# tmd (c43b1)

Targeted Molecular Dynamics

The Targeted Molecular Dynamics (TMD) method introduces a holonomic

constraint that reduces the rmsd with a predefined target at each MD

step. Three flavors of the method are available: the original TMD

method (J. Schlitter, M. Engels, P. Kruger, E. Jacoby and A. Wollmer,

Mol. Sim. 10, 291 (1993)), the zeta-TMD method (Q. Cui, to be

published), and the restricted perturbation - TMD method (A. van der

Vaart and M. Karplus, J. Chem. Phys. 122, 114903 (2005)).

The methods are implemented for the LEAP integrator, and Berendsen's

thermostat must be used; CHARMM needs to be compiled with the "TMD"

keyword present in the pref.dat file.

* Syntax | Syntax of the dynamics command

* Description | Description of the keywords and options

The Targeted Molecular Dynamics (TMD) method introduces a holonomic

constraint that reduces the rmsd with a predefined target at each MD

step. Three flavors of the method are available: the original TMD

method (J. Schlitter, M. Engels, P. Kruger, E. Jacoby and A. Wollmer,

Mol. Sim. 10, 291 (1993)), the zeta-TMD method (Q. Cui, to be

published), and the restricted perturbation - TMD method (A. van der

Vaart and M. Karplus, J. Chem. Phys. 122, 114903 (2005)).

The methods are implemented for the LEAP integrator, and Berendsen's

thermostat must be used; CHARMM needs to be compiled with the "TMD"

keyword present in the pref.dat file.

* Syntax | Syntax of the dynamics command

* Description | Description of the keywords and options

Top

Syntax for the Targeted Molecular Dynamics commands

TMDInitialize { [ INRT integer ] } [ atom-selection ] [ atom-selection ]

{ [ DINC real ] }

{ [ FRMS real ] }

{ [ ITMD integer ] }

{ [ FTMD integer ] }

{ [ ENER integer ] }

{ [ MAXF real ] }

{ [ MAXB real ] }

{ [ SUMP integer ] }

{ [ ZETA ] }

{ [ CZETa real ] }

{ [ ZTOL real ] }

{ [ ZMIT integer ] }

INRT 1000 Number of step that one needs to get rid of artificial

rotational motion in TMD simulation.

DINC 0.0 RMS increment in TMD simulation (ignored in RP-TMD

simulation).

FRMS 1.0D-6 Stop dynamics when a rmsd of FRMS with the target is reached.

ITMD -1 Write analysis file to this unit (-1: don't write).**

FTMD -1 Frequency of writing analysis file.**

ENER 0 Write potential energy difference due to the TMD constraint

to the analysis file if >0 (very expensive!).**

MAXF -1.0 If positive: perform a RP-TMD simulation with the value of

MAXF as the maximum perturbation.

If negative: perform an "original" TMD simulation.**

MAXB -1.0 If MAXF<0: ignored.

If MAXF>0 and MAXB>0: Use MAXF as the maximum perturbation

when C>0, use MAXB as the maximum perturbation when C<0,

where C = Sum ( |P_i| Cos(p_i F_i)).

If MAXF>0 and MAXB<0: Always use MAXF as the maximum

perturbation.**

SUMP 0 If MAXF<0: ignored.

If 0: the maximum perturbation is the maximum atomic

perturbation (MAXF = max |p_i|).

Else: the maximum perturbation is a sum over all

atoms (MAXF = Sum |p_i|).**

ZETA Indicates Zeta form of TMD is being used (default: not active)

CZETa 1.0 Zeta form expotential factor.

ZTOL 1.0E-10 Tolerance used in calculating Zeta-TMD constraint.

Positions are solved for (Zeta-Zeta0) < ZTOL.

ZMIT 1000 Number of iterations allowed in the Zeta-TMD constraint

subroutine.

1st atom-selection Apply TMD fit to the selected atoms only.

2nd atom-selection Apply TMD perturbation to the selected atoms only.

** For TMD/RP-TMD only.

Syntax for the Targeted Molecular Dynamics commands

TMDInitialize { [ INRT integer ] } [ atom-selection ] [ atom-selection ]

{ [ DINC real ] }

{ [ FRMS real ] }

{ [ ITMD integer ] }

{ [ FTMD integer ] }

{ [ ENER integer ] }

{ [ MAXF real ] }

{ [ MAXB real ] }

{ [ SUMP integer ] }

{ [ ZETA ] }

{ [ CZETa real ] }

{ [ ZTOL real ] }

{ [ ZMIT integer ] }

INRT 1000 Number of step that one needs to get rid of artificial

rotational motion in TMD simulation.

DINC 0.0 RMS increment in TMD simulation (ignored in RP-TMD

simulation).

FRMS 1.0D-6 Stop dynamics when a rmsd of FRMS with the target is reached.

ITMD -1 Write analysis file to this unit (-1: don't write).**

FTMD -1 Frequency of writing analysis file.**

ENER 0 Write potential energy difference due to the TMD constraint

to the analysis file if >0 (very expensive!).**

MAXF -1.0 If positive: perform a RP-TMD simulation with the value of

MAXF as the maximum perturbation.

If negative: perform an "original" TMD simulation.**

MAXB -1.0 If MAXF<0: ignored.

If MAXF>0 and MAXB>0: Use MAXF as the maximum perturbation

when C>0, use MAXB as the maximum perturbation when C<0,

where C = Sum ( |P_i| Cos(p_i F_i)).

If MAXF>0 and MAXB<0: Always use MAXF as the maximum

perturbation.**

SUMP 0 If MAXF<0: ignored.

If 0: the maximum perturbation is the maximum atomic

perturbation (MAXF = max |p_i|).

Else: the maximum perturbation is a sum over all

atoms (MAXF = Sum |p_i|).**

ZETA Indicates Zeta form of TMD is being used (default: not active)

CZETa 1.0 Zeta form expotential factor.

ZTOL 1.0E-10 Tolerance used in calculating Zeta-TMD constraint.

Positions are solved for (Zeta-Zeta0) < ZTOL.

ZMIT 1000 Number of iterations allowed in the Zeta-TMD constraint

subroutine.

1st atom-selection Apply TMD fit to the selected atoms only.

2nd atom-selection Apply TMD perturbation to the selected atoms only.

** For TMD/RP-TMD only.

Top

Description of the Targeted Molecular Dynamics Commands

To invoke TMD, the TMDInitialize command should be given before the

DYNAmics command. After TMDInitialize, the target structure should be

read in with the "READ COOR TARG" command. All TMD variables and the

target coordinates are cleared after the DYNAmics command; before

a restart you should invoke the TMDInitialize command (and the "READ

COOR TARG" command) once again.

Two atom selections are used with the TMDInitialize command. The first

selection is used to define the atoms used in fitting both targets

to the current structure (done every INRT steps). The second selection

is used to define the atoms which the TMD constraint is applied.

If only one selection is given in TMDI, this selection will be used for

both fitting and applying the constraint. To run TMD in its original

form, one must use 'select all end' for both selections. One should

perform an overlay of the structures before the simulation.

Please note that the "standard" TMD and RP-TMD methods are parallellized;

the 'Zeta-TMD' has not been parallellized.

A) Original TMD method.

For doing Targeted Molecular Dynamics (TMD), one needs to define

a moving coordinate and a target coordinate. You slowly pull the

moving structure towards the target structure by gradually decreasing

the RMS distance between two. The 'pulling' speed is defined

by the user.

Commands:

OPEN UNIT 88 WRITE CARD NAME tmd.dat

TMDINITIALIZE ITMD 88 FTMD 10 FRMS 1.2 INRT 10 DINCRE 0.0004 -

SELE ALL END SELE ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

B) Zeta-TMD method.

To constrain dynamics between two target structures (between a

starting and ending structure of a conformational transition,

for example), the 'Zeta' form of the constraint function is used:

Zeta(t) - Zeta0(istep) = 0 (contraint),

where

Zeta(t) = -1/(1+EXP(-CZETA*RMSD1(t))) + 1/(1+EXP(-CZETA*RMSD2(t)))

Zeta0(istep) = Zeta0(istep-1) - DINC

RMSD1 = (mass weighted) root-mean-squared difference (RMSD)

between the current structure and TARG (using atoms

defined by second selection in TMDI)

RMSD2 = RMSD between the current structure and TAR2

The two target structures are read using READ COOR TARG and READ COOR TAR2,

respectively. The sign of DINC (incrementation of the contraint function)

determines which structure the molecule is pulled towards, and which one is

pushed away from, during dynamics run. For DINC > 0, TMD pulls the molecule

towards TAR2. For DINC < 0, TMD pulls the molecule towards TARG.

The starting value of Zeta0 is based on the coordinates at the start of

the dynamics run; istep is the current step number in the dynamics run.

The ZETA keyword must be used in the DYNA command line for this. If two

targets are read in, but ZETA is not specified, then only the one TARG

structure is used in the TMD algorithm and the Zeta form of the

constraint is not used.

The Zeta form is useful, since it is more effective at pulling molecules

towards target structures than other relative constraint forms, such as

((RMSD1 - RMSD2) - rho) = 0, where the difference in RMSDs may be well

defined, but the current structure may be far from both target structures.

Also, transitions are not limited to paths which only allow for the RMSD

to one target structure to decrease monotomically.

ZTOLerance and ZMITerations are used in the minimization scheme for

calculating the coodinates which satisfy the TMD constraint. They are

similar to the cooresponding terms in the SHAKE algorithm.

The constrained RMSD for one-target TMD is not allowed to go below zero.

Similarily, the restriction |Zeta| <= -1/(1+EXP(-CZETA*RMSD0))+1/2 is

is used, where RMSD0 is the RMS Difference between the two target

structures. Once these values are reached during dynamics, the contraint

value for RMSD (or Zeta) is held at this limiting value.

This subroutine outputs RMSD1, RMSD2, and the actual Zeta value (which

is within +/- ZTOL of Zeta0(istep)), with PRNLEV >= 5. For each dynamics

step, this is likely to print out a few times, due to the iterative scheme

used between this subroutine and the SHAKE subroutine.

Commands:

TMDINITIALIZE INRT 1 DINC -0.0003 -

ZETA CZETA 1.0 ZTOL 1.0E-8 ZMIT 1000 -

SELECT ALL END SELECT ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

OPEN READ UNIT 2 CARD NAME init.crd

READ COOR UNIT 2 CARD TAR2

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

C) Restricted perturbation - TMD method.

In this method, the coordinate displacement (perturbation) is limited

to a preset value; given this displacement, the rmsd with the target

is minimized at each step. This procedure prevents the crossing of

large energy barriers, and may increase the efficiency of the calculation.

Either the total perturbation Sum |p_i| or the maximum atomic

perturbation Max |p_i| can be restricted.

The function C = Sum ( |p_i| Cos(p_i,F_i) ) is a good indicator of

barrier crossings: when C is negative, a barrier has probably been

crossed. To reduce barrier crossings, the perturbation can be decreased

when C<0 (this will increase the simulation time).

Note that in this method, the rmsd fluctuates along the trajectory and

the length of the simulation may vary (simulations may get "stuck" when

very small perturbations are used).

See J. Chem. Phys. 122, 114903 (2005) for a more detailed discussion

of the algorithm, and a comparison of the RP-TMD method with the

standard TMD method.

Commands:

OPEN UNIT 88 WRITE CARD NAME tmd.dat

TMDINITIALIZE ITMD 88 FTMD 10 FRMS 1.2 INRT 10 MAXF 0.001 -

MAXB 0.0008 SUMP 1 -

SELE ALL END SELE ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

Examples: tmdtest32.inp (serial & parallel), and tmd_zeta.inp (serial).

Description of the Targeted Molecular Dynamics Commands

To invoke TMD, the TMDInitialize command should be given before the

DYNAmics command. After TMDInitialize, the target structure should be

read in with the "READ COOR TARG" command. All TMD variables and the

target coordinates are cleared after the DYNAmics command; before

a restart you should invoke the TMDInitialize command (and the "READ

COOR TARG" command) once again.

Two atom selections are used with the TMDInitialize command. The first

selection is used to define the atoms used in fitting both targets

to the current structure (done every INRT steps). The second selection

is used to define the atoms which the TMD constraint is applied.

If only one selection is given in TMDI, this selection will be used for

both fitting and applying the constraint. To run TMD in its original

form, one must use 'select all end' for both selections. One should

perform an overlay of the structures before the simulation.

Please note that the "standard" TMD and RP-TMD methods are parallellized;

the 'Zeta-TMD' has not been parallellized.

A) Original TMD method.

For doing Targeted Molecular Dynamics (TMD), one needs to define

a moving coordinate and a target coordinate. You slowly pull the

moving structure towards the target structure by gradually decreasing

the RMS distance between two. The 'pulling' speed is defined

by the user.

Commands:

OPEN UNIT 88 WRITE CARD NAME tmd.dat

TMDINITIALIZE ITMD 88 FTMD 10 FRMS 1.2 INRT 10 DINCRE 0.0004 -

SELE ALL END SELE ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

B) Zeta-TMD method.

To constrain dynamics between two target structures (between a

starting and ending structure of a conformational transition,

for example), the 'Zeta' form of the constraint function is used:

Zeta(t) - Zeta0(istep) = 0 (contraint),

where

Zeta(t) = -1/(1+EXP(-CZETA*RMSD1(t))) + 1/(1+EXP(-CZETA*RMSD2(t)))

Zeta0(istep) = Zeta0(istep-1) - DINC

RMSD1 = (mass weighted) root-mean-squared difference (RMSD)

between the current structure and TARG (using atoms

defined by second selection in TMDI)

RMSD2 = RMSD between the current structure and TAR2

The two target structures are read using READ COOR TARG and READ COOR TAR2,

respectively. The sign of DINC (incrementation of the contraint function)

determines which structure the molecule is pulled towards, and which one is

pushed away from, during dynamics run. For DINC > 0, TMD pulls the molecule

towards TAR2. For DINC < 0, TMD pulls the molecule towards TARG.

The starting value of Zeta0 is based on the coordinates at the start of

the dynamics run; istep is the current step number in the dynamics run.

The ZETA keyword must be used in the DYNA command line for this. If two

targets are read in, but ZETA is not specified, then only the one TARG

structure is used in the TMD algorithm and the Zeta form of the

constraint is not used.

The Zeta form is useful, since it is more effective at pulling molecules

towards target structures than other relative constraint forms, such as

((RMSD1 - RMSD2) - rho) = 0, where the difference in RMSDs may be well

defined, but the current structure may be far from both target structures.

Also, transitions are not limited to paths which only allow for the RMSD

to one target structure to decrease monotomically.

ZTOLerance and ZMITerations are used in the minimization scheme for

calculating the coodinates which satisfy the TMD constraint. They are

similar to the cooresponding terms in the SHAKE algorithm.

The constrained RMSD for one-target TMD is not allowed to go below zero.

Similarily, the restriction |Zeta| <= -1/(1+EXP(-CZETA*RMSD0))+1/2 is

is used, where RMSD0 is the RMS Difference between the two target

structures. Once these values are reached during dynamics, the contraint

value for RMSD (or Zeta) is held at this limiting value.

This subroutine outputs RMSD1, RMSD2, and the actual Zeta value (which

is within +/- ZTOL of Zeta0(istep)), with PRNLEV >= 5. For each dynamics

step, this is likely to print out a few times, due to the iterative scheme

used between this subroutine and the SHAKE subroutine.

Commands:

TMDINITIALIZE INRT 1 DINC -0.0003 -

ZETA CZETA 1.0 ZTOL 1.0E-8 ZMIT 1000 -

SELECT ALL END SELECT ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

OPEN READ UNIT 2 CARD NAME init.crd

READ COOR UNIT 2 CARD TAR2

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

C) Restricted perturbation - TMD method.

In this method, the coordinate displacement (perturbation) is limited

to a preset value; given this displacement, the rmsd with the target

is minimized at each step. This procedure prevents the crossing of

large energy barriers, and may increase the efficiency of the calculation.

Either the total perturbation Sum |p_i| or the maximum atomic

perturbation Max |p_i| can be restricted.

The function C = Sum ( |p_i| Cos(p_i,F_i) ) is a good indicator of

barrier crossings: when C is negative, a barrier has probably been

crossed. To reduce barrier crossings, the perturbation can be decreased

when C<0 (this will increase the simulation time).

Note that in this method, the rmsd fluctuates along the trajectory and

the length of the simulation may vary (simulations may get "stuck" when

very small perturbations are used).

See J. Chem. Phys. 122, 114903 (2005) for a more detailed discussion

of the algorithm, and a comparison of the RP-TMD method with the

standard TMD method.

Commands:

OPEN UNIT 88 WRITE CARD NAME tmd.dat

TMDINITIALIZE ITMD 88 FTMD 10 FRMS 1.2 INRT 10 MAXF 0.001 -

MAXB 0.0008 SUMP 1 -

SELE ALL END SELE ALL END

OPEN READ UNIT 2 CARD NAME target.crd

READ COOR UNIT 2 CARD TARG

CLOSE UNIT 2

DYNA RESTART LEAP TCONST TCOUPL 0.5 TREFER 300.0 ...

Examples: tmdtest32.inp (serial & parallel), and tmd_zeta.inp (serial).