Validation of docking protocol

The validity of the docking protocol and the scoring function employed was confirmed by redocking the co-crystallized ligand of 7PCD (70I) at the active site. The docked pose was then superimposed with the ligand in the crystal structure, as shown in Fig. 3, and the root mean square deviation (RMSD) was calculated. The calculated RMSD was 0.5624 Å, which is far lower than the recommended 2 Å, indicating that the scoring function and docking protocol used in predicting ligand binding are highly reliable and valid.

Fig. 3

Superimposed structure of the docked pose (in green) and the co-crystallized ligand (in gray)

Similarity-based search

This study conducted a similarity-based search for compounds with potential HER2 TK inhibitory activity. The search was performed on the PubChem online database using Lapatinib as the query item, resulting in a library of 696 compounds out of over 1 billion compounds on the PubChem database that displayed 90% similarity to Lapatinib.

Molecular docking of PubChem-curated compounds and generation of structural analogs

Following filtering of compounds by RO5, the compounds were imported along with the standard (Lapatinib) into the Maestro software for docking against 7PCD. The docking followed two precision protocols (SP and XP) to identify the best ranking compounds and reduce false positives. Binding affinities of all docked compounds are detailed in supplementary Table 1. At the end of XP docking, compound 90,196,902 stood out as the best ranking compound with a binding energy of − 16.907 kcal/mol compared to − 9.805 kcal/mol recorded for Lapatinib. More details on the interaction of Compound 90,196,902 are seen in Table 1 and Fig. 4.

Table 1 Binding affinity and interaction profile of Compound 90,196,902 with HER2 TK active siteFig. 4

3D and 2D interaction of Compound 90,196,902

Sequel to the identification of Compound 90,196,902 as the best ranking compound among the PubChem-curated compounds, its structure was used as template in generation of random ligands using the DataWarrior software. These ligands were set to be similar to Compound 90,196,902 in structure and conformation, and more importantly, they retained the crucial functional groups on Compound 90,196,902 that were involved in its interaction with the HER2 TK. A total number of 1000 compounds were generated, and they were filtered by RO5 down to 443 compounds.

Molecular docking of generated compounds

The filtered compounds were subjected to SP docking, resulting in an output of 223 compounds with a binding affinity of ≥ − 9.985 kcal/mol. Based on the SP docking results, all compounds with a binding affinity of ≥ − 9.985 kcal/mol were further subjected to XP docking to draw more precise conclusions from its results, given its stringent scoring function. Following the XP docking, the top 5 ranking compounds (Compound_56, Compound_81, Compound_115, Compound_175, and Compound_339), along with Lapatinib, were subjected to IFD. The Maestro Prime refinement algorithm was utilized to refine several amino acid residues numbering up to 68 in the TK domain of HER2. Following the induced fit docking of these selected compounds, the best ranking pose was selected for the top 3 compounds as seen in Fig. 5, based on their docking score. The binding affinity of other compounds docked by SP, XP, and IFD can be found in supplementary tables 2, 3, and 4, respectively.

Fig. 5

3D and 2D interaction of Compound_56 (A), Compound_81 (B), Compound_339 (C), and Lapatinib (D)

Additionally, the active pocket containing the refined amino acid residues for the induced fit docking of the selected top-ranking compounds and Lapatinib was superimposed along with the original unrefined active pocket, as illustrated in Fig. 6. RMSD was calculated for 7PCD-Compound_56(15.43 Å), 7PCD-Compound_81(9.38 Å), 7PCD-Compound_339 (11.67 Å), and 7PCD-Lapatinib (9.33 Å). This comparison aimed to elucidate the changes induced in the conformation (arrangement) of residues in the binding pocket due to ligand, compared to those in the unbound protein. The binding energy and interaction profile of the selected compounds are detailed in Table 2 and Fig. 5.

Fig. 6

Superimposed original (gray) and refined HER2 TK domain active pocket used for induced fit docking of Compound_81 (green), Compound_56 (cyan), Compound_339 (purple), and Lapatinib (pink)

Table 2 IFD binding affinity (Kcal/mol) and interactions of the top 5 ranking compounds and Lapatinib in the HER2 TK active sitePost-docking MMGBSA calculation

Binding free energies (∆G_bind) reflect the overall strength of interaction between each compound and the target. Consequently, the binding free energy for Lapatinib, Compound_56, Compound_81, and Compound_339 was determined post-molecular docking using MM/GBSA calculations and findings are as seen in Table 3. The resulting binding free energies for Lapatinib, Compound_56, Compound_81, and Compound_339 were found to be − 64.48 kcal/mol, − 62.72 kcal/mol, − 79.46 kcal/mol, and − 67.63 kcal/mol, respectively. Analysis of the ∆G_bind_Hbond component reveals the contribution of hydrogen bonding to the overall binding affinity. Compound 339 exhibited the highest hydrogen bonding contribution (− 3.55 kcal/mol), followed by Compound 81 (− 2.33 kcal/mol), Compound 56 (− 2.25 kcal/mol), and Lapatinib (− 0.72 kcal/mol).

Table 3 MM/GBSA calculation of Lapatinib and the three promising drug candidates

Furthermore, it was noted that van der Waals interactions (∆G_bind_vdW) and solvation energies (∆G_bind_Solv) significantly contributed to the overall binding energy recoded. The most substantial van der Waals interactions were demonstrated by Lapatinib (− 68.58 kcal/mol), while the least were exhibited by Compound 339 (− 58.14 kcal/mol). The solvation energies varied across the compounds, with Compound 81 demonstrating the highest solvation energy (32.31 kcal/mol) and Lapatinib exhibiting the lowest (34.65 kcal/mol).

ADME and toxicity analysis

A total number of 20 descriptors were used in the analysis of ADME and toxicity properties of the selected compounds and Lapatinib as seen in Table 4. In terms of absorption values, all the compounds exhibited favorable properties, as indicated by their Caco-2 permeability, human intestinal absorption (HIA), and oral bioavailability. However, Lapatinib and Compound_339 showed a high probability (94 and 70%) of being a p-glycoprotein inhibitor.

Table 4 ADMET evaluation of Lapatinib and the three prospective drug candidates

Regarding distribution within the body, plasma protein binding (PPB) and blood–brain barrier (BBB) permeability were key descriptors. All compounds, including Lapatinib, demonstrated high PPB probabilities (> 90%) and very low BBB permeability probability.

In terms of metabolism, the evaluation focused on the Cytochrome P450 (CYP450) enzyme isoforms, assessing whether the compounds could induce or inhibit these enzymes. From the findings of this study, it was pointed out all the selected compounds and Lapatinib have low CYP450 induction probability. On the other hand, all compounds, including Lapatinib, were shown to inhibit all isoforms, except CYP450 3A4, which Lapatinib showed a moderate probability (44%) of inhibiting.

Excretion plays a vital role in removing unwanted substances and preventing the accumulation of toxic by-products of metabolism in the body. In line with this, clearance rates were predicted and All compounds showed favorable clearance rates. Toxicity descriptors were also evaluated to assess the toxic potential of the selected compounds. It was revealed from the result of toxicity analysis that Lapatinib has a high probability of hERG inhibition, while the selected compounds have a low probability except Compound_56 which also had a high probability of being an hERG inhibitor. Furthermore, Lapatinib stood out for its pronounced ames toxicity, a trait conspicuously absent in the three compounds under scrutiny. Unfortunately, all compounds, including Lapatinib, were found to be hepatotoxic, genotoxic and capable of inducing drug-induced liver injury, with Lapatinib exhibiting the most severe effects. Lastly, none of the selected compounds or Lapatinib displayed very low probability of being carcinogenic.

Molecular dynamics (MD) simulation

To provide better insight into the integrity and stability of molecules, including proteins and ligands, the complex of Compound_56, Compound_81, and Compound_339 and the standard drug Lapatinib with 7PCSD were subjected 100 ns of simulation. Parameters such as RMSD, RMSF, ROG, and H-bonds in an attempt to gain a better insight into their structural stability.

Root Mean Square Deviation (RMSD).

RMSD is an MD metric representing the degree of deviation of molecules from their original position over time. The graph in Fig. 7 illustrates the RMSD spectrum of 7PCD-Compound_56, 7PCD-Compound_81, 7PCD-Compound_339, and 7PCD-Lapatinib. The RMSD graph shows that 7PCD-Compound_81 exhibited the highest deviation, with a mean RMSD value of 0.454 nm, and 7PCD-Compound_81 can be inferred to have the lowest structural stability. 7PCD-Lapatinib displayed a steady spectrum throughout the simulation time, with minimal disturbance at several intervals (~ 27 ns, ~ 51 ns, and ~ 78 ns) and a mean RMSD of ~ 0.364 nm ± 0.042 nm. 7PCD-Compound_339 was shown to have a more stable RMSD spectrum, with a low mean RMSD of ~ 0.319 nm ± 0.020 nm. A spike was observed at ~ 9 ns of the simulation, and this complex maintained an equilibrated spectrum for the remaining ~ 91 ns of simulation time. Lastly, 7PCD-Compound_56 displayed the slightest deviation in the RMSD graph spectrum, having the lowest mean RMSD value of ~ 0.291 nm ± 0.017 nm. This suggests that 7PCD-Compound_56 has the highest structural stability among the compound set.

Fig. 7

Superimposed RMSD spectrums of the three promising drug candidates and the standard, Lapatinib in the HER2-TK domain (7PCD) active pocket

Root mean square fluctuation (RMSF)

It is of great importance to have an insight into the behavior of the individual amino acid residues in each of the protein–ligand complexes. Therefore, RMSF was calculated for the four complexes and the result is as presented in Fig. 8. It can be observed from the graph that all the complexes, including 7PCD-Lapatinib, experienced fluctuation in the same pattern at similar amino acid residues, with very little differences (as seen at residue 871 for 7PCD-Compound_339, residue 757 for 7PCD-Compound_81 and residue 814 for 7PCD_Lapatinib). 7PCD-Compound_339, 7PCD-Compound_81, and 7PCD-Compound_56 had final mean RMSF of ~ 0.144 ± 0.082 nm, ~ 0.169 ± 0.084 nm, and ~ 0.142 ± 0.082 nm, respectively, while the standard compound 7PCD-Lapatinib had a mean RMSF of ~ 0.173 ± 0.090 nm. It can be deduced from this that 7PCD-Lapatinib had the highest average RMSF and can be said to have low structural stability. On the other hand, 7PCD-Compound_56 had the lowest mean RMSF value.

Fig. 8

Superimposed RMSF graphs of the three promising drug candidates and the standard, Lapatinib in the HER2-TK domain (7PCD) active pocket

Radius of gyration (ROG)

The radius of gyration was used as a measure of the degree of compactness of the system during the simulation period. Having average ROG values of 1.980 ± 0.009 nm, 1.981 ± 0.015 nm, 1.931 ± 0.014 nm, and 1.971 ± 0.016 nm, respectively, and from the ROG spectrum depicted in Fig. 9, it can be deduced that 7PCD-Compound_56, 7PCD-Compound_81, 7PCD-Compound_339, and 7PCD-Lapatinib experienced minimal gyrations with very little difference in their average ROG value. However, 7PCD-Compound_81, though with a low, close to stable ROG spectrum, had the highest mean gyration value and can be said to have maintained the least compact conformation during the simulation. In contrast, 7PCD-Compound_339 experienced the least fluctuation during the simulation, with a mean ROG value even lower than the standard. It can be concluded that this complex was the most compact with the least degree of conformational mobility.

Fig. 9

Superimposed ROG graphs of the three promising drug candidates and the standard, Lapatinib in the HER2-TK domain (7PCD) active pocket

Hydrogen bond (H-Bond)

In this study, intermolecular hydrogen bonds between the three selected compounds, Lapatinib, and the target protein, 7PCD, were evaluated. The result of this evaluation is seen in Fig. 10. From the evaluation, the resulting average number of hydrogen bonds for 7PCD-Compound_56, 7PCD-Compound_81, 7PCD-Compound_339, and 7PCD-Lapatinib was 3.732 ± 0.727, 1.438 ± 1.336, 3.476 ± 1.034, and 1.030 ± 0.647, respectively. It is evident that 7PCD-Lapatinib has the lowest average number of hydrogen bonds among the four complexes at any given time, indicating potentially lower stability. Notably, 7PCD-Compound_56 exhibited the highest average number of intermolecular hydrogen bonds, solidifying its position as the most stable complex in comparison to the standard and other compounds.

Fig. 10

Superimposed H-bond graphs of the three promising drug candidates and the standard, Lapatinib in the HER2-TK domain (7PCD) active pocket