Sitagliptin

Synthesis, In Vitro Evaluation, and Computational Simulations Studies of 1,2,3Triazole Analogues as DPP-4 Inhibitors

Duy-Viet Vo1, Kwon Ho Hong2, Jongkook Lee1 and Haeil Park1*

Abstract

Novel 1,2,3-triazole analogues (S7~S10) were synthesized and evaluated for their inhibitory activity against hDPP-4. All the 1,2,3-triazole analogues exhibited moderate in vitro hDPP-4 inhibitory activities (265~780 nM). These results are somewhat less potent compared to those of known 1,2,3-triazole analogues (S1~S6, 14~254 nM). S2 and S3 manifested excellent potency against hDPP-4 with IC50s of 28 and 14 nM, respectively.
The role of the 1,2,3-triazole moiety in binding the molecule to the target was investigated using combined 10 1,2,3-triazole analogues (S1~S10). Molecular dynamics (MD) simulations following the aforementioned docking phase were performed to elucidate potential binding modes of sitagliptin’s 1,2,3-triazole analogues in hDPP-4, with the use of a cocrystal structure of hDPP-4 with sitagliptin (PDB ID: 1X70). Docking and MD simulations of the complexes of hDPP-4 with sitagliptin, S2 and S3 suggest that Glu205, Glu206, Tyr662, and Tyr666 would be the key amino acid residues for the binding of the molecules with the receptor. Especially, S2 and S3 showed additional strong - interaction between Phe357 and 1,2,3-triazole. Same strong - interaction is also observed between Phe357 and the 1,2,4-triazole ring of sitagliptin. Furthermore, additional interactions with Tyr547, Cys551, and especially Arg358 would enhance the binding affinity of the compounds for the pocket of the enzyme. . In overall, in vitro hDPP-4 inhibitory activities of synthetic 1,2,3-triazole analogues were well matched with results of computational simulations studies.

1. Introduction

Dipeptidyl peptidase-4 (DPP-4) inhibitors have emerged as a novel class of antidiabetic drugs which prevent the proteolytic degradation of incretin hormones and stimulate the glucose-dependent insulin secretion associated with a reduction in body weight and the risk of hyperglycemia.1 Inhibition of DPP-4 results in an increase in the half-life of glucagon-like peptide-1 (GLP-1) and in a sustained physiological action of this hormone, leading to an improvement in the glycemic control.2 Sitagliptin is the first approved DPP-4 inhibitor in the United States.
Intensive efforts have been made to investigate type 2 antidiabetic drugs with extended half-lives and better pharmacological profiles.3,4 Previous structure-activity relationship (SAR) studies of sitagliptin revealed that the modification of the 1,2,4-triazolopiperazine substructure is crucial to its bioactivity. Based on these findings, diverse DPP-4 inhibitors comprising different heterocycle scaffolds were explored and proved to improve bioactivities (Figure 1).5-7
The successful precedent SAR results led us to explore novel heterocycle scaffolds as 1,2,4-triazolopipeprazine surrogates of sitagliptin. As part of this research we designed novel sitagliptin analogues (S1~S6) with 1,2,3triazole heterocycle scaffold since the electron-withdrawing effect of nitrogen in 1,2,3-triazole may increase the number and strength of interactions between the ligand and hDPP-4 receptor. Our previous SAR study showed that the bulky alkyl group at C4 position of the 1,2,3-triazole ring exhibited better inhibitory activity than the small alkyl group (tert-Bu > Me > H). Our previous result also showed that compounds with different orientation of nitrogen atom in the pyridine (Py) ring at C4 position exhibited different bioactivities (4-Py > 3-Py > 2-Py) as shown in Figure 1.8 In summary, our previous SAR study suggested two important information as followings: the steric hindrance (size) of C4 substituent as well as orientation of nitrogen atom of pyridine ring at C4 position seemed to play critical role for bioactivity.
In continuation of our SAR study on DPP-4 inhibitors, novel 1,2,3-triazole analogues (S7~S10) were designed, synthesized and evaluated for their bioactivity. Structures of 1,2,3-analogues S7 and S8 were gained by switching the N-alkyl backbone and the C-4 substituent of analogues S2 and S3. Regioisomeric 1,2,3-triazole analogues S7 and S8 were expected to have different interactions with hDPP-4 receptor than those of analogues S2 and S3. 1,2,3-Triazole analogues with 5- and 6-quinoline (Qu) rings instead of 3- and 4-pyridine rings at N-1 (S9/S10) were designed to investigate the influence of steric hindrance and also orientation of nitrogen atom at N-1 substituents (quinoline ring vs. pyridine ring) to bioactivity. In addition, we performed computational simulations of 10 known and novel 1,2,3-triazole analogues (S1~S10) with the crystal structure of human DPP-4 in order to obtain insights into binding mode, binding strength and bioactivity. Herein, we report the synthesis and in vitro biological evaluation (S7~S10), molecular docking and molecular dynamics (MD) simulations of 1,2,3-triazole analogues (S1~S10).

2. Results and discussion

2.1. Chemistry

1,2,3-Triazole analogues (S1~S10) for SAR study were prepared as follows: The analogues (S1~S6) were synthesized from the commercially available (R)-3- (benzyloxy)carbonyl)amino-4-(2,4,5trifluorophenyl)butanoic acid in 5 steps as described in our previous report.8 The regioisomeric analogues (S7~S10) were synthesized from (R)-3- (benzyloxy)carbonyl)amino-4-(2,4,5-trifluorophenyl)butanoic acid (I) in 5 steps. Esterification of the starting material with cerium ammonium nitrate (CAN),9 followed by DIBAL reduction provided the aldehyde intermediate (III).10 The aldehyde intermediate was converted into the corresponding alkyne intermediate (IV) with Bestmann-Ohira reagent.10 Reaction of the alkyne key intermediate with corresponding arylazides in “click reaction” conditions,11 followed by the Cbz deprotection12 gave the corresponding 1,2,3-triazole analogues with propan-2-amine backbone (S7~S10) as shown in Scheme 1.

2.2. DPP-4 inhibition assay

We studied the structural features of the butan-2-amine backbone in compounds S1~S6. The compounds bearing the N-pyridyl group exhibited better potency than compounds with alkyl groups. Also our data showed that there was a relationship between the DPP-4 inhibitory activities and the orientation of the nitrogen atom in the N-pyridyl group, which was likely to play an important role in the interaction with the binding site (IC50 S1~S3, R1: 4-pyridyl > 3-pyridyl > 2-pyridyl). Moreover, analogues (S2 and S3) containing the 3-pyridyl or the 4-pyridyl group showed excellent DPP-4 inhibitory activities (IC50 = 28 and 14 nM, respectively). Among them, compound S3 displayed a strong IC50 value comparable to that of sitagliptin (IC50 = 7.8 nM). Our results also disclosed that the size of the R1 alkyl group affected on DPP-4 inhibitory activities (tert-Bu > Me > H).
As shown in Table 1, propan-2-amine analogues (S7 and S8) possessing N-pyridyl were identified as weak DPP-4 inhibitors (IC50 > 300 nM) compared with butan-2-amine backbone analogues (S1~S3; IC50 < 150 nM). Replacement of N-pyridyl with 5-quinolinyl improved the DPP-4 inhibitory activities slightly (S9; IC50 = 265 nM), but introduction of the 6-quinolinyl group significantly decreased the biological activities (S10; IC50 = 780 nM). To sum up for conducting the further SAR and computational simulations of the sitagliptin analogues, DPP-4 inhibitory activities of propan-2-amine backbone analogues (S7~S10) shown in Table 1 exhibited much reduced bioactivities compared to those of butan-2-amine backbone analogues (S1~S6) in the aforementioned literature.8 2.3. Molecular Docking Studies 2.3.1. Choosing the DPP-4 receptor The crystal structure of the DPP-4 receptor (PDB ID: 1X70, http://www.rcsb.org) used in this study consists of two symmetrical chains of the same protein, chains A and B. They were separately prepared for our computational study. The chain A was selected as representative for two reasons: (1) the obtained re-docked poses of the native ligand (HET code: 715) into the receptor deviated from the original X-ray pose less than those received when the chain B was used as template, and (2) the corresponding docking scores were better, as shown in Table 2. 2.3.2. Molecular docking of 10 sitagliptin analogues into the chain A The docking scores of 10 sitagliptin analogues were distributed from -34.5657 to -22.7799 kJ/mol. Three structures S4~S6 (of the butan-2-amine backbone, each containing an alkyl group) gave higher scores (indicating worse docking solutions) than analogues bearing the same backbone but containing other Npyridyl substituents (S1~S3). This is supported by bioactivity results, revealing that S4~S6 were indeed less potent than S1~S3 (Figure 3). The butan-2-amine backbone analogues (S1~S6) have lower scores than the propan-2-amine analogues (S7~S10), which also agrees with experimental biological data (Figure 3). Analogues (S1~S10) were examined for their in vitro inhibitory activity against the recombinant human DPP-IV using sitagliptin (SGT) as the reference sample. In the 3D docking model, the propan-2-amine backbone was not well-suited to the DPP-4 ligand-binding pocket as the docking poses of all the compounds bearing this backbone (S7~S10) did not coincide with that of the reference sitagliptin structure of the original X-ray pose (Figure 4). S7~S10 also did not form as much strong binding affinity as the Hbonding interactions between ligand and ligand-binding pocket compare with the butan-2-amine backbone analogues, as shown in Figure 4. This explains why the analogues containing the butan-2-amine backbone (S1~S6) were deemed more active than those with the propan-2-amine structure (S7~10). The analogue S3 had the lowest docking score and its docking pose coincided with that of sitagliptin, as shown in Figure 5. Besides, compound S3 formed an H-bonding interaction between the pyridine moiety and the residue Arg358 of DPP-4 and three H-bonding interactions (Glu205, Glu206, and Tyr662) as well as one π-π interaction (Phe357), but S8 did not form any corresponding H-bonding interaction except three H-bonding interactions (Glu205, Glu206, and Tyr662) with the 2-amine group. It was clear that the Hbonding with Arg358 was necessary for a possible bioactivity. The aforementioned observations were backed up by the fact that the analogue S3 (IC50 = 14 nM) exhibited a comparable inhibitory activity compared to that of sitagliptin (IC50 = 7.8 nM).8 For butan-2-amine backbone compounds, the alkyl 1,2,3-triazole substituted analogues exerted a lower inhibitory activity than N-(1-yl-1H-1,2,3-triazole-4-yl)pyridine substituted analogues, because the Npyridine group gave lower docking scores and the docking poses deviated from that of sitagliptin less than those obtained from the alkyl group, as shown in the 3D protein-ligand interactions (Figure 4 and Figure 6). The SAR data for compounds S1~S3 show that the orientation of the H-bonding site (nitrogen atom) that coincided with that of sitagliptin also plays an important role in the inhibitory activity (IC50 of S1~S3 with nitrogen position in R1: para>meta>ortho). For the alkyl group cases, S4~S6, when the alkyl group is large and branched (from -H to -CH3 and -tert-butyl), the corresponding bioactivity increased, as the van der Waals interactions increased. However, the observed improvement in bioactivity was small because van der Waals interactions were normally weak.
The reduced bioactivity of the analogues S7~10 might be attributed to the unfavorable conformation of the propan-2-amine moiety. The replacement of the N-pyridine group with a sterically hindered quinoline ring decreased the inhibitory activity (S7, S8 > S9 » S10). The difference between the inhibitory activities of S9 and S10 implies that the steric factor and the orientation of the H-bonding site (5-quinolinyl vs. 6quinolinyl) might be important for the resulting bioactivity of the compounds.
S10 gave one of the highest docking scores and had the lowest bioactivity among analogues. As shown in the 3D model in Figure 6, S10 exposes most atoms (6-quinolinyl and 1,2,3-triazole ring) to the outside of the pocket, its structure did not totally coincide with sitagliptin, and also the nitrogen atom of 6-quinolinyl could not form an H bond with Arg 358. Therefore, the bioactivity of S10 was significantly reduced compared to those of S7, S8 and S9.

2.4. Molecular dynamics simulations studies

Novel sitagliptin analogues with the 1,2,3-triazole heterocyclic scaffold were designed to investigate the role of the triazole moiety in binding the molecule to the target and to search for a surrogate of the 1,2,4triazolopiperazine substructure of sitagliptin that might enhance the π-π interaction with Phe357 (Figure 7). The designed sitagliptin 1,2,3-triazole analogues were synthesized as shown in Scheme 1, and had their inhibitory activity against the recombinant human DPP-4 evaluated by biological assays. We also performed molecular dynamics (MD) simulations following the aforementioned docking phase to obtain insights into potential binding modes of sitagliptin’s 1,2,3-triazole analogues in hDPP-4, with the use of a cocrystal structure of hDPP-4 with sitagliptin (PDB ID: 1X70, Figure 7).
MD simulations of the complexes of hDPP-4 with sitagliptin, S1, S2, and S3 showed H-bonding interactions with Glu205, Glu206, Tyr662, and Tyr666 as the common key binding interactions between these compounds and the binding pocket. In addition to these interactions, sitagliptin consistently interacted with Phe357, Tyr547, and Cys551 during the MD simulations. This agrees with its binding mode in the hDPP-4 crystal structure (PDB ID: 1X70), except for a water-mediated H-bonding interaction with Cys551: the interaction between Cys551 and sitagliptin is not observed in the X-ray pose (Figures 8 and 9).
The interaction frequency of S1 with Tyr547 significantly increased after the system was equilibrated while its interaction with His126 was observed in the early time points of the MD simulation but disappeared after 2 ns. S2 formed π-π interactions with Phe357 consistently during the MD simulations, reinforcing binding interactions with the binding pocket. S3 formed consistent additional interactions with Phe357 and Arg358: the H-bond between the 4-pyridyl group of S3 and Arg358 and its π-π interactions with Phe357 further reinforced its binding affinity with the pocket. The results from the computational simulations are consistent with those from the hDPP-4 enzyme assay (Figures 8 and 9).

3. Conclusion

All synthesized 1,2,3-triazole analogues (S1~S10) exerted submicromolar inhibitory activities against hDPP-4. Especially, S2 and S3 manifested excellent potency against hDPP-4 with IC50s of 28 and 14 nM, respectively. For the alkyl group cases, S4~S6, when the alkyl group is large and branched (from -H to -CH3 and -tertBu), the corresponding bioactivity increased, as the van der Waals interactions increased. However, the observed improvement in bioactivity was small because van der Waals interactions were normally weak. The reduced bioactivity of the analogues S7~10 might be attributed to the unfavorable conformation of the propan-2-amine moiety. The replacement of the N-pyridine group with a sterically hindered Nquinoline ring decreased the inhibitory activity (S7, S8 > S9 » S10). The difference between the inhibitory activities of S9 and S10 implies that the steric factor and the orientation of the H-bonding site (N-5quinolinyl vs. N-6-quinolinyl) might be important for the resulting bioactivity of the compounds.
Docking and MD simulations of the complexes of hDPP-4 with sitagliptin, S2 and S3 suggest that Glu205, Glu206, Tyr662, and Tyr666 would be the key amino acid residues for the binding of the molecules with the receptor. Especially, S2 and S3 showed additional strong - interaction between Phe357 and the 1,2,3triazole. Same strong - interaction is also observed between Phe357 and the 1,2,4-triazole ring of sitagliptin. Furthermore, additional interactions with Tyr547, Cys551, and especially Arg358 would enhance the binding affinity of the compounds for the pocket of the enzyme. In conclusion, in vitro hDPP-4 inhibitory activities of 1,2,3-triazole analogues were well matched with results of computational simulations studies.

4. Experimental section

4.1. Molecular Docking studies

4.1.1. Ligand Preparation

Ten sitagliptin analogues were prepared as ligands for docking simulations. SMILES chains of the ligands from ChemBioDrawUltra 12.0 were converted to the corresponding 3D structures using ChemBio3DUltra 12.0 and saved in sd file format. The ligands were next subject to a preparation process using SybylX 2.0.13,14 Their energy was first minimized using the Conj Grad method and the standard Tripos force field. The maximum number of iterations to perform was 10,000. Energy minimization was automatically stopped once the energy difference between iterations was smaller than 0.0001 kcal.mol-1.. The Gasteiger-Huckel charges were employed to calculate atomic formal charges. The output structures would then enter a short molecular dynamics (MD) simulation phase. Each was heated at high temperature (700 K) during 1000 ps, and was then cooled down to 200 K in another 1000 ps. The MD process was automatically repeated 10 times to give various necessary configurations of the ligands, after which energy minimization was carried out once again to find the structure with the optimized energy. Finally, the ligands after preparation were converted to sd file format using Sybyl-X 2.0.

4.1.2. DPP-4 Model Preparation

In this study, DPP-4 was imported from RCSB PDB (PDB ID: 1X70, http://www.rcsb.org) with the sitagliptin inhibitor. 1X70 contains a dimer of two chains, and only the monomeric unit was used in the docking studies. The DPP-4 model in complex with ligand was protonated; tethered, and minimized by the LigX tool in MOE,15 and the ligand was subsequently removed.

4.1.3. Molecular Docking

Ten selected DPP-4 inhibitors were docked into the ligand-binding pocket of DPP-4 (PDB ID: 1X70) using the FlexX package in LeadIT 2.1.2 to provide insights into molecular recognition via protein-ligand interactions. In the present process, the triangle matching algorithm was chosen for the place base fragment, while the maximum numbers of solutions per iteration and per fragmentation were defined as 1000 and 200, respectively. After docking, the best docking scores and protein-ligand interactions of the best docking scores poses were employed to explain the bioactivities of 10 synthesized analogues. Compounds with good docking scores were more likely to be the hits for DPP-4 inhibitory activity.16,17

4.2. Molecular Dynamics simulations studies

Docking poses of the compounds in hDPP-4 were used to prepare the systems for MD simulations using Desmond: water molecules and salts (0.15 M MgCl2) were added (orthorhombic box with the volume of 10 Å x 10 Å x 10 Å). The MD simulations of the complexes of hDPP-4 with inhibitors in 0.15 MgCl2 solution water were performed for 5 ns with 2 fs time steps. After that, 18 ns MD simulations of the complexes were performed using Desmond 3.0 employing 128 cores in a high performance computing system. Default option in Desmond/2011 for integration, ensemble, interaction, restraints, output, and Misc were used except Thermostat and Barostat methods in Ensemble; Berendsen method was used as Thermostat and Barostat methods with relaxation time of 1.0 ps and isotropic coupling style for the Barostat.18-21

4.3. DPP-4 inhibition assay

To measure the activity of DPP4, fluorogenic assay was employed using Gly-Pro-AMC, which is cleaved by the enzyme to release fluorescent aminoethylcoumarin (AMC). Compounds (S7~S10) were incubated with recombinant DPP4 or Caco-2 lysate and 50 μM Gly-Pro-AMC in a buffer containing 25 mM Tris/HCl, pH 8.0 (1 mg/ml bovine serum albumin was added only for recombinant human DPP4). The assay was performed at 25 °C for 1 h in a total reaction volume of 200 μl. Liberated AMC was detected using an excitation wavelength of 360 nm and an emission wavelength of 465 nm.

4.4. Chemistry

All chemicals, solvents and reagents were obtained from commercial suppliers and used without further purification, unless specified. Reactions were monitored by thin-layer chromatography performed on glass packed silica gel plates (60 F-254) with UV light. Flash column chromatography was performed with silica gel (100–200 mesh). 1H-NMR (300 MHz & 400 MHz) and 13C-NMR (100 MHz) spectra were recorded on Bruker DPX 400 spectrometers (MA, USA), fully decoupled and chemical shifts are reported in parts per million (ppm) downfield relative to tetramethylsilane as an internal standard. Peak splitting patterns are abbreviated as s (singlet), br s (broad singlet), d (doublet), t (triplet), q (quartet), dd (doublet of doublet), and m (multiplet). Mass spectra were recorded on API 3200 MS system of AB SCIEX. Melting points were recorded on Fisher-Johns microscopic scale melting point apparatus (NJ, USA). Optical rotation data were collected by Jasco DIP-1000 digital polarimeter (MD, USA).

References

1. Y.J. Chyan, L.M. Chuang, Dipeptidyl peptidase-4 Inhibitors: an evolving treatment for type 2 diabetes from the incretin concept. Recent Pat. Endocr. Metab. Immune Drug Discov. 1, 15 (2007).
2. E.J. Verspohl, Novel therapeutics for type 2 diabetes: Incretin hormone mimetics (glucagon-like peptide-1 receptor agonists) and dipeptidyl peptidase-4 inhibitors. Pharmacol. Ther. 124, 113 (2009).
3. N. Li, L.-J. Wang, B. Jiang, X.-Q. Li, C.-L. Guo, S.-J. Guo, D.-Y. Shi, Recent progress of the development of dipeptidyl peptidase-4 inhibitors for the treatment of type 2 diabetes mellitus. Eur. J. Med. Chem. 151, 145 (2018).
4. L. Mendieta, T. Tarrago, E. Giralt, Recent patents of dipeptidyl peptidase IV inhibitors. Expert Opin. Ther. Pat. 2011, 21, 1693.
5. H.J. Kim, W.Y. Kwak, J.P. Min, J.Y. Lee, T.H. Yoon, H.D. Kim, C.Y. Shin, M.K. Kim, S.H. Choi, H.S. Kim, E. K. Yang, Y.H. Cheong, Y.N. Chae, K.J. Park, J.M. Jang, S.J. Choi, M.H. Son, S.H. Kim, M. Yoo, B.J. Lee, Discovery of DA-1229: A potent, long acting dipeptidyl peptidase-4 inhibitor for the treatment of type 2 diabetes. Bioorg. Med. Chem. Lett. 21, 3809 (2011).
6. T. Biftu, G. Scapin, S. Singh, D. Feng, J.W. Becker, G. Eiermann, H. He, K. Lyons, S. Patel, A. Petrov, R. Sinha-Roy, B. Zhang, J. Wu, X. Zhang, G. Doss, N.A. Thornberry, E. Weber, Rational design of a novel, potent, and orally bioavailable cyclohexylamine DPP-4 inhibitor by application of molecular modeling and X-ray crystallography of sitagliptin. Bioorg. Med. Chem. Lett. 17, 3384 (2007).
7. T. Biftu, X. Qjan, P. Chen, D. Feng, G. Scapin, Y.-D. Gao, J. Cox, R.S. Roy, G. Eiermann, H. He, K. Lyons, G. Salituro, S. Patel, A. Petrov, F. Xu, S.S. Xu, B. Zhang, C. Caldwell, J.K. Wu, A.E. Weber, Novel tetrahydropyran analogs as dipeptidyl peptidase IV inhibitors: Profile of clinical candidate (2R,3S,5R)-2-(2,5difluorophenyl)-5-[2-(methylsulfonyl)-2,6-dihydropyrrolo[3,4-c]pyrazol-5(4H)-yl]tetrahydro-2H-pyran-3amine, Bioorg. Med. Chem. Lett. 23, 5361 (2013).
8. S. Gundetti, J. Lee, H. Park, Novel 1,2,3-Triazole Analogs of Sitagliptin as DPP4 Inhibitors. Bull. Kor. Chem. Soc. 37, 1156 (2016).
9. A. Kuttan, S. Nowshudin, M.N.A. Rao, Ceric ammonium nitrate (CAN) mediated esterification of N-Boc amino acids allows either retention or removal of the N-Boc group. Tetrahedron Lett. 45, 2663 (2004).
10. H.D. Dickson, S.C. Smith, K.W. Hinkle, A convenient scalable one-pot conversion of esters and Weinreb aamides to terminal alkynes. Tetrahedron Lett. 45, 5597 (2004).
11. J. Bonnamour, J. Legros, B. Crousse, D. Bonnet-Delpon, Synthesis of new trifluoromethyl peptidomimetics with a triazole moiety. Tetrahedron Lett. 48, 8360 (2007).
12. M. Drag, J. Oleksyszyn, Synthesis of α1-(Cbz-aminoalkyl)-α2-(hydroxyalkyl)phosphinic esters. Tetrahedron Letters 46, 3359 (2005).
13. ChemBioDraw Ultra, PerkinElmer Inc. Available online: https://www.perkinelmer.com/category/chemdraw (accessed on 18 March 2016).
14. SYBYL-X, Tripos International. Available online: https://www.certara.com (accessed on 18 March 2016).
15. Chemical Computing Group Inc. Molecular Operating Environment (MOE). Available online: https://www.chemcomp.com (accessed on 18 March 2016).
16. BioSolveIT–GmbH. FlexX (Protein–Ligand Docker) user & technical reference as part of LeadIT 2.0. Germany, 2012; p. 15. Available online: https://www.biosolveit.de/FlexX (accessed on 18 March 2016).
17. M. Rarey, B. Kramer, T. Lengauer, G. Klebe, A fast flexible docking method using an incremental construction algorithm. J. Mol. Biol. 261, 470 (1996).
18. Protein Preparation Wizard; Epik, Schrödinger, LLC, New York, NY, 2018.
19. LigPrep, Schrödinger, LLC, New York, NY, 2018.
20. Glide, version 7.8, Schrödinger LLC, New York, NY, 2018.
21. Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY, 2018. Maestro-Desmond Interoperability Tools, Schrödinger LLC, New York, NY, 2018.