Systematic evaluation of the collected structures
In this study, we examined an in-house dataset consisting of 300 structures with resolutions ranging from 1.05 to 3.50 Å. Although these structures were originally used to study ambiguous electron densities around crystal ligands, we repurposed them for our current investigation. We systematically sorted these 300 structures using various approaches to facilitate our analysis. Initially, our visual inspection allowed us to shortlist approximately 100 structures where the ligands protruded towards the solvent. To further validate the solvent exposure of these ligands, we calculated their SASA. We selected ligands with a SASA of ≥ 15%, which resulted in the elimination of 70 candidate PDB entries.
Next, we generated the SRM and evaluated its interaction with the respective crystallographic ligands. This step led to the exclusion of 19 additional structures that lacked interactions with the SRM. Ultimately, we were left with a dataset of 11 structures, all of which had resolutions below 2.2 Å. At this resolution, we expected the electron density around the ligands to be sufficiently clear.
To ensure the accuracy of ligand fitting, we also evaluated the 2Fo-Fc electron density maps of the selected ligands. The details of the structures and the SASA values of the ligands in the selected PDB entries are shown in Tables 1 and 2, respectively. The entire workflow is illustrated in Fig. 1.
Table 1
Details of PDB Structures used in the current study
Sl No | PDB ID | Name of structures | Resolution in Å | Space group | R work | R free |
1. | 5EI4 | First domain of human bromodomain BRD4 in complex with inhibitor 8-(5-Amino-1H-[1, 2, 4]triazol-3-ylsulfanylmethyl)-3-(4-chlorobenzyl)-7-ethyl-3,7-dihydropurine-2,6-dione | 1.05 | P 21 21 21 | 0.171 | 0.192 |
2. | 3VNG | Crystal Structure of Keap1 in Complex with Synthetic Small Molecular based on a co-crystallization | 2.10 | P 21 21 21 | 0.196 | 0.225 |
3. | 2I4H | Structural studies of protein tyrosine phosphatase beta catalytic domain co-crystallized with a sulfamic acid inhibitor | 2.15 | C 1 2 1 | 0.169 | 0.246 |
4. | 3KRR | Crystal Structure of JAK2 complexed with a potent quinoxaline ATP site inhibitor | 1.80 | C 2 2 21 | 0.167 | 0.206 |
5. | 3PBB | Crystal structure of human secretory glutaminyl cyclase in complex with PBD150 | 1.95 | H 3 | 0.181 | 0.239 |
6. | 7F7W | JAK2-JH2 (JAK2/FLT3 inhibitor) | 1.83 | P 1 21 1 | 0.207 | 0.238 |
7. | 1Q0Z | Crystal structure of aclacinomycin methylesterase (RdmC) with bound product analogue, 10-decarboxymethylaclacinomycin A(DcmA) | 1.95 | P 1 21 1 | 0.157 | 0.185 |
8. | 6GL8 | Crystal structure of Bcl-2 in complex with the novel orally active inhibitor S55746 | 1.40 | P 21 21 21 | 0.178 | 0.194 |
9. | 1MMQ | Matrilysin complexed with Hydroxamate inhibitor | 1.90 | P 31 2 1 | 0.179 | NR* |
10. | 5ACL | EGFR kinase domain mutant "TMLR" with compound 24 | 1.49 | P 21 21 2 | 0.158 | 0.187 |
11. | 6G0G | KRAS-169 Q61H GPPNHP | 1.48 | P 21 21 2 | 0.193 | 0.220 |
NR*: Note Reported |
Table 2
Solvent-accessible surface area of the ligands present in the selected PDB Structures.
Sl No | Ligand name as per the PDB entry | PDB ID | Total SASA of ligand in Å2 | % Of ligand SASA in Å2 |
1. | 5NV | 5EI4 | 672.413 | 21 |
2. | FUU | 3VNG | 640.158 | 42.34 |
3. | UA1 | 2I4H | 825.092 | 42.41 |
4. | DQX | 3KRR | 811.189 | 16.02 |
5. | PBD | 3PBB | 637.913 | 20.24 |
6. | 36H | 7F7W | 832.607 | 18.22 |
7. | AKA | 1Q0Z | 1115.222 | 26.30 |
8. | F3Q | 6GL8 | 899.223 | 33.59 |
9. | RRS | 1MMQ | 750.370 | 33.33 |
10. | SAS | 5ACL | 688.272 | 29.70 |
11. | SAS | 6G0G | 688.272 | 26.69 |
SRM helps to determine the biologically relevant poses
Our molecular docking studies, conducted both in the presence and absence of SRMs, revealed a crucial role for SRMs in retaining biologically relevant/crystallographically significant poses. While comparing the docked poses with crystal poses, significant deviations in the orientation of the ligands were observed for most of the selected structures except 5ACL, 6GL8 and 6G0G in the absence of SRMs (Fig. 2A). On the other hand, in the presence of SRMs, the docking programs were able to generate binding poses closely resembling the crystal poses (Table 3).
Table 3
The RMSD and MM-GBSA calculated for the ligand in the presence and absence of symmetry-related molecules.
Sl No | PDB ID | RMSD in Å | MM-GBSA |
Symmetry | Symmetry |
- | + | - | + |
1. | 5EI4 | 7.12 | 0.84 | -76.19 | -87.63 |
2. | 3VNG | 8.26 | 0.89 | -53.42 | -56.01 |
3. | 2I4H | 7.93 | 0.29 | -62.88 | -107.24 |
4. | 3KRR | 8.66 | 0.49 | -104.96 | -119.52 |
5. | 3PBB | 7.59 | 0.25 | -77.77 | -100.89 |
6. | 7F7W | 8.54 | 1.00 | -86.03 | -92.82 |
7. | 1Q0Z | 8.71 | 0.29 | -137.15 | -141.28 |
8. | 6GL8 | 6.71 | 0.43 | -102.21 | -113.36 |
9. | 1MMQ | 6.32 | 0.96 | -98.97 | -112.05 |
10. | 5ACL | 8.09 | 0.36 | -53.02 | -58.25 |
11. | 6G0G | 7.77 | 1.00 | -31.48 | -66.36 |
Furthermore, an analysis of the binding free energies indicated that all entries exhibited better (more negative) binding free energies in the presence of SRMs when compared to their absence (Fig. 2B). This suggests that the binding poses were correctly predicted in the presence of SRMs, thereby favoring the biologically relevant/crystallographically significant conformations. This study underscores the importance of considering SRMs in molecular docking to achieve accurate and reliable predictions of ligand binding poses.
Mode of action of ligands: crystal structure vs. docked structures in the presence and absence of SRM
To delve deeper into the molecular mechanisms underpinning the ligand interaction with the protein structure, we analyzed the docking results in the absence of SRMs, focusing on the highest RMSD cases. The docking studies revealed multiple interactions for the ligands in the absence of SRMs. In all selected cases, the docked poses of ligands in the presence of SRMs closely resembled the crystal pose (Fig. 3).
In the case of 2I4H, the active site features a phosphate-binding element formed by the backbone residues from 1904 to 1910 and the side chain of D1910[23]. In the crystal structure, the ligand engages in hydrogen bonding with key residues such as R1910, S1905, V1908, and A1906, which are crucial for stabilizing the ligand within the active site. Additional active site residues, including D1910, C1904, V1908, Q1948, and H1871, also significantly contribute towards the ligand interactions. The phosphate-binding site, flanked by residues N1735, Y1733, and I1736, plays a pivotal role in ligand binding, with the p-ethyl phenylsulfamic acid moiety of the ligand forming hydrogen bonds with H1871, D1870, and R1910. The H1871 is particularly notable for its π-stacking interaction with the ligand's phenyl ring moiety. The enzyme's WPD-loop undergoes a significant conformational change upon ligand binding, facilitating the proper alignment of the catalytic residue D1870 and ensuring precise ligand fit within the active site(Fig. 3A) [24].
In contrast, in the docked structure without SRM, the p-ethyl phenylsulfamic acid adopts the same conformation as in the crystal structure due to being captured in a small cleft of the protein. However, the rest of the ligand orients entirely in the opposite way when compared to the crystal pose (Fig. 3B). Despite this, the docked structure maintains hydrogen bonding with several key residues such as S1905, A1906, V1908, G1909, R1910, D1870, and H1871. In the absence of SRM, the π-stacking interaction with H1871 is absent. This absence likely affects the proper alignment and conformational transition of the WPD-loop, in turn resulting in increased RMSD values for the ligand.
In the case of 5EI4, the WPF shelf—a conserved triad of tryptophan, proline, and phenylalanine—provides a hydrophobic platform that stabilizes ligand binding. The gatekeeper residue, isoleucine, shapes the acetyl-lysine binding pocket, influencing ligand access and contributing to selectivity and affinity. In the crystal structure, the ligand binds to the enzyme through several key interactions. The triazole moiety of the ligand forms a hydrogen bond with the residue D88, anchoring the ligand securely within the binding pocket. The triazolopyrimidinyl fragment is oriented within the ZA channel (formed by the Z and A helices in the protein) and forms van der Waals contacts with residues L92 and Q85, along with hydrogen bonds with N140. The pyrimidine ring in the ligand strongly interacts with Q85[25]. These interactions provide specificity for the ligand towards the active binding domain (Fig. 3C).
In the absence of SRM, the docked structure reveals an altered interaction pattern. The triazolopyrimidinyl fragment and triazole moiety are oriented in the opposite directions compared to the crystal structure. This orientation leads to a different pattern of hydrogen bonds between the ligand and active site residues. In the docked pose, different portions of the ligand form hydrogen bonds with N140 and D88. These interactions might be less effective, likely contributing to the higher RMSD, causing the ligand to adopt a conformation that deviates significantly from the optimal binding pose observed in the crystal structure (Fig. 3D).
In 1Q0Z, the active site is located at the interface of two domains. The ligand binds with its hydrophobic part extending deep into the pocket and its carbohydrate moiety at the entrance. Hydrophobic and stacking interactions occur between the ligand and the residue F134. The primary amino sugar forms additional interactions with residues M103, H219, Y220, and L222, and a hydrogen bond with D135 (Fig. 3E) [26].
While analyzing the docked pose in the absence of SRM, the ligand exhibited similar interactions within the binding cavity of the proteins. Specifically, a hydrogen bond with D135 and a π-stacking interaction with F134 were noted, consistent with the interactions seen in the presence of SRMs. However, the absence of SRM allowed more conformational freedom to the sugar units protruding towards the solvent. This increased flexibility led to an altered conformational orientation of this portion of the ligand, as reflected in the higher RMSD values. This deviation indicates that the ligand adopts a different binding pose in the absence of SRM, which may impact its binding affinity and overall stability within the active site (Fig. 3F).
The absence of SRMs appears to impact the stability and accuracy of ligand binding, altering interaction patterns and increasing RMSD values. These findings suggest that SRMs play a critical role in stabilizing key interactions within the protein-ligand complex, which in turn enhance the docking protocols and improve the accuracy of ligand binding predictions in computational studies. However, in the specific cases of PDB IDs: 6G0G, 6GL8 and 5ACL, we observed only minimal deviations for the ligands even in the absence of SRMs (Fig. 4A to C). To delve deeper into this phenomenon, we conducted a comprehensive structural analysis to understand the reasons behind the reduced RMSD for these molecules. Our investigation revealed that these ligands are confined within a relatively shallow and linear binding pocket. This particular structural characteristic of the binding site significantly influences ligand behavior, where the ligands adopt a parallel orientation within these binding sites (parallel orientation is with respect to the binding site), which inherently restricts their movement of ligand within the binding pockets, resulting in lower RMSD values.
MD Simulations confirmed the importance of SRMs
To gain deeper insights into the dynamic behaviour of protein-ligand complexes, MD simulations were performed under two distinct conditions: in the presence of SRMs (case-1) and in their absence (case-2). For the simulations, three specific structures were selected based on their highest RMSD values observed after the docking step in the absence of SRMs: PDB IDs 2I4H (RMSD = 7.64Å), 5EI4 (RMSD = 6.63Å), and 1Q0Z (RMSD = 2.97Å). The rationale behind selecting these structures was to investigate if the ligands could achieve the crystal conformation by undergoing structural rearrangement during the MD simulations. Upon comparing the docked poses of case-1 and case-2 with the crystal pose, it was observed that case-2 exhibited a higher RMSD than case-1. This increased RMSD in case-2 is likely attributed to the enhanced conformational freedom of the ligand regions exposed to the solvent surface. In contrast, in case-1, the presence of SRMs significantly restricts this conformational freedom of the ligand that is exposed to the solvents, leading to a more stable ligand conformation.
To our surprise, none of the selected ligands achieved the crystal conformation during MD simulations. However, these ligands exhibited very low RMSD profiles throughout the simulations. During the 100 ns simulations, the stability of both the protein and the ligands was well maintained, even in case-2 (Fig. 5). The lower RMSD observed in case-2 systems can be attributed to the other stable interactions formed between the ligands and the protein over the simulation period. We also assessed the ligand dynamics for both case-1 and compare with that of the crystal poses, revealed very low RMSD profile. It should be noted that the binding pose selected for the MD simulations was determined through molecular docking studies. This suggests that SRMs play a crucial role in maintain the biologically relevant/crystallographically significant pose.
A critical analysis of docked poses and MD trajectories was carried out to understand the differences in binding modes between case-1 and case-2. In the case of 1Q0Z, both case-1 and case-2 reproduced almost the same interactions between the ligand and the deeper cleft of the binding pocket. A common hydrogen bond with D135 and a face-to-face stacking interaction with F134 were observed in both case-1 and case-2 throughout the MD simulation, resulting in a stable orientation of the ligand inside the binding cavity. However, the differences between case-1 and case-2 became apparent in the protruding part of the ligand. In case-1, additional interactions were observed with residues from the SRM, such as A196, E197, and S280, which stabilized the ligand in a pose very similar to the crystallographic one. In contrast, in case-2, the formation of a hydrogen bond with L222 led to a different orientation of ligand, which remained stable throughout the MD simulation (Fig. 6A). This detailed comparison highlights how the presence of SRMs in case-1 reinforces the primary interactions within the binding pocket to adopt a conformation close to the crystallographic pose. Conversely, in the absence of SRMs, as seen in case-2, the ligand adopts a different stable orientation due to alternative interactions, emphasizing the significant role of SRMs in modulating ligand conformation and stability within the binding site (Fig. 6B).
In the case of 5EI4, the simulations for case-1 accurately reproduced the binding pose of the ligand ‘5NV’ within the binding pocket. The ligand’s two moieties, 1H-1,2,4-triazol-3-amine (moiety-1) and chlorobenzene (moiety-2), were observed to protrude towards the solvent surface. In case-1, the orientation of moiety-1 was determined by a strong hydrogen bond with D88 and an interaction with D96 from the SRM, while moiety-2 was stabilized primarily by a π-π stacking interaction with K99 from the SRM. In contrast, in case-2, these two moieties underwent a ~ 50° flip to the opposite side compared to the crystal pose resulting into a stacking interaction with residue W81 through moiety 1 and strong hydrogen bond with N140. This reorientation and the resulting stable interactions were further supported by the RMSD values extracted from the MD simulations of case-1 (Fig. 6C) and case-2 (Fig. 6D) relative to the crystal pose. In case-1, the presence of SRMs facilitated more extensive and stable interactions, closely mirroring the crystallographic pose. Conversely, in case-2, the absence of SRMs led to alternative interactions and a notable reorientation of the ligand, demonstrating the critical role of SRMs in maintaining ligand conformation and stability within the binding pocket.
In the case of 2I4H, the simulations of both case-1 and case-2 reproduced a stable hydrogen bond interaction between the ligand 'UA1' and residues located deep inside the active site, including R1910, G1909, V1908, A1906, S1905, and D1870. Similar to the observations in 1Q0Z, the protruding part of the ligand adopted a different orientation in case-2. In case-1, a stable hydrogen bond with N1734, along with interactions with R1715 and D1720, stabilized the ligand in a position closely resembling that of the crystallographic pose. These interactions effectively anchored the ligand within the binding site, minimizing conformational changes and maintaining a low RMSD. Conversely, in case-2, the ligand's portion which is protruding towards the solvent formed hydrogen bonds with K1811 and H1871, and exhibited a classical π-π stacking interaction. These interactions resulted the docked pose in a different orientation when compared to the crystal pose. Despite this deviation, all interactions remained stable throughout the simulation period, which was reflected in the lower RMSD for this ligand. In case-1, the presence of SRMs not only reinforced primary interactions but also introduced additional stabilizing contacts that guided the ligand towards a conformation similar to the crystallographic pose (Fig. 6E). In contrast, case-2 demonstrated that in the absence of SRMs, the ligand achieved stability through alternative interactions, leading to a different but stable orientation. These findings highlight the dynamic interplay between ligands and their binding environments, influenced significantly by the presence or absence of SRMs (Fig. 6F).