFR of 5 and RPRInt of 10, the averages will be
calculated every 5 steps and written out every 50 (RPRInt*ICFR) steps.
(See testcase for examples.) TEMP specifies the temperature in degrees
Kelvin at which the free energy is to be calculated (default 300).
PEVEry specifies the period (in ICFR number of steps) for writing out
the usual tsm output file containing the energies and the internal
coordinates (default 1--file is written every ICFR steps).

The "on-the-fly" output file is formatted as follows:
RUNAV> 5 168.5417 1.84955880 2.02536215
RUNAV> 5 168.0417 0.84931404 0.90105846
RUNAV> 5 167.0417 -0.69647324 -0.62795693
RUNAV> 5 166.5417 -1.21666615 -0.91659629
RUNAVI> 1 167.54174244
RUNAVI> 2 155.47953779
RUNAV> 10 170.7559 1.62569412 1.84119225
RUNAV> 10 170.2559 0.77839900 0.83143112
RUNAV> 10 169.2559 -0.67517686 -0.62242291
RUNAV> 10 168.7559 -1.20371332 -0.99498136
RUNAVI> 1 169.75592284
RUNAVI> 2 155.45919138

For the "RUNAV>" lines, the 2nd column gives the number of data points
included in the averages. The second line gives the average value of
the internal coordinate after a particular perturbation. (For perturba-
tions involving multiple internal coordinates, the value of only the
only the first internal coordinate specified in the input file is given).
The third and fourth columns give the free energy change and average
energy change, respectively, for the given perturbation.
For the "RUNAVI>" lines, the first column gives the number of the
internal coordinate (in its order of appearance in input file) and
the second column gives the average value for that coordinate over
the unperturbed trajectory. All coordinates involved in the
perturbations (i.e. specified by the MOVE command) are listed.

16. END

Terminates the perturbation setup. At this point the program
does additional error checking and prints out values of some parameters.

-------------------------------------------------------------------------

TSM CLEAr

A separate command ( NOT!! a setup command ) to clear logical
flags and release HEAP memory allocated for perturbation data structures.
It is not necessary to use this command unless you have more than one
dynamics run in a single job and want to reset or turn off the
perturbation. Definitely invoke this command before entering the
perturbation setup a second time.



Top

Post-Processing of Perturbation Output



* PSyn | Syntax for setting up the post-processing
* PPost | Description of the keywords and options for setting up
post-processing of the perturbation calculation
* PProc | Processing perturbation data
* PEnd | The END command of post-processing


Top

Syntax for Post-Processing Commands


1. TSM POST [PSTAck <int>] [PLOT] [TI] [NODEriv] [COMPonents] [ENDPoints]

[IC] [MAXP <integer>] [MAXW <integer>] [SURF] [MAXS <integer>]
[NODEriv] [INTE]

2. PROCess FIRSt <int> [NUNIt <int>] BINSize <int>
[CTEM] [TEMP <real>] [DELTa <real>]
[BEGin <integer>] [STOP <integer>] [SKIP <int>] [NMAX <int>]
LAMBda <real> [ONE] [ZERO] [UMBRella] [EAVG]

3. END



Top

Description of Post-Processing Commands


Post-processing of perturbation data is initiated by the following
command:

TSM POST [PSTAck <int>] [PLOT] [TI] [COMPonents] [ENDPoints]

[IC] [MAXP <integer>] [MAXW <integer>] [SURF] [MAXS <integer>]
[INTE]

[NODEriv]


Summary of Parameters:
----------------------

1. Chemical Perturbation Post-processing Parameters

PSTAck: Array size for plotting x,y points. Default 100.
Needed for thermodynamic integration and/or
plotting.
PLOT: Create PLT2 output.
TI: Thermodynamic integration:
Delta A = int 0 to 1 <dE/dLambda>.
COMP: Do Vdw, Elec and Intern component analysis.
ENDP: Calculate TI integral to endpoints, i.e., full
(0,1).


2. Internal Coordinate (IC) Perturbation Post-processing Parameters

IC: Specifies that ic perturbation output will be
processed.

MAXP: Maximum number of ic perturbations.
MAXW: Maximum number of ic perturbation windows (NWIN in
SAVI command).
SURF: Generate thermodynamic surfaces from ic
perturbation data.
MAXS: Maximum number of points in surface.
NODEriv: Only calculate the free energy.
INTE: Calculate average interaction energies from ic
perturbation data.

3. NODEriv Subcommand

NODEriv: Only calculate the free energy.


Chemical Perturbation Post-processing
-------------------------------------

There are two methods of calculating the relative free energies
and relative temperature derivative properties: the perturbation method
and the Thermodynamic Integration method (» pdetail).
By default the perturbation method is used. The optional parameter TI
specifies the thermodynamic integration technique. PSTAck determines
the size of arrays to be allocated from the stack. For the default
perturbation method this allocates space for plotting values. The TI
method requires arrays for actually calculating the thermodynamic
properties. The parameter PLOT indicates that output is created for
PLT2 data files. They must be edited out of the output file. NODEriv
indicates that only the free energy is to be calculated and not Delta E
or Delta S.


Internal Coordinate Perturbation Post-processing
------------------------------------------------

The MAXP, MAXW, and MAXS parameters are used to allocate memory for
the processing of the perturbation data. The integer following the
keyword MAXP is the maximum number of perturbed internal coordinates in
the data files to be processed (e.g. the number of MOVE commands in the
perturbation dynamics input files). The default MAXP value is 1. The
integer following MAXW is the maximum number of subintervals in each
window (e.g. the NWIN value on the SAVI command command line in the
dynamics input files). The default MAXW value is also 1. If the SURF
keyword is present on the TSM POST command line, then thermodynamic
surfaces will be constructed using the thermodynamic differences computed
from the perturbation data. We say more about this below. The integer
following MAXS is the maximum number of points in a thermodynamic surface.
If all of the N data files to be processed using PROC commands (see below)
have the same number of subintervals, m, the MAXS value is equal to mN +
1. The default MAXS value is 100.

In addition to the free energy differences, the internal energy and
entropy differences are computed by default using finite-difference
temperature derivatives. However, the calculation of the derivative
properties may be turned off using the NODE keyword. If the NODE keyword
is present on the TSM POST IC command line, the internal energy and
entropy differences are not computed. If the INTE keyword is present, the
average interaction energies of the solute partition (the atoms specified
by the INTE selection in the MOVE command) with the bath partition (the
remaining atoms), as well as the average total energies in the unperturbed
and perturbed systems are computed and printed in the output file from the
post-processing run. By default the average interaction energies and
average total energies are not computed.


Top
The TSM POST command is followed by one or more PROC commands which
specify the processing of perturbation data files, and terminated with the
END command. The syntax of the PROC command is as follows:

PROC FIRSt int [NUNIts int] BINSize int [CTEMp] [TEMP real] [DELTa real]

LAMBda <real> [ONE] [ZERO] [UMBRella] [EAVG] [SKIP <int>]
[NMAX <int>]

[BEGIn int] [STOP int]


Summary of Parameters:
----------------------

FIRSt Fortran unit number of first file.

NUNIT Number of fortran i/o units. One can 'tack' on trajectories
from separate files in the manner used for trajectory commands
in correl. The files must be opened for read access prior
initiating the post processing with the TSM POST command, and
the units must be numbered consecutively, starting with FIRSt.
This is due to the fact that the post-processor command reader
does not handle MISCOM commands (» MISC). This
probably should be corrected. Only formatted files are
handled currently, remember to open the files as formatted. We
also note that it does not matter which order multiple files
in a given window are processed. The post-processing program
does not check to see if the dynamics steps are contiguous. It
only checks to see that all of the files in a given window
have the same header, as they should. Default = 1.

BINSize The number of data points per bin for error calculation.

CTEMp A flag to indicate that average temperature is to be
calculated. Because this is being calculated while the other
averages are being accumulated the temperature is not used in
calculating the thermodynamic properties. To use the average
temperature in the thermodynamic calculations, the user has to
process the data twice, manually specifying the average temp-
erature from the first processing run as the TEMP value in the
second run. By default the average temperature is not computed.

TEMP The temperature for calculating properties. Default = 298.

DELTa The temperature increment for finite difference derivatives
(calculate delta E and delta S). No meaning if TI is
specified. A level 0 warning is issued. The default is 10
degrees.


The following parameters are only used when processing chemical
perturbation data.

LAMBda Lambda prime. For calculating <exp-beta(E(lambda')-E(lambda)>
and related quantities. No meaning if TI is specified.
Level 0 warning issued. (» pdetail).

ONE Indicates that lambda is exactly 1. Overides input lambda in
file. This is used only in TI. In the case on non-linear
lambda dependence the derivatives due to reactant terms are
identically zero. This provides a solution to the lambda ->
zero catastrophe.

ZERO Indicates that lambda is exactly 0. Overides input lambda in
file. Commands ONE and ZERO are mutually exclusive and are
used only in the TI post processor. Level 0 warning issued.
This command is used only in TI. In the case on non-linear
lambda dependence the derivatives due to product terms are
identically zero. This provides a solution to the lambda ->
zero catastrophe.

UMBRella A flag to indicate that umbrella sampling is used. A check is
made of a parameter line in each data file to see if umbrella
sampling was indeed used.

EAVG calculate <Etot.> and uncertainty for this value of lambda
ignored if TI.

SKIP skip first nskip records.

NMAX maximum number of points to read. skips skip number of values
first.


The following parameters are only used when processing internal coordinate
perturbation data.

BEGI specifies number of first dataset to use in accumulating
averages.

STOP specifies number of last dataset to use in accumulating
averages.


Processing Chemical Perturbation Data
-------------------------------------

The PROCess command is usually issued several times. When using
the perturbation method one would issue it at least once for every
lambda. In all of our work so far, we have employed double-wide sampling
in that for each value of lambda whereupon dynamics are run, we "perturb"
both up and down from lambda to lambda prime (i.e. both less than and
greater than lambda, definitely » pdetail). The
program rewinds the files after each PROCess command. For each lambda
-> lambda prime perturbation, a separate PROCess command is issued. For
the TI method, one PROCess command per lambda is used and
<dE(lambda)/dlambda> is calculated.



Processing Internal Coordinate Perturbation Data
------------------------------------------------

The PROC command is used to specify the processing of perturbation
data from a single window. The PROC command is usually used several times
and the results from the various windows are usually constructed into
thermodynamic surfaces.

The user may specify that a subset of all the data read is to be
used in the calculation of the averages and thermodynamics. This option
is useful for examining the convergence of the thermodynamic properties.
The integers following the BEGI and STOP keywords are the numbers of the
first and last datasets (not dynamics steps), respectively, to be used for
processing the data for a given window. By default, all of the data is
used. If the limits are set using the BEGI and STOP keywords, they are
only used on the data processed by the particular PROC command which set
the limits (e.g. the defaults are reinstated after each PROC command).

The integer following the BINS keyword is the number of datasets per
batch, n, used in the calculation of the statistical uncertainties by the
the method of batch averages (» pdetail). This number
must be specified as there is no default value. We typically use BINS
100.


Top
END

This command terminates the post-processing. When this command
is received the averages generated by the issuing the PROCess commands
are combined and total values of the thermodynamic properties are
computed and output.

For chemical perturbations, if the TI method is used, a spline
polynomial is fit to the averages and integrated over limits determined
by the minimum and maximum lambda's. The lambda values do not have to be
processed in order since the program will sort them. It is the user's
responsibility to cover the whole range for lambda = 0 to 1 (if that is
the intention). Though there are cases where a range that does not
include those two endpoints may be useful (e.g. mixing linear TI and
linear perturbation *NOTE details: (pdetail).), discontinuous gaps in the
lambda curve do not make sense. In the section *NOTE details: (pdetail).
Input examples and explanations of the different methods are given.

For internal coordinate perturbations, if surface construction has
been requested by including the SURF keyword on the TSM POST IC command
line, the average internal coordinate and thermodynamic values (free
energy, internal energy, and entropy differences) are sorted for construc-
tion of the surfaces. The sorting is done according to increasing values
of the average internal coordinate of the first perturbed internal coordi-
nate (e.g. the i.c. specified in the first MOVE command issued in the
dynamics input file). Then the thermodynamic surfaces are constructed by
combining the differences in the thermodynamic properties. Finally, the
surfaces are printed to the output file of the post-processing run. If
more than one internal coordinate is perturbed, an identical set of sur-
faces is printed out as functions of each perturbed internal coordinate.
The surfaces may be simply cut from the output file using an editor for
subsequent plotting (e.g. using the PLT2 program).