The preparation of the structure of COX-2 followed the protocol of Furse et al [Kristina E. Furse, Derek A. Pratt, Ned A. Porter, and Terry P. Lybrand, Biochemistry. 2006 March 14; 45(10): 3189-3205]. The crystal structure for the mouse COX-2/ AA complex (pdb entry 1CVU), was used to generate the initial model. Since in this crystal structure, AA is bound in a catalytically unproductive, inverted orientation, a previous theoretical model for COX-2 with a properly oriented AA molecule was used (pdb entry 1DCX). Unfortunately, in the latter case, the docked conformation of AA in this complex appears to be inconsistent with the proposed reaction mechanism.
The following part was taken from Furse et al. It represents the protocol for COX docking and is already cited. If you think, I can rephrase it or we can simply skip it.
The molecular dynamics calculations were carried out with the Yasara program1 running on double core Intel processors under Windows XP program. The simulation box was of 107.11 Å, 75.44 Å, 85.73 Å respectively for the a, b, and c axis under periodic boundary conditions.
The simulations were carried out under the NPT ensemble at 298K and 1 atm by coupling the system with a Berendsen thermostat2 and by controlling the pressure in the in the manometer pressure control mode.2
The AMBER03 3 force field were used for R and S isomers, and new force fields parameters were generated using the Autosmiles method.4 Briefly, the geometry of monomers were optimized by semi-empirical AM1 method using the COSMO solvation model.5 Partial atomic charges were calculated using the same level of theory by the Mulliken point charge approach,6 and were then improved by applying the 'AM1 Bond Charge Correction'.
Electrostatic interactions were calculated with a cutoff of 7.86 A, and the long-range electrostatic interactions were handled by the particle mesh Ewald (PME) 7 algorithm using a sixth-order B-spline interpolation and a grid spacing of 1 Å. After removal of conformational stress by a short steepest descent minimization, the procedure continued by simulated annealing (time step 1 fs, atom velocities scaled down by 0.9 every 10th step) until convergence was reached, i.e. the energy improved by less than 0.0002 kcal/mol per atom during 200 steps.
The binding energy is obtained by calculating the energy of the bound state and subtracting the energy of receptor and ligand at infinite distance. The water molecules in the first solvent shell have to reorganize and lose conformational freedom, which decreases their entropy. Therefore to calculate the free energy we esteemed the entropy assigning 0.155 kcal/mol for Å2.
Results and discussion
The binding of the AArOH is very similar to the one observed for AA. The closest interactions are with ARG , TYR, TYR . The hydroxyl group is well hosted in a relatively wide pocket with only 3 residues at circa 3 A of distance: VAL B 2349, SER B 2353, LEU B 2359. The position of these aminoacids accounts for the invariance of binding energies of the two forms AAsOH and AArOH.
The values of the binding energy for the whole series are listed in table XXX
Fatty acid
kcal/mol
activity
AA
59.44
99.0
OAsOH
79.04
94.9
DHA
130.40
87.4
gLNA
119.99
79.2
OA
73.15
75.3
EPA
69.63
72.8
AAsOH
57.14
60.7
aLNArOH
60.02
57.5
aLNA
111.68
56.7
LA
61.08
36.6
gLNAsOH
35.22
32.8
EPAsOH
41.70
32.1
DHArOH
82.98
31.9
LAsOH
31.39
27.9
The binding energies do not correlate satisfactorily with the in vivo activities. We have performed an extended computational and Quantitative-Structure Activity Relationship (QSAR) investigation on these molecules to define the molecular features required for exhibiting high activity. For all the molecules listed in Table XXX, we have calculated more than 500 molecular descriptors, covering topological, structural, and electronic properties. None of the explored parameters showed a correlation when considered singularly with the fatty acid series. Therefore we performed a QSAR investigation via genetic functions (GF). One of the main advantages of GF analysis is the production of multiple models. We generated a total of 100 QSAR equations that consist of one to four descriptors among the QSAR random models. Applying the "test set method" strategy we generated models using only 11 out of 14 molecules. All the models obtained as outcome of the different choice of the training set, have been tested against the three molecules left out.
Compound DHArOH showed an activity much smaller than predicted and is responsible of r2 =0.42. Omitting fatty acid DHArOH resulted in a correlation of r2=0.91. This is a strong indication that DHArOH acts through a mechanism different from the one used by the other fatty acids.
Among the models, the one with the strongest predictive power of the in vivo activity (IVA) was the following:
IVA = 8.88DipY2 + 0.83AlogP2-277
Structures
Binding kcal/mol
activity
HOMO
LUMO
Molecular surface area
AlogP
AA
59.44
99
-9.73
-7.66
421.89
10.51
OAsOH
79.04
94.9
-10.06
-7.56
404.84
9.15
DHA
130.4
87.4
-9.42
-7.58
427.87
11.30
gLNA
119.99
79.2
-9.71
-7.48
391.78
9.72
OA
73.15
75.3
-9.87
-7.70
393.86
9.72
EPA
69.63
72.8
-9.79
-7.61
405.36
10.51
AAsOH
57.14
60.7
-9.64
-7.49
435.32
9.94
aLNArOH
60.02
57.5
-9.63
0.85
404.15
7.38
aLNA
111.68
56.7
-9.58
0.96
394.89
7.95
LA
61.08
36.6
-9.76
1.21
412.38
7.95
gLNAsOH
35.22
32.8
-9.73
-7.26
398.50
9.15
EPAsOH
41.7
32.1
-9.80
-7.51
411.79
9.94
DHArOH
89
31.9
-9.73
-7.60
455.28
10.73
LAsOH
31.39
27.9
-9.76
0.97
420.73
7.38
Molecules with low LUMO energy values are more able to accept electrons than molecules with high LUMO energy values. The LUMO energy value is increased with the presence of electron donating groups. By increasing in LUMO energy softness is decreased and thus the molecules can inhibit enzyme strongly.
(1) Krieger, E.; Darden, T.; Nabuurs, S.; Finkelstein, A.; Vriend, G. Proteins 2004, 57, 678-683.
(2) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Di Nola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684-3689.
(3) Duan, Y.; Wu, C.; Chowdhury, S.; Lee, M.; Xiong, G.; Zhang, W.; Yang, R.; Cieplak, P.; Luo, R.; Lee, T. J.Comput.Chem. 2003, 24, 1999-2012.
(4) Jakalian, A.; Jack, D.; Bayly, C. J. Comput. Chem. 2002, 23, 1623-1641.
(5) Klamt, A. J. Phys. Chem. 1995, 99, 2224-2235.
(6) Stewart, J. J. P. J. Comp. Aided Mol. Des. 2000, 4, 1-103.
(7) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. B 1995, 103, 8577-8593.
