Calculating Free Energies Using Adaptive Biasing Force Method

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 14

Calculating free energies

using Adaptive biasing force


method

Joanne Chiu
Department of Mechanical Engineering,
Stanford
2009/6/3
Introduction
Free Energy
Determination of the most stable
conformation
Computing the reaction rate
Determination of probabilities at equilibrium
First step in building a course grained model
of protein
Examples
Fuel Cell reaction rate
Protein-ligand binding and ion partitioning
across the cell membrane

2009/6/3
However
Free Energy Calculation
Needs first and second derivatives of the
order parameter with respect to Cartesian
coordinates
Expensive computation and long
calculation time
Accuracy

2009/6/3
Thermodynamics Integration
2009/6/3

o
c
c

c
c
=
=

}
J
kT
U
d
dA
dq q e kT A
q U
ln
) (
)) ( ( ln ) (
) (
Lecture 08
Assume there is no momentum in the
Hamiltonian of the system.
J is already the first derivative of q, thus
we need to get the second derivative of
q.
Adaptive Biasing Force
Based on computing the mean force
on
A free computation using the equation:
2009/6/3

( )
2
1
1
) (
|
|
.
|

\
|
c
c
=
=

k
k
k
q m
m
m
dt
d
d
dA


wher
e
( )
i i
i
q m
dt
d
dq
dU
=
The governing equation of
motion:
Adaptive Biasing Force (Cont.)
Applying a biasing force
2009/6/3
V
b
F

Take a long time to go


over the energy
barrier
Much easier to reach
another state
Local minimum
Global minimum
Numerical Results
2009/6/3
Glycophorin A: key interactions
between the two helical segments



Simulation Setup
Two molecules connected with a
spring and the potential is given by:




Because there is no long range force,
free energy should be the same as
potential energy.
2009/6/3
( )
2
2 1 0
2
x x
k
U =
1
x
2
x
Free energy calculation
Thermodynamics Integration
2009/6/3
( )
( )
( )
( )
( )
( )
1 2
2
1 2
2
2
1 2
2
1 2
0
2
1 2
0
2
1 2
0
0
) (
) (
ln ) (
ln ) (
x x x
dq e e
dq x x x e e
dx
x dA
dx e e kT x A
dx x x x e kT x A
x x x
U
x x x
U
x x x
U
U
+ =
+
=
=
+ =
}
}
}
}
+

o
o
o
o
|
|
o
|
|
o
|
|
|
Replace
impulse by
Gaussian
distribution
Thermodynamics Integration







Calculated free energy (U) will be slightly
different from the potential energy(U
0
).

2009/6/3
( )
( )
( )
2
1 2 0
2
2
2
ln ) (
ln ) (
2
1 2 0
2
1 2
0
x x x U U
dx e kT x A
dx e e kT x A
x x x U
x x x
U
+ + =
=
=
}
}
|
.
|

\
|
+ +
+

o
o
|
o
|
|
Free energy calculation
Adaptive Biasing Force
2009/6/3
( ) ( ) ( )
( )
1 2 1 2
2
2
2
1
2
1
1 2
1 2 1 2
2
1
2
1 ) (
2
1
) (
F F v v
dt
d
d
x dA
x x q m
m
v v m
dt
d
m
dt
d
d
x dA
v v x x
k
k
k
=
|
.
|

\
|
=
=
|
|
.
|

\
|
c
c
+
|
|
.
|

\
|
c
c
=
|
|
.
|

\
|
c
c
=
= =
= =

Assume the mass of two particles are 1:


0 1 2 3 4 5 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Reaction Coordinate
E
n
e
r
g
y


Potential
ABF
ABF simulation result
2009/6/3
( )
2
2 1 0
2
1
x x U =
2
1
= x 5
2
= x
Potential
Match
Conclusion
ABF is easier to implement and only
need to evaluate the first derivatives
and the derivatives with time.
With applying biasing force, the phase
space can be explored more
completely.
The free energy converges faster than
Thermodynamics Integrator.
2009/6/3
Future Work
Comparing the simulation results of TI
and ABF.
Changing the potential to create a low/
relatively high energy barrier and
comparing the results .
Adding biasing force to see if the
accuracy is improved.
Adding long range force (more
molecules) to calculate the new free
energy by applying both methods.
2009/6/3

You might also like