Mottadi Rupa and Thirumurthy Madhavan†
Abstract
Non-small cell lung cancer (NSCLC) is the leading cause of cancer-related mortality worldwide. Phosphoinositide 3- kinases (PI3Ks) play important role in Non-Small Cell Lung Cancer. PI3Ks constitute a lipid kinase family which modulates the function of numerous substrates involved in the regulation of cell survival, cell cycle progression and cellular growth. Herein, we describe the ligand based pharmacophore combined with molecular docking studies methods to identify new potent PI3K inhibitors. Several pharmacophore models were generated and validated by Guner-Henry scoring Method.
The best models were utilized as 3D pharmacophore query to screen against ZINC database (Chemical and Natural) and the retrieved hits were further validated by fitness score, Lipinski’s rule of five. Finally four compounds were found to have good potential and they may act as novel lead compounds for PI3K inhibitor designing.
Keywords: PI3K, Non-small Cell Lung Cancer, 3D Pharmacophore.
1. Introduction
Phosphoinositide 3-kinases (PI3Ks) catalyze phos- phorylation of the 3-OH group of phosphatidylinositol (PI) to generate phosphatidylinositol 3-phosphate (PIP), phosphatidylinositol 3,4-bisphosphate (PIP2), and phos- phatidylinositol 3, 4, 5- trisphosphate (PIP3), which act as second messengers[1]. In the appropriate cellular con- text, these three lipids can regulate a remarkably diverse array of physiological processes, including cell growth, survival, differentiation, and chemotaxis[2]. There are three classes of PI3Ks—classes I, II and III—based on their sequences and substrate specificities. Class I PI3Ks are further divided into class IA enzymes (PI3Kα, β, and δ), which have a p85 regulatory subunit and three different catalytic subunits (p110α, β, and δ), and a class IB enzyme (PI3Kγ) with a p101 regulatory subunit and p110γ catalytic subunit, which are activated by tyrosine kinases or G protein-coupled receptors to gen- erate PIP3.
Recently, it has been confirmed that angiogenesis, which is essential for tumor growth and metastasis, selectively requires the p110α isoform of PI3K to con- trol endothelial cell migration[3]. All these facts suggest that PI3K p110α is a potential and attractive target for cancer therapy and hence spark great interest in the dis- covery and development of inhibitors. The aim of this study was to identify new potent inhibitors and to inves- tigate important chemical features responsible for inhi- bition of PI3K receptor. Studies have suggested that pharmacophore model is widely used to explore the important chemical features and to find new class of chemical entities[4-6]. Hence in this study, ligand based 3D pharmacophore models were generated and the best hypotheses were selected based on potency validation and Guner-Henry scoring methods. The obtained hits were subsequently filtered by fitness score, Lipinski's rule of five and ADME (Absorption, Distribution, Metabolism and Excretion) properties. Virtual screening was found to be successful method especially when combined with molecular docking studies to eliminate the false positive results. Therefore, molecular docking was performed to identify the binding modes and inter- action of the hits in the active site of PI3K.
Department of Bioinformatics, School of Bioengineering, SRM Univer- sity, SRM Nagar, Kattankulathur, Chennai 603203, India
1 7.744
2 6.552
3 8.154
4 8.301
5 6.677
6 7.698
7 7.154
9 7.055
10 7.823
11 7.853
12 6.259
13 7.356
14 7.408
15 6.795
16 7.585
18 7.000
19 7.657
20 7.853
21 7.619
22 7.366
23 7.408
24 7.148
26 6.769
27 6.283
28 5.920
29 6.744
30 7.886
31 7.154
32 8.960
33 8.045
35 7.958
36 8.945
37 8.000
38 8.154
39 8.394
40 8.301
41 8.221
42 8.221
43 8.301
44 8.301
2.2. Computational Details
The chemical structure of PI3K inhibitors was sketched using sketch molecule function whose partial atomic charge was calculated using GasteigerHuckel method and its energy minimization was performed using Tripos force field in SYBYL. The pharmacophore model generation and screening was performed using Maestro software. Molecular docking was performed using SYBYL-X2.1 package.
2.3. Pharmacophore Model Generation
In the absence of macromolecule structure, ligand based pharmacophore modeling is the key computa- tional strategy for facilitating the drug discovery pro- cess[8]. Hence, ligand based pharmacophore modeling was carried out for the known 46 PI3K inhibitors using the PHASE module which is the highly flexible system for pharmacophore model generation in Maestro[9]. Ini- tially structures were cleaned and various conformations were generated by using Ligprep module. PHASE pro- vides the set of six pharmacophore features namely, hydrogen bond acceptor (A), hydrogen bond donor (D),
active, inactive, or intermediate based on the biological activity values. Out of 46 compounds, we have set 8 molecules as active (pIC50 > 8.300), 19 as inactive molecules (pIC50 < 7.500), and rest of them as inter- mediate (neither active nor inactive). Searching for common hypothesis was performed using both five and six pharmacophoric sites. The survival scores (both
‘‘survival’’ and ‘‘survival- inactive’’) for each hypothe- sis were calculated with default parameters. Ample dif- ferences between these scores for particular hypothesis suggest that the hypothesis can correctly distinguish between active and inactives.
2.4. Pharmacophore Validation
Pharmacophore validation was performed to evaluate the accuracy and reliability of the generated models.
Two types of validation was performed namely, potency validation and Güner-Henry score validation.
2.5. Potency Validation
Potency validation was performed using decoy set to test whether the model is good enough to pick the active
46 8.397
2.6. Güner-Henry Score Validation
Güner-Henry scoring method is applied for quantifi- cation of model selectivity and evaluation of model effectiveness of similarity search. This scoring evokes the actives from a molecule dataset consisting of known active and inactive molecules. This scoring system ranges from 0 to 1, where 0 specifies a null model and 1 specifies an ideal model. The score is expected to be greater than 0.7[11]. The formulas used for calculating GH score are given below:
Güner-Henry score =
where Ha is the number of actives in the hits list (true positives), A is the number of active compounds in the database, Ht is the number of hits retrieved, D is the number of compounds in the database, %A is the per- centage of known active compounds obtained from the database, %Y is the percentage of known actives in the hits list, E is the enrichment of the concentration of actives by the model relative to random screening with- out a pharmacophoric approach. Güner-Henry score is considered as a relevant metric, it take into account both percent yield of actives in a database (%Y) and the per- cent ratio of actives in the hit list (%A).
2.7. Pharmacophore Based Virtual Screening Pharmacophore based virtual screening is the fastest and accurate techniques and can be efficiently used to identify new leads. The best pharmacophore hypotheses were taken as template for retrieving potent molecules with novel chemical structures and desired chemical features from ZINC natural products and chemical products. The obtained hits were further filtered by applying Lipinski's rule of five, fitness score, ADME properties and molecular docking.
pharmaceutically relevant properties for small drug-like molecules. In this work, we considered all the properties predicted by Qikprop such as amine, amidine, acid, amide, rotor, rtvFG, CNS, mol MW, dipole, SASA, FOSA, FISA, PISA, WPSA, volume, donorHB, accptHB, glob, QPpolrz, QPlogPC16, QPlogPoct, QPlogPw, QPlogPo/w, QPlogS, QPlogHERG, QPPCaco, QPlogBB, QPPMDCK, QPlogKp, IP(ev), EA(ev), metab, QPlogKhsa, Human Oral Absorption, SAFluo- rine, SAamideO, PSA, NandO are within the recom- mended range. The reason behind considering all these properties is to make sure that whether these particular chosen molecules from database can stand the Lipinski's rule, as well as pharmacokinetics and pharmacodynam- ics properties.
2.9. Molecular Docking
Molecular docking provides visualization of potential binding orientation of the hits with the important resi- dues of PI3K. Docking was performed using Surflex- dock interfaced with SYBYLX2.0.[13] The Surflex dock uses an empirically derived scoring function that is based on the binding affinities of the protein ligand complexes. The protein was prepared using structure preparation tool and energy minimizationwas performed for 100 steps utilizing Powell method and Tripos force field. The binding site residues were used for generation of protomol. Protomolrepresents the unique and impor- tant factor of the docking algorithm, representing the interaction of the ligand with the binding site of the protein. It implements the Hammerhead's empirical scoring function with the molecular similarity method to create putative poses of ligand fragments. The Sur- flex scoring function which is based on the binding affinities of protein ligand complexes takes into account several terms, including hydrophobic, polar, repulsive, entropic and solvation. The docking scores are expressed in terms of -lg10Kd units to evaluate the Ha 3A Ht( + )
( )
4 Ht A× ×
--- 1 Ht Ha– D A– ---
⎝ – ⎠
⎛ ⎞
%A Ha 100×
--- %YA ; Ha 100×
--- EA ; Ha Ht⁄ A D⁄ ---
= = =
No. Hypothesis Survival
-inactive Post-hoc Site Vector Volume Selectivity
Matches Energy Activity Inactive
1 ADHRR 3.713 1.219 3.713 0.89 0.994 0.825 1.849 8 0 8.301 2.494
2 AADHR 3.712 1.334 3.712 0.89 0.994 0.825 1.63 8 0 8.301 2.378
3 AADHH 3.685 1.421 3.685 0.87 0.999 0.815 1.572 8 0 8.301 2.264
4 ADHHR 3.684 1.418 3.684 0.87 0.999 0.816 1.885 8 0 8.301 2.266
5 AAADH 3.557 1.196 3.557 0.86 0.945 0.757 1.379 8 1.978 8.301 2.361
6 DHHRR 3.552 1.34 3.552 0.82 0.985 0.75 2.145 8 1.409 8.301 2.212
7 AHHRR 3.548 1.331 3.548 0.81 0.986 0.75 2.042 8 1.409 8.301 2.217
8 AAHHR 3.545 1.33 3.545 0.8 0.986 0.756 1.86 8 1.409 8.301 2.216
9 DDHRR 3.474 1.249 3.474 0.78 0.971 0.723 2.097 8 0 8.301 2.225
10 ADDHR 3.467 1.244 3.467 0.77 0.978 0.722 1.867 8 0 8.301 2.223
11 ADDRR 3.451 1.233 3.451 0.72 0.972 0.758 1.606 8 0.001 8.945 2.218
12 AADDR 3.449 1.229 3.449 0.71 0.98 0.761 1.35 8 0.001 8.945 2.22
13 AADDH 3.438 1.42 3.438 0.69 0.987 0.758 1.395 8 0 8.945 2.018
14 ADDHH 3.425 1.413 3.425 0.73 0.983 0.71 1.784 8 0 8.301 2.012
15 DDHHR 3.418 1.407 3.418 0.73 0.978 0.712 2.107 8 0 8.301 2.01
16 AAAHR 3.368 1.142 3.368 0.7 0.966 0.701 1.768 8 4.19 8.96 2.226
17 AAHRR 3.366 1.361 3.366 0.69 0.963 0.717 1.839 8 2.856 8.301 2.005
18 AADRR 3.36 1.204 3.36 0.68 0.96 0.719 1.425 8 0.708 8.397 2.156
19 AAADR 3.352 1.2 3.352 0.66 0.968 0.727 1.181 8 0.708 8.397 2.151
20 AAADR 3.316 1.143 3.316 0.62 0.954 0.741 1.182 8 4.19 8.96 2.173
21 AAADD 2.474 0.485 2.474 0.25 0.898 0.327 0.926 8 2.107 8.301 1.988 A-acceptor, D-donor, H-hydrophobic, P-positives, N-negatives, R-aromatic ring.
The bold values indicate the selected hypothesis.
Survival score: weighted combination of the vector, site, volume, and survival scores, and a term for the number of matches and the minimum value of this score is 1.0, Survival-inactive-survival score: for actives with a multiple of the survival score for inactives subtracted, Post-hoc: This score is the result of rescoring, and is a weighted combination of the vector, site, volume, and selectivity scores, Site: measures how closely the site points are superimposed in an alignment to the pharmacophore of the structures that contribute to this hypothesis, based on the RMS deviation of the site points of a ligand from those of the reference ligand, Vector: measures how well the vectors for acceptors, donors, and aromatic rings are aligned in the structures that contribute to this hypothesis, Volume: Measures how much the volumes of the contributing structures overlap when aligned on the pharmacophore, Selectivity: The selectivity is the negative logarithm of the fraction of molecules in the Index that match the hypothesis. A selectivity of 2 means that 1 in 100 molecules
3.2. Potency Validation
A database comprising of 1047 molecules (1000 decoys and 46 known PI3K inhibitors) were used for potency validation. Potency validation was performed to check which pharmacophore model was able to pick more active inhibitors with less number of decoys.
Among the selected twenty one hypotheses, we found the hypotheses DDHHR has picked 26 actives with 12 decoys which indicates these two pharmacophore mod- els are more reliable in picking the active molecules.
hypothesis DDHHR is showing higher GH score than others and hence, these pharmacophore hypotheses were selected and used for screening. The selected phar- macophore hypothesis DDHHR is represented in Fig.
1(a) and (b).
3.4. Pharmacophore Based Virtual Screening The selected pharmacophore hypothesis DDHHR was employed as 3D query against ZINC Chemical and Natural compounds. Compounds whose chemical group maps with the pharmacophore sites were captured as
Table 3. Pharmacophore validation using Güner-Henry Scoring method
S.No Hypothesis Hits Ha Ht A D %A %Y E GH Score
1 ADHRR 170 27 197 27 1027 100 13.71 5.21 0.29
2 AADHR 199 27 226 27 1027 100 11.94 4.54 0.27
3 AADHH 136 27 163 27 1027 100 16.54 6.31 0.32
4 ADHHR 121 27 148 27 1027 100 18.24 6.94 0.34
5 AAADH 147 26 173 27 1027 96.29 15.08 5.71 0.31
6 DHHRR 84 27 111 27 1027 100 24.32 9.25 0.41
7 AHHRR 95 27 122 27 1027 100 22.13 8.41 0.37
8 AAHHR 141 27 171 27 1027 100 15.78 6.01 0.32
9 DDHRR 22 26 48 27 1027 96.29 54.16 20.6 0.63
10 ADDHR 51 26 77 27 1027 96.29 33.76 12.8 0.47
11 ADDRR 47 26 73 27 1027 96.29 35.61 13.5 0.48
12 AADDR 67 26 93 27 1027 96.29 27.95 10.6 0.42
13 AADDH 79 26 105 27 1027 96.29 24.76 9.41 0.39
14 ADDHH 24 26 50 27 1027 96.29 52 19.8 0.62
15 DDHHR 12 26 38 27 1027 96.29 68.42 26 0.75
16 AAAHR 115 26 141 27 1027 96.29 18.43 7.01 0.34
17 AAHRR 53 27 80 27 1027 96.29 33.75 12.8 0.48
18 AADRR 130 26 156 27 1027 96.29 16.66 6.33 0.32
19 AAADR 124 26 150 27 1027 96.29 17.33 6.58 0.32
20 AAAHH 80 26 106 27 1027 96.29 24.52 9.32 0.39
21 AAADD 150 26 176 27 1027 96.29 14.77 5.61 0.3
The bold values indicate the goodness of the hypothesis based on Guner-Henry score.
Ha is the number of actives in the hits list (true positives), Ht is the number of hits retrieved. A is the number of active compounds in the database, D is the number of compounds in the database, %A is the percentage of known active
hits. We found out 16046 hits matching DDHHR hypothesis. We chose only the hits which attained the fitness score greater than or equal to 1.5 because the fit- ness score describes how well the pharmacophore site
obtained and these hits were subjected to ADME pre- diction. The binding and interaction of compound 36 and the four hits which have the highest activity were illustrated in Fig. 2.
Fig. 1. (a) DDHHR the best pharmacophore hypotheses produced by the PHASE module. Pharmacophore sites are colored in blue, thick blue, red, brown and greencontours represent the H-bond donor (D), positively ionizable (P), negatively ionizable (N), aromatic ring (R) and Hydrophobic (H) groups respectively. (b) DDHHR the best pharmacophore hypotheses produced by the PHASE module.The pink dotted lined indicates the distance between the pharmacophore features (Å).
selecting the compound for in vitro studies. So, we investigated the drug likeliness of 2548 hits by using Qikprop module. The recommended ranges are calcu- lated on the basis of molecular characteristics of a par- ticular compound as well as pharmacokinetic and pharmacodynamics behavior of a molecule from in- vitro and in-vivo studies. The hits which follow Lipin- ski’s rule and other ADME properties were alone selected which in turn gave 39 hits. The hits were checked for structural redundancy and duplicate struc- tures were eliminated which yielded 24 hits whose binding interactions were studied using docking.
PI3K, we performed molecular docking using Surflex dock to find the suitable orientation of the hits within the binding site. We also characterized the binding site of PI3K which comprises of the following residues HIS39, PHE87, PHE88, PHE111, PHE112, ASN114, MET115, PHE116, TYR166, ARG170, TYR184, ASP185, VAL186, LEU187, GLN203, LEU206, LYS210, PHE211, TRP259, GLY260, TYR262, HIS263, PHE265, SER266, LEU267, GLU269, ALA270, TRP283, LEU286, PRO287, PHE288, VAL289, THR290, SER291, LEU292, ALA293 and PHE294.
Fig. 2. Binding poses and interaction with active site residues of highly active molecule(compound 36) and the selected hits from ZINC database.
10 ZINC02124114 1.821 8.655 -2.637 4.003
11 ZINC04165433 1.584 7.470 -0.975 5.303
12 ZINC04620922 1.786 5.578 -1.015 3.259
13 ZINC05918661 1.556 7.398 -0.633 4.074
14 ZINC05918663 1.567 6.757 -0.527 4.542
15 ZINC05918997 1.654 7.874 -0.600 4.843
16 ZINC18068762 1.634 5.215 -0.450 2.064
17 ZINC08430124 1.743 7.421 -1.681 3.206
18 ZINC08430407 1.926 6.212 -0.787 5.741
19 ZINC08437659 1.576 6.701 -0.498 5.377
20 ZINC08442089 1.554 6.924 -0.782 2.804
21 ZINC08680885 1.799 6.138 -0.616 2.050
22 ZINC08739938 1.677 7.788 -0.385 4.542
23 ZINC08740291 1.958 6.822 -1.004 4.256
24 ZINC08744114 1.66 7.860 -0.661 2.831
25 ZINC08973405 1.682 7.814 -0.886 5.559
26 ZINC09130648 1.656 6.655 -0.772 1.221
27 ZINC15676063 1.614 6.775 -0.845 4.193
28 ZINC15957672 1.618 6.310 -1.448 1.378
29 ZINC18046579 1.613 6.343 -0.975 3.259
30 ZINC18102470 1.625 6.701 -0.498 4.542
31 ZINC18166941 1.521 6.822 -0.450 4.843
32 ZINC70673872 1.79 6.757 -1.015 2.053
33 ZINC79204073 1.528 7.814 -0.546 2.106
34 ZINC79204151 1.576 5.578 -0.527 3.259
35 ZINC79204151 1.574 7.470 -0.975 1.378
36 ZINC85592478 1.551 6.614 -0.787 2.064
37 ZINC85592495 1.507 6.757 -1.681 4.542
38 ZINC85592496 1.621 5.578 -0.498 5.303
39 ZINC85625485 1.562 7.421 -1.015 3.259
Surflex score: Total Surflex Dock score expressed as elog(Kd), Crash score: The degree of inappropriate penetration
5 ZINC00983266 1.587 435.924 2 8.7 3.491 -7.16 -1.517
6 ZINC01009450 1.573 488.349 2 7.2 3.917 -6.324 -1.544
7 ZINC17835543 1.614 398.418 5 7 1.855 -4.141 -1.91
8 ZINC02062354 1.879 438.457 1 7.5 3.203 -7.84 -2.6
9 ZINC02082198 1.809 427.843 1 6 3.932 -6.613 -1.61
10 ZINC02124114 1.821 447.935 2 7.2 3.381 -5.059 -0.981
11 ZINC04165433 1.584 487.557 3 8 3.313 -6.906 -1.825
12 ZINC04620922 1.786 347.2 1 4.5 4.234 -5.767 -0.069
13 ZINC05918661 1.556 434.469 3 5 4.057 -6.527 -1.392
14 ZINC05918663 1.567 434.469 3 5 4.375 -7.079 -1.418
15 ZINC05918997 1.654 436.897 3 5 4.067 -6.627 -1.284
16 ZINC18068762 1.634 368.391 5 6.25 1.519 -3.522 -1.704
17 ZINC08430124 1.743 427.843 1 6 3.893 -6.331 -1.564
18 ZINC08430407 1.926 475.536 1 9.5 3.028 -5.972 -2.208
19 ZINC08437659 1.576 489.609 3 5.75 4.278 -7.507 -1.644
20 ZINC08442089 1.554 430.505 3 5 4.066 -6.705 -1.465
21 ZINC08680885 1.799 428.483 2 10 1.897 -6.629 -2.64
22 ZINC08739938 1.677 380.323 3 4 3.898 -6.03 -0.84
23 ZINC08740291 1.958 492.574 2 9.75 3.21 -4.841 -1.423
24 ZINC08744114 1.66 338.362 3 4 2.884 -4.657 -0.872
25 ZINC08973405 1.682 372.81 5 5.5 1.928 -4.27 -1.693
26 ZINC09130648 1.656 417.261 5 5.5 2 -4.374 -1.691
27 ZINC15676063 1.614 434.429 2 7.25 2.666 -4.282 -0.759
28 ZINC15957672 1.618 476.531 3 8.75 1.473 -2.987 -1.543
29 ZINC18046579 1.613 422.363 5 6.25 2.363 -4.838 -1.67
30 ZINC18102470 1.625 471.233 5 5.5 3.191 -6.177 -1.522
31 ZINC18166941 1.521 350.297 3 3.25 3.486 -5.097 -0.666
32 ZINC70673872 1.79 534.611 3.5 10.5 3.122 -5.491 -1.761
33 ZINC79204073 1.528 698.817 6 9.5 6.584 -8.637 -2.797
34 ZINC79204151 1.576 698.817 6 10.5 6.111 -9.039 -3.489
35 ZINC79204151 1.574 698.817 6 10.5 6.111 -9.04 -3.489
36 ZINC85592478 1.551 617.651 4 14.75 2.36 -5.224 -2.623
37 ZINC85592495 1.507 486.518 4 11.75 1.442 -3.74 -2
38 ZINC85592496 1.621 488.533 4 11.75 1.587 -3.707 -1.987
39 ZINC85625485 1.562 563.689 5 7.45 4.552 -4.926 -2.615
mol MW: Molecular weight of the molecule, accptHB: Estimated number of hydrogen bonds that would be accepted by the solute from water molecules in an aqueous solution, DonorHB: estimated number of hydrogen bonds that would be donated by the solute to water molecules in an aqueous solution, QPlogBB: predicted brain/blood partition coefficient, QPlogS: predicted aqueous solubility, log S. S in mol dm_3 is the concentration of the solute in a saturated solution that
binding site residues were used for protomol generation.
In order to validate the docking results known PI3K inhibitors was also docked into binding site of the receptor using Surflex dock program. After running sur- flex dock the scores of the active docked conformers were ranked in the molecular spreadsheet. SurflexDock results mainly contain 3 information, (a) total score (surflex score) which is the total SurflexDock score expressed as log (Kd) to represent binding affinities which include hydrophobic, polar, repulsive, entropic and solvation, (b) Crash value which is the degree of inappropriate penetration by the ligand into the protein and of interpenetration between ligand atoms (self- clash) that are separated by rotatable bonds, (c) Polar value is a contribution of the hydrogen bonding and salt
TYR166, GLN203, LYS210, TYR262, GLU269 and THR290. Focusing on surflex score and interaction of hits similar to known inhibitors, twenty eight molecules were selected. The scores obtained for these twenty four molecules are tabulated in Table 4 and its associated ADME predictions are shown in Table 5.
4. Conclusion
In this study, we described the strategy for identifying new inhibitors for PI3K using combined pharmacoph- ore based virtual screening and docking. The best phar- macophore was generated and validated through Guner- Henry scoring method and it was used as query to per- form virtual screening against ZINC database (Chemi-
10 ZINC04165433 7.96 4 ARG93,PHE119
11 ZINC04620922 5.02 0
12 ZINC05918661 5.77 2 SER681,GLN475
13 ZINC05918663 5.42 2 ARG93,PHE119
14 ZINC05918997 5.39 3 GLY122,LYS672,ILE841
15 ZINC08430124 5.97 3 ARG93,GYS672,ASP843
16 ZINC08430407 5.48 3 LYS672,HIS676
17 ZINC08437659 5.37 3 GLU76,THR482
18 ZINC08442089 5.70 3 GLY122,ILE841,LYS672
19 ZINC08739938 5.00 5 LYS672,MET675,HIS676
20 ZINC08740291 6.25 2 GLU76
21 ZINC08744114 5.82 4 LYS678,HIS676,ASP843
22 ZINC15676063 6.52 4 GLN475,ASP843,ILE841
23 ZINC15957672 5.90 5 GLY122,GLU76,LYS672,ILE841
24 ZINC1816641 4.84 3 GLY122,LYS672,ASP843
References
[1] D. A. Fruman, R. E. Meyers, and L. C. Cantley,
“Phosphoinositide kinases”, Annu. Rev. Biochem., Vol. 67, pp. 481-507, 1998.
[2] R. Katso, K. Okkenhaug, K. Ahmadi, S. White, J.
Timms, and M. D. Waterfield, “Cellular function of phosphoinositide 3-kinases: implications for devel- opment, homeostasis, and cancer”, Annu. Rev. Cell Dev. Biol., Vol. 17, pp. 615-675, 2001.
[3] M. Graupera, J. Guillermet-Guibert, L. C. Foukas, L. K. Phng, R. J. Cain, A. Salpekar, W. Pearce, S.
Meek, J. Millan, P. R. Cutillas, A. J. H. Smith, A.
J. Ridley, C. Ruhrberg, H. Gerhardt, and B. Van- haesebroeck, “Angiogenesis selectively requires the p110alpha isoform of PI3K to control endothelial cell migration. Nature”, Nature, Vol. 453, pp. 662- 666, 2008.
[4] H. Z. Xie, L. L. Li, J. X. Ren, J. Zou, L. Yang, Y.
Q. Wei, and S. Y. Yang, “Pharmacophore modeling study based on known spleen tyrosine kinase inhib- itors together with virtual screening for identifying novel inhibitors”, Bioorg. Med. Chem. Lett., Vol.
19, pp. 1944-1949, 2009.
[5] M. O. Taha, N. Atallah, A. G. Al-Bakri, C. Paradis- Bleau, H. Zalloum, K. S. Younis, and R. C.
Levesque, “Discovery of new MurF inhibitors via pharmacophore modeling and QSAR analysis fol- lowed by in-silico screening”, Bioorg. Med. Chem.
2008.
[7] J. Pothier, M. A. Riederer, O. Peter, X. Leroy, A.
Valdenaire, C. Gnerre, and H. Fretz, “Novel 2-(2- (benzylthio)-1H-benzo[d]imidazol-1-yl)acetic acids:
discovery and hit-to-lead evolution of a selective CRTh2 receptor antagonist chemotype”, Bioorg.
Med. Chem. Lett.., Vol. 22, pp. 4660-4667, 2012.
[8] S. Y. Yang, “Pharmacophore modeling and applica- tions in drug discovery: challenges and recent advances”, Drug Discov. Today, Vol. 15, pp. 444- 450, 2010.
[9] S. L. Dixon, A. M. Smondyrev, E. H. Knoll, S. N.
Rao, D. E. Shaw, and R. A. Friesner, “PHASE: a new engine for pharmacophore perception, 3D QSAR model development, and 3D database screening”, J. Comput. Aid. Mol. Des., Vol. 20, pp.
647-671, 2006.
[10] M. M. Mysinger, M. Carchia, J. J. Irwin, and B. K.
Shoichet, “Directory of useful decoys, enhanced (DUD-E): better ligands and decoys for better benchmarking”, J. Med. Chem., Vol. 55, pp. 6582- 6594, 2012.
[11] J. M. Yang and T. W. Shen, “A pharmacophore- based evolutionary approach for screening selective estrogen receptor modulators”, Proteins, Vol. 59, pp.
205-220, 2005.
[12] QikProp, Version 3.2, Schrodinger LLC, New York, USA, 2009.
[13] SYBYL Software, Version 7.3, Tripos Associates Inc, St. Louis, USA, 2006.
