Structural Screening of HIV-1 Protease/Inhibitor

Docking by Non-parametric Binomial Distribution Test

Piyarat Nimmanpipug1, Vannajan Sanghiran Lee1, Jeerayut Chaijaruwanich2,

Sukon Prasitwattanaseree3, and Patrinee Traisathit3

1
Deparment of Chemistry, Faculty of Science, Chiang Mai University,
Chiang Mai 50200, Thailand
npiyarat@chiangmai.ac.th, vannajan@chiangmai.ac.th
2
Department of Computer Science, Faculty of Science, Chiang Mai University,
Chiang Mai, 50200, Thailand
jeerayut@cs.science.cmu.ac.th
3
Department of Statistics, Faculty of Science, Chiang Mai University,
Chiang Mai, 50200, Thailand
s.prasit@chiangmai.ac.th, patrinee@chiangmai.ac.th

Abstract. Attempts have been made to predict the binding structures of the
human immunodeficiency virus-1 protease (HIV-1Pr) with various inhibitors
within the shortest simulation time consuming. The purpose here is to improve
the structural prediction by using statistical approach. We use a combination of
molecular docking and non-parametric binomial distribution test considering
the combination of binding energy, hydrogen bonding, and hydrophobic-
hydrophilic interaction in term of binding residues to select the most probable
binding structure. In this study, the binding of HIV-1Pr and two inhibitors:
Saquinavir and Litchi chinensis extracts (3-oxotrirucalla-7, 24-dien-21-oic acid)
were investigated. Each inhibitor was positioned in the active site of HIV-1Pr
in many different ways using Lamarckian genetic algorithm and then score each
orientation by applying a reasonable evaluation function by AutoDock3.0
program. The results from search methods were screened out using non-
parametric binomial distribution test and compared with the binding structure
from explicit molecular dynamic simulation. Both complexes from statistical
selected docking simulation were found to be comparable with those from X-
ray diffraction analysis and explicit molecular dynamic simulation structures.

Keywords: HIV-1 protease, Docking, Binomial Distribution Test, Saquinavir,

3-Oxotirucalla-7, 24-Dien-21-Oic Acid.

1 Introduction
Human Immunodeficiency Virus-1 Protease (HIV-1Pr) has an important role in the
replication of HIV-1 by processing two precursor polyproteins, Pr55gag and
Pr160gag-pol, into structural proteins and replication enzymes. This enzyme is a type
of aspartic protease, and has a C2 symmetric homodimer. Each monomer consists of
99 amino acid residues and contributes a loop structure containing the active site triad

S. Hochreiter and R. Wagner (Eds.): BIRD 2007, LNBI 4414, pp. 119–130, 2007.
© Springer-Verlag Berlin Heidelberg 2007
120 P. Nimmanpipug et al.

Asp25(25’)-Thr26(26’)-Gly27(27’). A cavity for the insertion of the substrate is

formed by the loop structures containing the active site triads and flap regions (flaps)
which are presumably related to the entry and affinity of the substrate to the enzyme
[1-3].
Up to now, much research effort has been focused on computational methods for
the prediction of this difficult-to-obtain structural information. In general, docking
study has been widely used to generate the enzyme-substrate structures. The
candidate from docking typically selected from the lowest energy structure. However,
this criteria of energy screening is not always give the correct binding structure,
especially for the highly flexible ligand [4]. The purpose here is to improve the
structural prediction by using statistical approach; non-parametric binomial
distribution test to include the essential amino acids in the binding pocket.
For preliminary study, the prediction was done for the known structure from X-ray
diffraction analysis in the case of HIV-1Pr – Saquinavir complex, a peptidomimetic
inhibitor of HIV protease in order to proof the efficiency of this method. In addition,
since several research groups develop a number of HIV-1Pr inhibitors from natural
product but its HIV-1Pr complex is still unknown, efforts have been also done for the
unknown HIV-1Pr – inhibitor complex. For our investigation, the binding structure
of HIV-1Pr and Litchi chinensis seed extracts (3-oxotrirucalla-7, 24-dien-21-oic acid)
was predicted and compared with the simulated from molecular dynamic (MD). The
inhibitory activity of 3-oxotrirucalla-7, 24-dien-21-oic acid is reported in IC50 of 20
mg/L by Chao-me Ma and co-workers in 2000 [5] and this compound was found in
chemical constituents in seed of Litchi chinensis by Pengfei Tu and co-workers in
2002 [6]. The extracts from such a natural product still have relatively high IC50, in
order to get the lower IC50 value the molecular modeling is, therefore, a very useful
tool.

2 Methods
2.1 Preparation of HIV-1Pr

The initial HIV-1Pr structure was obtained from HIV-1Pr with saquinavir at 2.3 Å
resolution (1HXB entry in PDB database) [7]. The structural water and inhibitor from
the selected PDB database were then removed in the preparation of the HIV-1Pr.
Hydrogen atoms were added in the structure and a short minimization run was
performed to remove any potentially bad contacts with the program package
AMBER, Version 7 [8,9]. A cutoff distance at 12 Å for non-bonded pair interaction
was used in the minimizations. Then the obtained structure model was protonated at
catalytic Asp25.

2.2 Preparation of Inhibitors: Saquinavir and 3-Oxotirucalla-7, 24-Dien-21-Oic

Acid

Geometry of saquinavir (compound 1) and 3-oxotirucalla-7, 24-dien-21-oic acid

(compound 2), as shown in Fig. 1, were optimized using semi-empirical calculation,
AM1 in the program package Spartan’04.
 Structural Screening of HIV-1 Protease/Inhibitor Docking 121

3-oxotirucalla-7,24-dien-21-oic

saquinavir acid

HOOC CH3
H CH3
O N CH3 H CH3
O CH3
H CH3 CH3
N
N N N
H H H CH3
O O OH
H O
NH2 H
H3C CH3

Flexible parts (in dash line) in molecular docking

O
H CH3
O N CH3
O
HO
CH3 CH3
H CH3
H CH3
N
N N N
H H CH3
O O OH
H H CH3
NH2
O
H
H3C CH3

Fig. 1. Chemical structures of saquinavir (compound 1) and 3-oxotirucalla-7, 24-dien-21-oic

acid (compound 2)

2.3 Preparation of HIV-1Pr – Inhibitor Complex by Molecular Docking and

Molecular Mechanics Methods

Each structure of HIV-1Pr – inhibitor was obtained by docking compound 1 and

compound 2 to HIV-1Pr, respectively. HIV-1Pr was kept rigid while inhibitors are
flexible (21 flexible torsions in compound 1 and 13 flexible torsions in compound 2),
as shown in dash line in Fig.1, and Gasteiger charges were used. Grid maps have been
calculated using the module AutoGrid in AutoDock 3.0 program [10-12] for protease
structure. The center of grid was assigned at the center of the cavity, between two
catalytic aspartics. A number of grid points in x y z, is 60 60 60 with the spacing
0.375 Å. This parameter set covers the active site extensively and let the ligand move
and explored the enzyme active site without any constraints regarding the box size.
The inhibitor was positioned in the active site of HIV-1 protease in many different
ways using Lamarckian genetic algorithm (LGA). The solvation effect was also
122 P. Nimmanpipug et al.

included in this docking study via atomic solvation parameters for each atom in the
macromolecule of AutoGrid 3.0 module. Each hydrogen bonded to N and O atom in
hydrogen grid maps was calculated using self-consistent hydrogen bonding 12-10
parameters in the AutoGrid.

2.4 Molecular Dynamic Simulations

The energy minimized conformation of HIV-1Pr – compound 2 from previous

docking calculation was used as starting structure. The molecular mechanics
potential energy minimizations and MD simulations were carried out with the
program package AMBER, Version 7 [8,9]. Calculations were performed using the
parm99 force field reference for HIV-1Pr and compound 2. The atom types for
compound 2 were assigned by mapping their chemical properties (element,
hybridization, bonding schemes) to AMBER atom type library and the Gasteiger
charges were used.
The enzyme-inhibitor complex was solvated with WATBOX216 water model
(9298 water molecules) in cell dimension 61.06 x 66.56 x 75.88 Å [3] and treated in
the simulation under periodic boundary conditions. All of MD simulations reported
here were done under an isobaric-isothermal ensemble (NPT) using constant pressure
of 1 atm and constant temperature of 298 K. The volume was chosen to maintain a
density of 1 g/cm3. A cutoff distance (12 Å) was applied for the non-bonded pair
interaction. Three sodium and eight chloride ions were added to neutralize and buffer
the system. The potential energy minimizations were performed on the systems using
the steepest descent method and followed by conjugate gradient method. After a short
minimization, molecular dynamic simulations were performed to get an equilibrium
structure. The temperature of the whole system was gradually increased by heating to
298 K for the first 60 ps, and then it was kept at 298 K from 61-1800 ps. The
temperature was kept constant according to the Berendsen algorithm [13]. All
trajectories were kept and analyzed in detail.

2.5 Non-parametric Binomial Distribution Test

The enzyme-inhibitor (compound 1 and compound 2) complexes from 100 runs

molecular docking were collected in order to explore all probable binding structures.
The vicinity residues in the binding pocket within a trial distance (3 Å) measured
from the inhibitor were selected as vital amino acids for enzyme-inhibitor complex
formation. As shown in Fig. 2, the average hydrogen bond formation of total 100 runs
is only 0-1 bond but 1-2 hydrogen bonds are found for the 25% lower of energy
ranking. Therefore, only the lower quartile structures will be taken into consideration
here. The present or absent of amino acid was treated as binomial variable x. In this
case, the probability of present or absent of amino acid was assumed to be 50%
equally. Binomial Distribution Test–value (BDT-value), P+(x) and P-(x), for the
observed number of present (r+) and absent (r-) can be calculated directly from this
binomial distribution:
 Structural Screening of HIV-1 Protease/Inhibitor Docking 123

P + (x) = P(x >= r +

when p =1/2)
N
⎛N⎞ r+
∑ ⎜ r ⎟ ( 0 .5 )
+
= +
( 0 .5 ) N − r (1)
⎝ ⎠
r +

P (x) = P(x >= r - when p =1/2)

-

If P+(x) < α amino acid x will be present, on the other hand, if P-(x) < α amino acid x
will be absent. In this case, α = 0.05 indicating that if more than 95% of cases were
found then the present of absent of the residue will be accepted.

40

35

30

25
Frequency

20

15

10

0
0 1 2 3
Number of hydrogen bond

Fig. 2. Comparison of hydrogen bond distribution between total docking structures (gray) and
the 25% lower of energy ranking structures (white)

3 Results and Discussion

3.1 Non-parametric Binomial Distribution Test of HIV-1Pr – Inhibitor

Complexes

As in most of the existing methods, the protein-ligand complex was composed of a

rigid protein structure and a flexible ligand. A flexible ligand has three translational
degrees of freedom, three rotational degrees of freedom and one dihedral rotation for
each 2 rotable bond. The docking search is computed over a 6+n dimensional space
where n is the number of rotable bonds in the ligand. Fig. 3 shows the combined
statistical-docking algorithm with the docking simulations. One docking trial allows a
protein and a ligand to dock into its binding site, a docking attempt consists of a series
of independent structures or the so-called clusters. For each cluster in the lower
quartile, all vicinity residues (acidic, basic, polar, and non-polar groups) were tested
124 P. Nimmanpipug et al.

docking an inhibitor with HIV-1Protease using LGA

100 run (selected residues within 3 Å)

dividing structures into 4 energy levels

the lowest energy level

non-parametric Binomial Distribution Test

potential candidate

Fig. 3. The statistical-based protein-ligand docking procedure

Table 1. Criteria for screening crucial residues of HIV-1Pr-compound 1 complex

Present residues Chain A: Asp25, Ala28, Chain B: Asp25’, Gly27’,

Gly48, Gly49, Ile50 Ala28’, Asp29’,

Gly48’, Gly49’,

Ile50’

Absent residues Chain A: Leu23, Met46, Chain B: Thr26’, Leu76’,

Leu76, Thr80, Thr80’, Arg87’

Val82, Asn83, Ile84

 Structural Screening of HIV-1 Protease/Inhibitor Docking 125

for the possibility of appearance. The criteria for essential-inessential residues were
concluded and reported in Table 1 (HIV-1Pr – compound 1 complex) and Table 2
(HIV-1Pr – compound 2 complex). Using these criteria (Table 1-2), the inessential
residues will be neglected.

Table 2. Criteria for screening crucial residues of HIV-1Pr-compound 2 complex

Present residues Chain A: Asp25, Gly27, Gly48 Chain B: Asp25’, Gly27’,

Ala28’, Asp29’,

Asp30’

Absent residues Chain A: Val32, Ile47, Val82, Chain B: Val32’, Ile47’

Ile84

Fig. 4. The structure of HIV-1Pr – compound 1 from (a) X-ray crystallography, (b) the
minimized structure after molecular docking, and (c) the statistical selected docking structure.
The binding residues were shown in stick.
126 P. Nimmanpipug et al.

3.2 Structural Comparison Between Docking Statistical Test, and X-ray

Crystallographic Structure of HIV-1Pr – Compound 1 Complex

The X-ray studied structure of HIV-1Pr complexed with compound 1 is shown in

Fig. 4a. The inhibitor was bound with protease enzyme using both catalytic and flap
regions at Asp25, Asp25’, and Gly48 respectively. Compared with experimental
investigation, the lowest energy structure from docking (Fig. 4b) give the
inappropriate binding structure in both its orientation and the binding sites. After
screening using amino acid residues in our non-parametric binomial distribution test
criteria, the more appropriate binding residues (Asp25’ and Gly48) were observed
in Fig. 4c and the more similar inhibitor structure orientation was also found as
shown from the superimposition of all atoms in enzyme–compound 1 complexes in
Fig. 5.

(a) (b)

Fig. 5. Superimposition of all atoms in enzyme–compound 1 complexes (a) between the X-ray
crystallographic structure (black) and the lowest energy docking structure (gray) (b) between
the X-ray crystallographic structure (black) and the statistical selected docking structure (gray)

3.3 Structural Comparison Between Docking and Explicit Water MD

Simulation of HIV-1Pr – Compound 2 Complex

Firstly, the 100 run docking calculations were used in a prior prediction of binding
affinities and to simulate a crystal geometry as a candidate of the ligand/protein
complex. However, from the obtained lowest energy minimized structure, the
direction of OH group in compound 2 points to the flap site (O:Gly48’) instead of
catalytic site of enzyme. The reason for this observation may due to the highly
flexibility of ligand and rigidity of HIV-1Pr leading to the inappropriate binding
structure from docking study. One way to solve problem is to take the flexibility of
both enzyme and ligand into account. Therefore MD simulation was performed using
this docking structure. The total energy (ET), Potential Energy (EP) and Kinetic
energy (EK) over simulations from 0-1800 ps were investigated and the equilibrium
was obtained after 600 ps. After the equilibrium stage, compound 2 was found to bind
to the enzyme at catalytic site with more than 89% hydrogen bond formation with
Asp25 and Asp29. As shown in Table 3, the energy minimized structure obtained
from molecular dynamics simulations directs CO and OH group of compound 2 to the
catalytic site of enzyme, O:Asp25 and N:Asp29, respectively. The structure of
 Structural Screening of HIV-1 Protease/Inhibitor Docking 127

Table 3. Possible hydrogen bond between compound 2 and HIV-1Pr after reach equilibrium in
explicit MD simulation

% H-bond formation Average distance / Å

Case 1: HIV-1Pr as donor and compound 1 as acceptor

Asp25:OD2H - compound1:O78 1.98 3.067

Asp29:NH - compound 1:O21 94.70 2.968

Asp30:NH - compound1:O21 1.73 3.211

Ala28’:NH - compound 1:O78 3.27 3.054

Case 2: compound 1 as donor and HIV-1Pr as acceptor

compound 1:O78H79 – Asp25:OD1 0.04 3.168

compound 1:O78H79 – Asp25:OD2 89.18 2.946

compound 1:O78H79 –
12.98 2.929
Asp25’:OD1

compound 1:O78H79 –Asp25’:OD2 2.00 2.949

HIV-1Pr – compound 2 from docking method and MD simulation were compared in

Fig. 6. The binding residues as discussed above were labeled and shown in stick.
Root mean square deviation (RMSD) of all atoms for complex of enzyme-
compound 1 in docked - MD structure and statistical docked-MD structure are 3.63 Å
and 3.84 Å respectively, indicating somehow conformational difference and
molecular displacement in inhibitor structure after both docking results (Fig.7).
However binding site of the statistical docked structure is the same as equilibrated
MD structure as shown in Fig. 8; both structures bound to Asp25 and Asp29.
128 P. Nimmanpipug et al.

Fig. 6. The structure of HIV-1Pr – compound 2 from docking method (top) and the minimized
final structure after explicit MD simulation (below). The binding residues were shown in stick.
 Structural Screening of HIV-1 Protease/Inhibitor Docking 129

(a) (b)
Fig. 7. Superimposition of all atoms in enzyme–compound 2 complexes (a) between the final
MD simulations structure (black) and the lowest energy docking structure (gray) (b) between
the final MD simulations structure (black) and the statistical selected docking structure (gray)

O21
O21
Asp29
ASP29 Asp29
O78
O78

ASP25
Asp25 Asp25

Fig. 8. Binding structure of enzyme-compound 2 complex from statistical selected docking

structure (left) compare to that from MD simulation (right). The binding residues were shown
in stick.

4 Conclusions
A combination of molecular docking and non-parametric binomial distribution test
considering the binding energy, hydrogen bonding, hydrophobic-hydrophilic
interaction in term of residue binding to select the most probable binding structure can
reduce simulation time consuming. In this study, binding mode and orientation in
HIV-1Pr – inhibitors (saquinavir and 3-oxotrirucalla-7, 24-dien-21-oic acid)
complexes were found to be comparable with those from X-ray diffraction and
explicit molecular dynamic simulation structures. Our statistical selected docking
simulation provides the significant improvement of enzyme-inhibitor binding
prediction.

Acknowledgements. We gratefully acknowledge a generous grant of computer time

from the Computational Simulation and Modeling Laboratory (CSML), Department
of Chemistry, Faculty of Science, Chiang Mai University, Chiang Mai, Thailand, and
Computational Chemistry Unit Cell (CCUC), Department of Chemistry, Faculty of
130 P. Nimmanpipug et al.

Science, Chulalongkorn University, Bangkok, Thailand, where the simulations were

performed. This project was supported by the Commission on Higher Education and
Thailand Research Fund (TRF), Thailand.

References
1. Lappatto, R., Blundell, T., Hemmings, A., Overington, J., Wilderspin, A., Wood, S.,
Merson, J. R., Whittle, P. J., Danley, D. E., Geoghegan, K. F., Hawrylik, S. K., Lee, S. E.,
Scheld, K. G., Hobart, P. M.: X-ray analysis of HIV-1 proteinase at 2.7 A resolution
confirms structural homology among retroviral enzymes. Nature, Vol. 342. (1989) 299-302
2. Navia, M. A., Fitzgerald, P. M.: D., McKeever, B. M., Leu, C. T., Heimbach, J. C.,
Herber, W. K., Sigal, I. S., Darke, P. L., Springer, J. P.: Three-dimensional structure of
aspartyl protease from human immunodeficiency virus HIV-1. Nature, Vol. 337. (1989)
20-615
3. Okimoto, N., Tsukui, T., Kitayama, K., Hata, M., Hoshino, T., Tsuda, M., J. Am. Chem.
Soc., 122, 5613 (2000)
4. Hassan, S. A., Gracia, L., Vasudevan, G., Steinbach, P. J.: Methods Computer simulation
of protein-ligand interactions: challenges and application. Methods Mol. Biol, Vol. 305.
(2005) 92-451
5. Ma, C.-m., Nakamura, N., Hattori, M., Kakuda, H., Qiao, J.-c., Yu, H.-l.: Inhibitory effects
on HIV-1 protease of constituents from the wood of Xanthoceras sorbifolia. . J. Nat. Prod.,
Vol. 63. (2000) 42-238
6. Tu, P., Luo, Q., Zheng, J., Zhongcaoyao, 33, 300 (2002)
7. Krohn, A., Redshaw, S., Ritchie, J. C., Graves, B. J., Hatada, M. H.: Novel binding mode
of highly potent HIV-proteinase inhibitors incorporating the (R)-hydroxyethylamine
isostere. J. Med. Chem, Vol. 34. (1991) 2-3340
8. Case, D. A., Pearlman, D. A., Caldwell, J. W., Cheatham III, T. E., Wang, J., Ross, W. S.,
Simmerling, C. L., Darden, T. A., Merz, K. M., Stanton, R. V., Cheng, A. L., Vincent, J.
J., Crow-ley, M., Tsui, V., Gohlke, H., Radmer, R. H., Duan, Y., Pitera, J., Massova, I.,
Seibel, G. L., Singh, U. C., Weiner, P. K., Kollman, P. A.: AMBER 7. University of
California, San Francisco (2002)
9. Pearlman, D. A., Case, D. A., Caldwell, J. W., Ross, W. S., Cheatham III, T. E., DeBolt,
S., Ferguson, D., Seibel, G., Kollman, P. A..: AMBER, a package of computer programs
for applying molecular mechanics, normal mode analysis, molecular dynamics and free
energy calculations to simulate the structural and energetic properties of molecules. Comp.
Phys. Commun, Vol. 91. (1995) 1-41
10. Goodsell, D. S., Olson, A. J.: Automated docking of substrates to proteins by simulated
annealing. Proteins: Structure, Function, and Genetics, Vol. 8. (1990) 195-202
11. Morris, G. M., Goodsell, D. S., Huey, R., Olson, A. J.: Distributed automated docking of
flexible ligands to proteins: parallel applications of AutoDock 2.4. J Computer-Aided
Molecular Design, Vol. 10. (1996) 293-304
12. Morris, G. M., Goodsell, D. S., Halliday, R. S., Huey, R., Hart, W. E., Belew, R. K.,
Olson, A. J., J. Comp. Chem., 19, 1639 (1998)
13. Berendsen, H. J. C., Postma, J. P. M., Gunsteren, W. F. van, DiNola, A., Haak, J. R.,
J. Chem. Phys., 81, 3684 (1984)