Molecular Dynamics Study of Oral Medicine Delivery Rates ...
Transcript of Molecular Dynamics Study of Oral Medicine Delivery Rates ...
Abstract—This paper presents an investigation of oral
medicine delivery from a non-traditional point of view, namely,
delving into the rates of delivery attributed by two different
drug particle’s shapes in cylindrical and spherical models. The
methodology employed is molecular dynamics (MD). The
approach leads to the development of an MD computer
simulation program for the establishment of generic parametric
databases for follow on scientific comparisons on the solubility
rates of particle models in various shapes of interests. Einstein’s
fluctuation-dissipation equation is referred to in the calculation
of diffusion coefficients of given particle models within the MD
simulation. This also facilitates the study on the impact on
solubility and diffusion of oral medicine particles due to the
rates of change of radius and of mass with different weights and
sizes. In addition, the dependency of solubility on a cylindrical
drug particle’s aspect ratio as well as the control on the kinetics
of drug release based on diffusion speeds are evaluated. The
study concludes on a comparison of the solubility of a sphere
versus that of a cylinder with the same mass. The results
demonstrate that with the same given mass, the solubility of a
spherical particle is less than that of the cylindrical shape. It is
also revealed that an increase in the diffusion coefficient for a
particle with fixed mass will in turn raise its solubility. However,
the solubility and the diffusion coefficient of a particle with
fixed radius are inversely proportional.
Index Terms—Molecular dynamics, oral medicine delivery,
solubility, absorption rate.
I. INTRODUCTION
Improving solubility of oral medicine in water is always an
unresolved issue attracting substantial research efforts over
the years. However, since the drug delivery mechanism is
rather complex, demand for mathematical models and studies
that are able to predict human absorption of oral medicine has
increased dramatically in the last decade [1].
Although the physiological conditions play a major role in
determining the amount of the drug absorbed by a particular
patient, there is no universally agreed method to either
directly or indirectly take all these factors into account.
Therefore, the boundary conditions such as the shape of the
drug particle become an important design consideration [2].
In oral drug delivery, human intestine acts as a membrane
for the drug absorption. In a typical drug delivery device, a
solute such as a bioactive drug molecule upon exposure to
body fluids diffuses out. Solving Fick's second law for
diffusion results in the following short time approximation
Manuscript received December 10, 2012; revised February 28, 2013. This
work was supported in part by the TSGC under Grant NI12012.
C. J. Ginny Soong is with the Computer Science Department, University
of Texas at Tyler, Tyler, TX75799 USA (e-mail: [email protected]).
Y. J. Lin is with the Mechanical Engineering Department, University of
Texas at Tyler, Tyler, TX 75799 USA (e-mail: [email protected]).
[3], [4].
21
4
Dt
M
M t
(1)
M
M t
denotes the fraction of the total solute contained in the
membrane that is released by time t, δ is the width of the
membrane and D is the solute diffusion coefficient across the
membrane; D is thus directly related to the kinetics of solute
release. There is thus great practical and fundamental interest
in investigating how the diffusion coefficient of the solute
depends on its size and the molecular structure to better
exercise control over the kinetics of drug release.
In an unbounded solution, the resistance to Brownian
motion of the solute is equal to the hydrodynamic drag
exerted by the continuum solvent. For a spherical solute of
radius rs, this leads to the Stokes-Einstein free-solution
diffusivity D0
s
Bo
r
TKD
6 (2)
where η is the solvent viscosity. However, for a solute with
dimensions similar to that of the pore, diffusion in the
cramped space of the pore causes the molecular friction
coefficient to exceed its value in free solution [5].
Molecular Dynamic simulation is a computational method
for calculating the equilibrium and transport properties of a
multi-body system. It is assumed that the nuclear motion
obeys the laws of classical mechanics [6].
In this work at first, the dissolution characteristics of the
drug particles is investigated using the Noyes–Whitney
equation which takes into account the cylindrical shape of the
particle. We then perform the MD simulations to find the
diffusion coefficient of a given sample. The coefficient of
diffusion was then used to predict the solubility and the
fraction absorbed.
The solubility analysis is one of the most important tools
that researchers frequently use to predict the absorption of a
given particle. Most of the current drug delivery models that
are used for the solubility analysis treat the particles to be
spheres. The most common way of preparing micron sized
drug particle is by grinding and milling of previously formed
larger particles. However, the particles formed by this
process are usually not spheres but are irregular or spindle in
shape.
II. DISSOLUTION OF CYLINDRICAL PARTICLES
Most of the drugs are given in solid dosage forms, and they
Molecular Dynamics Study of Oral Medicine Delivery
Rates Attributed by Two Drug Particles’ Shapes
C. J. Ginny Soong and Y. J. Lin
International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 3, May 2013
279DOI: 10.7763/IJBBB.2013.V3.213
International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 3, May 2013
280
must dissolve before they can be absorbed into the body.In
this section the dissolution profile for a cylindrical particle is
created in terms of a rate of change of mass and of a rate of
change of radius. The Noyes-Whitney equation for
dissolution is used as a basis for predicting the dissolution of
a cylindrical particle.
A. Factors Affecting Oral Medicine Dissolution
Some of the important factors affecting drug dissolution
area) surface area, b) diffusion layer thickness, c) diffusion
coefficient, and d) drug solubility.
In this work we derive the drug diffusion co-efficient using
the methods and techniques of MD simulation. We then use
the drug diffusion coefficient to study the rate of change of
mass and size of a drug particle.
B. Related Work
In their research on the effect of particle size on
distribution and oral absorption, Hintz et al [7] established a
model, for theoretically simulating the absorption, by using
their dissolution and permeation parameters. The
Noyes-Whitney type equation was used to describe the
dissolution model. The model was based on the assumption
that the basic drug particles are in spherical shapes of the
same size. The following relationship was derived for time
dependent diffusion layer thickness given by
3
1
i
si
M
Mrh
(3)
where ir is the initial radius. sM is mass left un-dissolved.
iM is the initial mass.
The dissolution rate is given by the following equation
V
tMC
M
tMM
hr
D
dt
tdM ds
i
si
i
s )()(3)( 32
(4)
wheredM is the mass of the drug dissolved in time t and is
given as
sid MMM
is the density of the drug.
V is the volume of the solute.
sC is the saturated solubility of the drug.
C. Dissolution Model for Cylindrical Particles
The following assumptions have been made, namely,i) it is
assumed that the dissolution occurs only on the outer
surface.ii) the particle retains its shape until it is fully
dissolved.iii) as mentioned above it is assumed that the
dissolution occurs uniformly on all sides. iv) the drug
primarily dissolves in the intestine instead of the stomach or
colon. v) the majority of the drug absorption occurs in GIT
rather than in the colon. vi) the surface area of the small
intestine is 800 2cm [8]. vii) the volume of the fluid (bile
and other gastrointestinal fluids) inside the GIT is 600 mL
under fasted states [9] and viii) the effective transit time of
the drug through the intestine is 4hours [10].
D. The Mathematical Model
In this section we develop two profiles, one for change of
mass and the other for change of radius of a given cylindrical
particle.
1) The rate of change of mass
In a cylinder depicted in Fig. 1 of given radius r and
height h , the volume V and surface area S of the cylinder
are given as
hrV 2 (5)
)(2 hrrS (6)
The surface area in terms of volume is given by
22
r
VrrS
(7)
The mass dissolved dM is given by
dVdM (8)
where is the density.
At time t the dissolution profile can be depicted as seen
in Fig. 1.
Fig. 1. Cylinder with height h and radius r at t = 0.
The dissolution rate of the thn particle can be written as
L
CCsDS
dt
dM nn )( (9)
wherenM is the mass of the
thn particle, nS is the surface
area. Further derivation yields
)1(2
3/2
V
MC
M
MM
Lr
D
dt
dM ds
n
sn
i
s
(10)
which gives us the rate of change of mass with respect to
time.
2) The rate of change of radius
Since
dt
dV
dt
dM
(11)
hence,
)(2 rr
h
tt
hh
r2
International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 3, May 2013
281
dt
drr
dt
rd
dt
dM
13
)( 23
(12)
Assuming that at 0t the initial concentration 0C , so
)1)((3
2
sC
L
D
dt
dr (13)
Since
V
MC d (14)
further derivation yields
])[1(3
2 33
sis rrCL
D
dt
dr
(15)
Referring to Equations (10) and (15), it can be seen that the
solubility of drug is directly dependent on the diffusion
coefficient and inversely proportion to the thickness of the
diffusion layer.
III. THE MD SIMULATION
In this section the methods and procedures utilized to
calculate the diffusivity of glucose-water mixture using the
mean square displacement data for the mixture, as well as the
various steps in the process of MD simulation are to be given.
A. Initialization
The initialization of the MD process consists of mainly
two parts: initialization of parameter and initialization of
atoms. Initialization of parameters consists of defining the
system of units and numerical algorithm used. The
initialization of atoms deals with the position, velocity and
other kinematics properties of the atoms.
B. The Predictor Corrector Method
In this paper we have adopted the Gear’s predictor
corrector method to evaluate the forces. The main advantage
of using this method is that it allows us to solve a
second-order ordinary differential equation (ODE) without
the need to convert it into a first-order ODE. Therefore, this
analysis can be done in a single step [11].
C. The Initial Condition
The location of a particle is measured with respect to a
space fixed frame and its initial value being assigned
according to the lattice structure. Fig. 2 shows the location
with a space-fixed frame.
The Initial velocities may be randomly assigned but they
have to follow physical criteria. Since no external force acts
on the system, the total linear momentum should be
conserved.
The values of initial acceleration can be computed from
Newtonian physics by using the initial positions and
velocities.
D. The Periodic Boundary Condition
The unit cell volume described above is the basis for our
simulation. A system containing N such atoms for the
simulation in a given volume V , can be visualized as the
same initial cube volume, being periodically repeated over
many times in all the three dimensions so that it can have
enough volume to contain all of the molecules. Since the
initial unit cell volume is being replicated, the replica cells
are called image cells[12].
Fig. 2. An atom location within a space-fixed frame.
Each image cell is an exact copy of the unit cell and
contains the same number of atoms as the initial unit cell. The
atoms in the image cell, called image atoms, are again a
replica of the atoms in the unit cell.
Since the system has no external force and energy and
mass are conserved a new atom must enter the simulation
cube with same velocity, momentum and potential energy.
Fig. 3 and Fig. 4 show us the simulation of such a process.
Fig. 3. Initial conditions for an image cell at t = 0 [12].
The value of the calculated diffusion coefficient is
comparable to the published value 10103.1 2s-1 [13]. This
L
L
X
Y
(0,0,0)
atomi
ir
Z
Fig. 4. Simulation for an image cell at t = Δt [12].
International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 3, May 2013
282
validates the value of the diffusion coefficient that was
calculated using our simulation code.
IV. MD SIMULATION AND RESULTS
Referring to Fig. 1, the average initial radius of the particle
was assumed to be 100μm [4]. The rate of change of mass for
different values of ( = 1, 2, 3) is plotted in Fig. 5.
Fig. 5. Change of mass vs. time.
The graph covers for T = 4hrs. Here we noted that the
particle doesn’t get dissolved completely in the given time
frame. This agrees with our perception of the human
digestion process that the food absorption occurs in the colon
and large intestine after it passes through the small intestine.
Also we notice that as we increase the value of
( rh / ), the rate of change of mass is increased. This means
that thinner needle like shape is more soluble than a thicker
plate/disc like shape which is consistent with our notion of
using a cylindrical model.
Fig. 6 presents the concentration of glucose vs. time.
Fig. 6. Concentration of glucose vs. time.
As expected the concentration of glucose increases rapidly
when the size of the particle is small Fig. 7 & Fig. 8 show
additional results regarding the rate of change of radius
versus time and initial radius. Finally we compare the rate of
change of mass vs. time for a spherical particle vs. a
cylindrical particle. From Fig. 9 we see that for a given mass
the cylindrical shaped particle dissolves at a much higher rate
than a spherical shaped particle. This implies that our novel
cylindrical drug particle model has merit in contributing to
speeding up drug delivery.
Fig. 7. Change of radius vs. time
Fig. 8. The rate of change of radii of particles with different initial radii.
Fig. 9. The change of mass vs. time for a spherical particle (dashed) vs. a
cylindrical particle (solid).
V. CONCLUSIONS
In this paper, an oral medicine delivery issue regarding
drug molecule’s geometrical models was investigated. To
facilitate the investigation of drug particle dissolution profile,
a dissolution model for cylinders was developed first.
Then, the diffusion coefficient of glucose was derived and
computational modeling based on the derivation was
conducted. This approach enabled us to make comparisons
between the dissolution profile of a cylinder and a sphere
with the same given mass. When performing comparisons,
the fourth order Runge-Kutta and the Gear’s Predictor
Corrector algorithms were employed in obtaining drug
particle’s dissolution profiles.
0 2000 4000 6000 8000 10000 12000 14000 160004
5
6
7
8
9
10x 10
-5
Time (seconds)------>
M -
----
->
M vs Time
0 2000 4000 6000 8000 10000 12000 14000 160000
1
2
3
4
5
6
7
8
9
10
Time (seconds)------>
Glu
cose
con
cent
ratio
n %
---
--->
Glucose concentration vs Time
0 2000 4000 6000 8000 10000 12000 14000 160001
2
3
4
5
6
7
8
9
10x 10
-5
Time ------>
Radiu
s -
----
->
Radius vs Time
0 2000 4000 6000 8000 10000 12000 14000 160005
6
7
8
9
10
11
12
13x 10
-5
Time ------>
Rad
ius
----
-->
Radius vs Time
0 2000 4000 6000 8000 10000 12000 14000 160006.5
7
7.5
8
8.5
9
9.5
10x 10
-5
Time (seconds)------>
Mass
----
-->
Mass vs Time
The simulation results showed that for a given mass the
dissolution occurs at a faster rate when the shape is in a
needle or cylindrical profile. For a cylinder with a given
volume the surface area is maximum when the height of the
cylinder is the same as the radius (δ =1).This entailed a
contradiction to our finding since with more surface contact
area, it is expected to have higher dissolution rate at δ =1 than
δ =3. This is because the rate of change of mass is inversely
proportional to radius. The study also demonstrated that for a
given volume as we increased the value of δ (1, 2, 3...), the
radius decreased.
Moreover, the rate of change of radius results was found to
be quite consistent with the anticipated trend. It was observed
that as we increased δ (1, 2, 3, etc.), the mass of the particle
was increased concurrently resulting in slower dissolution. In
the case of two particles with the same initial value of δ but
different initial radii, the dissolution is faster for the particle
with a smaller radius. Last but not the least, it was evidenced
from both the dissolution plots we obtained that they were
approximately linear instead of decreasing exponentially.
This may be due to the fact that the effects of absorption of
drug into the body and rising in the concentration of drug in
the surrounding medium were completely neglected. Another
important point being observed in this work was that for the
same initial mass, a cylindrical particle dissolves at a faster
rate than a spherical one. What is more, we know from the
latest manufacturing techniques for drug powders that the
shape of the particles is usually irregular [14]. This goes back
to our original question, namely, why do we always model
drug particles as spheres in most of the past research? Should
we think out of the box and start formulating oral medicine
particles in different shapes such as spindles or cylinders?
That’s why this work.
ACKNOWLEDGMENT
The authors would like to acknowledge the contribution to
the initial phase of this project by Mr. B. Nagirreddy. His
initial development efforts in the MD simulations facilitate
the follow on computational endeavors of the work.
REFERENCES
[3] W. W. Wood, “Early history of computer simulation in statistical
mechanics,” J. Comp. Simu., vol. 22, pp. 2-14, 1986.
[4] B. Nagireddy, “An effective drug delivery process using a novel
cylindrical particle model justified by molecular dynamics simulation,”
M. S. Thesis, University of Akron, Akron, OH, 2007.
[5] M. Bisrat and C. Nystrom, “Physicochemical aspects of drug release.
VIII. The relation between particle size and surface specific dissolution
rate in agitated suspensions,” Int. J. Pharm., vol. 47, pp. 223–231,
1988.
[6] M. Mosharraf and C. Nystrom, “The effect of particle size and shape on
the surface specific dissolution rate of micro-sized practically insoluble
drugs,” Int. J. Pharm., vol. 122, pp. 35–47, 1995.
[7] R. J. Hintz and K. C. Johnson, “The effect of particle size distribution
on the dissolution rate and oral absorption,” Int. J. Pharm., vol. 51, pp.
9-17, 1989.
[8] J. B. Dressman, “Comparison of canine and human gastrointestinal
physiology,” Pharm. Res., vol. 3, pp. 123-131, 1986.
[9] J. B. Dressman and C. Reppas, “In vitro-in vivo correlations for
lipophilic, poorly water-soluble drugs,” Eur. J. Pharm. Sci, vol. 11, no.
2, pp. S73-S80, 2000.
[10] B. Davies and T. Morris, “Physiological parameters in laboratory
animals and humans,” Pharm. Res., vol. 10, pp. 1093-1095, 1993.
[11] C. W. Gear, “The Numerical Integration of Ordinary Differential
Equations of Various Orders,” Argonne National Laboratory,
ANL-7126, 1966.
[12] D. C. Rapaport, The Art of Molecular Dynamics Simulation, 2nd ed.,
Cambridge University press, pp. 15-16.
[13] A. Vicente, M. Dluhý, E. Ferreira, M. Mota, and J. Teixeira,
“Masstransfer properties of glucose and O2 in Saccharomyces
cerevisiae flocs,” Biochemical Engineering Journal, vol. 2, issue 1, pp
35-43, Sep. 1998.
[14] T. L. Rogers, K. P. Johnston, and R. O. Williams III, “Physicalstability
of micronized powders produced by spray-freezing into liquid (SFL) to
enhance the dissolution of an insoluble drug,” Pharm Dev Technol., vol.
8, no. 2, pp. 187-97, 2003.
C. J. Ginny Soong was born in Taipei, Taiwan in
1958. Shereceived her BS degree from National
Taiwan University in Taipei, Taiwan in 1981 and the
Ph.D. degree from University of Illinois in Chicago,
USA in 1986 with her study focused on biochemistry.
She also earned a MS degree in computer science in
1999 from The University of Akron in Akron, Ohio,
USA.
After receiving her Ph.D. degree, she worked at
Northwestern University as a Postdoctoral Fellow, at
Cleveland Clinic Foundation as Scientist and Case Western Reserve
University as Project and Lab Manager. Her computer science training and
degree led her stepping into industry and was employed by Bridgestone
North America Company for 11 years as Senior Systems Analyst. She joined
The University of Texas at Tyler in 2010 as an Associate Professor in the
Computer Science Department where she has been contributing mainly to
the computer information system and bioinformatics programs. Her research
interests span in the areas of database and information systems,
bioinformatics, bio-computation, and engineering applications.
Dr. Soong has authored and co-authored many technical papers resulted
from her research outcomes including her current interests in
bio-computational efficacy study of drug delivery, as well as molecular
dynamics simulation program development. Recently, Dr. Soong was
awarded a NASA grant by Texas Space Grant Consortium for a support to
furthering her computing in life science research work. She has also received
the 2011 Outstanding Professor of Computer Science award in the College of
Engineering and Computer science at UT Tyler. Dr. Soong is a member of
ASEE, ACM and ISCA.
Y. J. Lin was born in Hsinchu, Taiwan in 1955.
Hereceived his BS degree from National Tsinhua
University in Hsinchu, Taiwan in 1978 and the Ph.D.
degree from University of Illinois in Chicago, USA in
1988 with his study focused on mechanical
engineering.
Since he graduated, he has worked in
industry,government research institute and
universities. He first worked as a postdoctoral research
associate at UI Chicago on a VA Hospital-sponsored project. In 1988, he
joined the mechanical engineering faculty at The University of Akron as an
Assistant Professor. He became the ME Graduate Program Coordinator in
1998 at UA. In 2007, He was appointed as the Innovative Learning Research
Director at NASA Safety Center in Cleveland, Ohio. He joined The
University of Texas at Tyler in 2008 as Professor and Chair of mechanical
engineering. In an international scope, He has been invited to serve on the
University External Advisory Board as Academic Programs Examiner in
Engineering and Science of Multimedia University in Malaysia since
2008.In addition, he has been serving on the editorial advisory boards of
Journal of Assembly Automation, International Journal of Industrial Robot,
and Cybernetics and Engineering Systems Open Journal for many years.His
research interests span in the areas of CAD/CAM/CAE, robotics,
mechatronics, manufacturing automation, smart materials and structural
health monitoring, bioCAD and biofabrication, and nano-device
development.
Dr. Lin has authored and co-authored over ninety technical papers since
the beginning of his academic career. Dr. Lin has received a Management
Excellence certificate in 2009 from NASA Safety Center, and was the
recipient of Outstanding Professor of Mechanical Engineering Award in
2010 in the College of Engineering and Computer Science at UT Tyler. Dr.
Lin is a member of ASME, IEEE, ISCA and ASEE.
Author’s formal
photo
International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 3, May 2013
283
[1] T. Hou et al., “Recent advances in computational prediction of drug
absorption and permeability in drug discovery,” Current Medicinal
Chemistry, vol. 13, pp. 2653-2667, 2006.
[2] M. Tubic et al., “In silico modeling of non-linear drug absorption for
the p-gp substrate talinolol and of consequences for the resulting
pharmacodynamic effect,” Pharmaceutical Research, vol. 23, no. 8,
August 2006.