3.1. Fingerprint-based 2D QSAR modeling
After building various 2D QSAR models, as discussed in Section 2.3, models that can accurately predict the active and inactive molecules in the test set were selected. Of the 147 models generated, only 87 could predict the test set of molecules without any error tabulated in Table 1 and were therefore used to predict the activity of 305168 database molecules. After predicting the activity, the top twenty-five molecules found to be active using most models were enlisted in Table 2. Out of the top twenty-five molecules, only seven molecules, 109586, 115183, 125036, 142354, 197105, 253878, and 256085, were predicted to be active in all 87 models.
3.2. Shape-based virtual screening
The top twenty-five molecules obtained from fingerprint-based 2D QSAR were screened using known Akt activators, as discussed in Section 2.4. The shape sim score for the selected molecules is listed in Table 3. Of the twenty-five molecules, 133584 showed the highest score in the five models obtained from quercetin-3-glucuronide (shape sim-0.3638), puerarin (shape sim-0.471), kaempferol-3-glucuronide (Shape sim-0.339), inositol 1,3,4,5-tetrakisphosphate (shape sim-0.421), and baicalin (shape sim-0.501), in the other two models chlorogenic acid (shape sim-0.285) and SC79 (shape sim-0.349) it had lower scores than the other molecules.
3.3. Pharmacophore modeling and virtual screening
3.3.1. Protein-ligand complex-based pharmacophore
The model generated by this methodology has three acceptor groups, A8, A11, and A15, which are aligned with the three phosphate groups of inositol 1,3,4,5-tetrakisphosphate which is the ligand bound to the PH domain of Akt1 (Fig. 1). The A15 is 5 Å apart from A8, A15 is 10.42 Å apart from A11, and A11 is 8.46 Å apart from A8. Among the top twenty-five selected molecules, only five showed similarities with the generated model, of which 133584 (fitness score-2.214) showed the highest fitness score, which also showed the highest shape similarity score in the shape-based screening approach. Other molecules that emerged as hits in this model were 115183 (fitness score-2.144), 109586 (fitness score-1.997), 12931 (fitness score-1.932), and 132665 (fitness score-1.768). The fitness score of all twenty-five selected molecules has been tabulated in Table 3.
3.3.2. Multiple ligand-based pharmacophore
Ten models were generated for multiple ligand-based pharmacophores (Table S2). Pharmacophore models were ranked according to the phase hypo score, among which AAAD_3, AAAD_5, and AAAD_9 had the highest phase hypo score of 1.17, but AAAD_5 had the highest BEDROC score of 0.85 among the three models; therefore, they were selected for screening. The
model has three acceptor groups (A2, A3, and A5), and one donor group D15, as shown in Fig. 2. The distance between the acceptor groups was less than 4 Å compared to the model obtained from Protein-ligand complex-based pharmacophore. It had a donor group aligned to the hydroxyl group in inositol 1,3,4,5-tetrakisphosphate. After the screening of the top twenty-five selected molecules, only nine molecules showed similarities to that of the model, out of which three molecules had fitness scores of more than 1.5, which are 109586 (fitness score-2.338), 115183 (fitness score-1.9), 126004 (fitness score-1.724). 109586 were positive in 86 models generated by 2D QSAR and had a high score for protein-ligand complex-based pharmacophores
3.4. Molecular docking and binding free energy calculation
After performing various extensive ligand-based approaches such as 2D QSAR, shape-based screening, and pharmacophore-based virtual screening, the top twenty-five molecules were selected for molecular docking. To perform structure-based approaches, the crystal structure of the full-domain Akt1 structure is required, which is currently not available in the active conformation. The only crystal structure of Akt1 in the active conformation is 1UNQ, which is the PH domain of Akt1 bound to inositol 1,3,4,5-tetrakisphosphate. Although the structure and ligand binding to the PH domain changes in the presence of other domains, owing to the lack of availability of the full domain crystal structure, 1UNQ was selected for further structure-based approaches. Before docking the selected molecules, the generated grid was validated by redocking the inbound ligand, and the RMSD was found to be 1.455. Next, selected molecules and already known activators, such as baicalin, puerarin, chlorogenic acid, kaemferol-3-glucuronide and quercetin-3-glucuronide, were docked into the pocket. Chlorogenic acid showed the highest score (-6.829 kcal/mol) among the known activators. Among the selected molecules, 197105 showed the highest score of -4.978 kcal/mol, showing interactions with Lys 14, Arg 23, Arg 25, and Asn 53. Other molecules, such as 197105, 261126, 253878, 256085, and 123435, had binding affinities between − 4.876 and − 4.605 and interacted with key residues. The docking score, binding energy, and 2D interaction diagram of the known activators and selected molecules have been tabulated in Table S4 and Table 4, respectively.
Using Prime-MMGBSA, the binding energies for the ligand were calculated, indicating the stability of the protein-ligand complex created during docking. The binding energy of the known activator ranged from − 40.96 to -26.38 Kcal/mol, amongst which quercetin 3-O-glucuronide had the highest binding energy of -40.96 Kcal/mol and SC79, the lowest of -26.38 Kcal/mol. For the selected molecule, 123435 had the highest binding energy of -39.89 Kcal/mol, but in molecular docking, it showed the lowest dock score (-4.605 Kcal/mol) among selected molecules
1.1. Predicted ADME properties
The top five molecules, 123435, 197105, 253878, 256085, and 261126, were selected based on docking score, binding energy, and binding interaction with key residues and were subjected to ADME analysis using the Qikprop module. ADME properties included molecular weight, QPlogPo/w, QPlogS, rule of five, QPPCaco, percentage human oral absorption, CNS, and QPlogBB, which has been tabulated in Table 4. All molecules have molecular weights within the acceptable range of 200–500 Daltons, QPlogPo/w, representing the partition between octanol and water, and values within the acceptable range of -2 to 6.5. The predicted aqueous solubilities (QPlogS) were 6.5–0.5. None of the molecules violated any of these five rules. Except for 197105, all molecules had good permeability through QPPCaco and percentage of human oral absorption. All molecules had QPlogBB permeability within an acceptable range, but the predicted CNS activity was negative.
1.2. Molecular docking and binding free energy calculation of derivatives
After performing MD simulation of the top five molecules 123435, 197105, 253878, 256085, and 261126 selected after the ligand-based screening, molecular docking, MM-GBSA and MD simulation were performed, and results have been discussed in section 3.7. After MD simulation, it was found that molecules 261126 and 123435 were more stable and had many interactions with key amino acid residues like Lys 14, Arg 23, Arg 25, Asn 53, and Arg 86. To further explore the effect of substitution on the scaffold of 261126 and 123435 in the ability to bind and form a stable complex with PH domain of Akt1 derivates and to identify more potent Akt activator derivatives were searched for 261126 and 123435 molecules by using PubChem database. The 261126 molecule had 664 derivates, and the 123435 molecule had 55 derivates. These derivates were downloaded, prepared into the 3D structure, and docked in 1UNQ pocket, and the top five molecules based on docking score were selected (Table 4). The docking score of the derivates was more than − 4.978 kcal/mol which was the highest for molecules identified from the Asinex gold platinum database. The docking score for 12289533 was − 6.3 Kcal/mol, the highest amongst the derivates and selected molecules. The binding energy of the derivates was also much better compared to that of selected molecules from the database, the highest being − 76.71Kcal/mol for 12289533. The deviates also have druggable properties, as shown in Table 5. Finally, five derivates, 12289533, 12785801, 83824832, 102479045, and 6972939, were chosen for MD simulation.
1.3. Molecular dynamic simulation analysis
The results of MD simulation have been illustrated for 123435 and 261126 and their derivates like 12289533, 12785801, 83824832, and 6972939 in Fig. 4 and Fig. 5 and discussed below. Other selected molecules like 197105, 253878, 256085, and 102479045 (derivate of 123435) were performed, but due to inadequate results, they have been not discussed in the manuscript but have been illustrated in Fig. 1S and Fig. 2S.
The protein 1UNQ - ligand 261126 complex system was used for the simulation contained 6408 molecules of water, and the charge of the system was neutralized with two counter sodium ions. During the MD simulation, all the crucial XP interactions were observed along with the additional interactions (Tyr 18 and Ile 19). However, loss of salt bridge with Lys 14 was observed, which was present in XP docking results. The system was stable for 60 ns with RMSD ranging from 4.5 to 7.5 Ǻ after 60 ns, RMSD fluctuated up to 12 Ǻ. Till 60 ns, a strong H-bond and water bridge interactions were retained with Lys 14, Arg 25, and Asn 53. However, the change in the RMSD after 60ns might have occurred due to the loss of H-bond and water bridge interactions with three crucial amino residues, Lys 14, Arg 25, and Asn 53. Therefore, new derivatives of the 261126 ligand were identified (83824832, 12785801, 12289533) and subjected to the MD simulation in order to acquire the most potent Akt activators.
Derivative of 261126 ligand bearing 10-oxo-10H-pyrano[2,3-f]-8-carboxylate substitution (12289533) was selected for simulation study employing 1UNQ protein and 12289533 ligand complex. A fluctuation in the protein was observed till 60 ns, with the RMSD value ranging from 3 to 7.5 Ǻ. After 60 ns, the system was quite stabilized, and the RMSD value of protein and ligand was found to be 3.5 and 5 Ǻ, respectively. A drift in the RMSD value till 60 ns was exhibited by the loss of strong H-bond and water bridge interactions with Arg 23, Arg 25, and Asn 53 residues. However, the system was stabilized after 60 ns due to H-bond and water bridge interactions between ligand and protein amino residue, Arg 25. Lys 14 and Arg 86 residues have contributed to the overall stability of the simulation system by retaining the H-bond and water bridge interactions throughout the simulation. Interaction with Arg 86 residue was pertained by the carboxylate group of the pyran ring system, which has not appeared in the case of the 261126 ligand simulation system. Therefore, the system was quite stable in comparison with the 261126 ligand. However, it was subordinate in comparison to the 83824832 ligand. This might be due to the loss of H-bond and water bridge interaction with Arg 23, Arg 25, Asn 53, Glu 17, Tyr 18, and Ile 19 residues. The results of the 1UNQ-12289533 complex system revealed that the selected compound might show more potency as an Akt activator in line 261126 ligand. Therefore, it should be selected for future studies to interpret its potency towards Akt.
Another derivative of 261126 ligand bearing 3-methyl carboxylate substitution along with the 6,8-dichloro group on the 4-oxo-4H-chromene-2-carboxylate scaffold is entitled as 12785801. The selected ligand was subjected to 100 ns simulation. The fluctuation in the RMSD was reported throughout the simulation. However, the system stabilized between 65 to 95 ns during the simulation period. After 95 ns RMSD value fluctuated up to 10.5 Ǻ. The alteration in the system might be reported due to the introduction of the methyl carboxylate group at 3rd position, as it led to ligand fluctuation. Besides this, inconsistent H-bond and water bridge interactions of the ligand with Asn 53 and Arg 86 might lead to a significant drift in the RMSD value throughout the simulation.
Protein (1UNQ)-ligand complex of 83824832 derivatives of 261126 was subjected to 100 ns simulation. The system comprises 6394 molecules of water and 986 molecules of heavy atoms. Three counter ions of sodium neutralized the charge of the system. The protein-ligand complex was stable throughout the simulation compared to the 261126 ligand with RMSD values 2.2 and 3.2 Ǻ, respectively. A slight drift was observed between 20 to 40 ns, but the RMSD deviation was less than 3 Ǻ which is well within the acceptable range. Stability to the 83824832 ligand system was contributed by introducing the 3-oxide group in the 261126 ligand composed of the 4H-chromene-2-carboxylate scaffold, illustrated in Table S4. The introduction of the oxido group at the 3rd position helped to retain all the crucial H-bond and water bridge interactions with Lys 14, Arg 23, Arg 25, and Asn 53 residues. Besides, the system also maintained additional H-bond and water bridge interactions with Glu 17, Tyr 18, and Ile 19 residues which might also provide stability to the complex. The overall stability of the complex exhibited that 83824832 could act as a potential activator for Akt.
MD simulation study was conducted for the 123435 ligand, which has phenyl, hydroxy, and propanoate groups substitution at the 2nd, 5th, and 7th position of the 4-oxo-4H-chromene moiety. The protein-ligand complex system was stable throughout the simulation as per the RMSD plot. However, the RMSD value of the protein and ligand was found to be 3.8 and 7 Ǻ, respectively. A large difference in the protein and ligand RMSD value was observed due to the introduction of the phenyl group at 2nd position, which led to the contribution of high ligand fluctuation. Besides this, introducing the phenyl group also contributed to the loss of H-bond and water bridge interactions with crucial amino acid residues, such as Lys 14, Asn 53, and Arg 86.
Another simulation study was executed for 6972939, a derivative designed from the 123435 ligand with 1UNQ protein to identify potent lead as an Akt activator. The ligand is composed of 2-phenyl and 5,7-bis(oxy)dipropionate substitution on the 4-oxo-4H-chromene scaffold. In this simulation study, 1UNQ-6972939 complex system composed of 6500 water molecules and 986 heavy atom molecules. Three counter sodium ions neutralized the charge of the complex. RMSD of protein and ligand was found to be 3.5 and 7.5 Ǻ. A drift in the RMSD value was observed from 0 to 50 ns, and the system was stabilized after 50 ns. The ligand has retained most of the crucial H-bond and water bridge interactions with Lys 14, Arg 23, Arg 25, Asn 53, and Arg 86 residues. However, a high RMSD value was reported for protein and ligand. The reason behind this might be the ligand fluctuation due to the introduction of the phenyl group at the 2nd position. The Phenyl group at the 2nd position led to the major fluctuation in ligand RMSF value. Therefore, this ligand could be subjected to further studies with lead optimization to acquite potent Akt activators.
The simulation revealed that 83824832 and 12289533 could be potential lead compounds as Akt activators in the PI3K/Akt/mTOR pathway. The study also predicted approximately the SAR (Structure-activity relationship) of the compounds bearing 4-oxo-4H-chromene moiety. The substitution of electron-withdrawing groups (EWG) might be favorable at the 2nd position, like the carboxylate group in 261126, 83824832, and 12289533, instead of the electron donating group (EDG) like phenyl in 123435 and 6972939 ligands for 4-oxo-4H-chromene moiety as Akt activator. Small EWD groups (like oxido groups) might be favorable for activity at 3rd position. Substitution of the halogen group at the 6th position might favor activity. However, substitution at the 5th and 7th positions might be unfavorable for the activity. Hence, the overall study will help design the most potent Akt activators of the 4-oxo-4H-chromene scaffold using computational lead optimization approaches.