Structure-Based Virtual Screening and Protein–Protein Docking Analysis of ERBB2 and Associated Proteins for Pediatric Cancer Therapeutic Approaches

Keywords

AutoDock Vina - docking - ERBB2 - EGFR - toxicity prediction - virtual screening

Introduction

The ERBB2 protein, also known as human epidermal growth factor receptor 2 (HER2), plays a pivotal role in cell proliferation, survival, and differentiation.[1] Dysregulation or overexpression of ERBB2 has been linked to various cancers, such as neuroblastoma, gastric, breast, and ovarian cancers, making it an attractive therapeutic target.[2] Human ERBB2 was initially identified as an oncogene in rat brain tumors induced by chemicals. Subsequent analysis of human tissues revealed ERBB2 amplification in specific cases of salivary carcinomas and breast cancers with poor prognosis. These early findings sparked significant interest in ERBB2's role in human cancer, leading to a multitude of studies investigating the biology and clinical relevance of ERBB receptor signaling.[3] [4] In recent years, structure-based virtual screening and in silico analysis have emerged as powerful approaches to identifying potential inhibitors for specific protein targets.

This comprehensive study aimed to identify novel inhibitors for the ERBB2 protein through structure-based virtual screening and in silico analysis. Leveraging the wealth of available protein structural data, computational tools, and advanced algorithms, we sought to identify small molecules with the potential to interact with key binding sites on ERBB2 and disrupt its activity.

By employing state-of-the-art computational techniques, including molecular docking, protein–protein docking, and binding free energy calculations, we conducted an extensive screening of a diverse chemical library. Our focus on ERBB2-specific binding sites aims to prioritize compounds with high binding affinities and favorable amino acid interactions, thereby increasing the likelihood of successful inhibition. The identification of novel ERBB2 inhibitors holds promise for developing targeted therapies that can effectively combat cancers associated with ERBB2 dysregulation.[5] These findings may contribute significantly to advancing personalized medicine and improving the overall efficacy of cancer treatments.

Overall, this study represents a crucial step toward harnessing the power of computational approaches to expedite the discovery of new and potent ERBB2 inhibitors, fostering advancements in precision oncology and targeted therapeutics. The implications of these findings in the context of cancer therapy and future directions for experimental validation and drug development are also discussed.

Methodology

Data Collection

Relevant protein sequence and structural data for ERBB2 and other targets were sourced from publicly available databases and repositories, including PubMed, RCSB-PDB, and UniProt.[6] Our primary objective was to identify potential inhibitors from diverse natural databases, such as Enamine and LifeChemicals for ERBB2.[7] [8] These databases are distinguished for their wealth of natural compound derivatives and medicinal value. The compounds pinpointed through virtual screening present viable candidates for subsequent experimental studies. The compound library was downloaded from the official website of the respective chemical compound database. A protein–protein docking protocol and pertinent data were acquired through a comprehensive literature survey.

Protein and Ligand Preparation

Protein and ligand preparation are essential steps in molecular docking studies, and AutoDock Vina serves as a powerful tool for this purpose. In the initial phase, the protein structure is prepared by removing any water molecules, heteroatoms, and cocrystallized ligands that are not part of the binding site. The protein is then assigned appropriate atom types, charges, and torsion angles, and polar hydrogens are added.[9] Careful attention is given to the correct protonation states of ionizable residues, ensuring accuracy in the simulation.

On the other hand, ligands are prepared by removing any counterions, water molecules, or other nonessential entities. The ligand's three-dimensional structure is refined by optimizing bond lengths, angles, and torsion angles. Proper charges, atom types, and hybridization states are assigned to the ligand's atoms, ensuring compatibility with the chosen force field. Additionally, for flexible ligand docking, multiple conformations of the ligand were generated for a single ligand to explore potential binding modes. Based on the binding energy calculated for each ligand conformation, the potential ligand and its binding pose were considered.

Overall, this meticulous preparation of both protein and ligand ensures a reliable and accurate docking simulation with AutoDock Vina, enabling the exploration of protein–ligand interactions and the prediction of potential binding poses, ultimately aiding in drug discovery and future molecular design efforts.

Structure-Based Virtual Screening

Structure-based virtual screening is a valuable computational approach employed to identify potential drug candidates by predicting their binding affinities to a target protein (ERBB2), using the AutoDock Vina tool.[10] In this study, the LifeChemicals and Enamine compound databases were utilized as valuable sources of diverse small molecules. The prepared protein structure and ligand molecules were utilized for the virtual screening studies.

The virtual screening was conducted by docking each compound from the LifeChemicals and Enamine databases into the active site of the ERBB2 protein. AutoDock Vina exhaustively sampled binding poses and ranked the compounds based on their calculated binding energies, reflecting the strength of their potential interactions with the protein. The top-ranking compounds, with the most favorable binding energies, were further analyzed to assess their predicted binding modes, hydrogen bonding patterns, and key interacting residues within the binding site. Default protocols of the AutoDock Vina were implemented throughout the virtual screening analysis.

This structure-based virtual screening using AutoDock Vina, coupled with the utilization of the LifeChemicals and Enamine databases, provided a systematic and efficient means to prioritize promising small molecules for potential ERBB2 inhibition.[11] [12] The results from this study contribute valuable insights into the realm of drug discovery, guiding experimental efforts toward the identification and development of novel therapeutic agents targeting ERBB2-associated diseases.[13]

Functional Partner Discovery with Bioinformatics Tools

Functional partners in a pathway are crucial components that interact with each other to execute specific biological processes, such as cancer pathogenesis. Identifying these partners is essential for understanding the intricate molecular mechanisms that govern cellular functions. Tools and servers such as STRING,[14] KEGG,[15] and REACTOME[16] provide valuable resources to unravel these interactions and uncover the network of relationships within a pathway. STRING is a powerful bioinformatics resource that specializes in predicting protein–protein interactions. It integrates various sources of experimental and computational data to construct a comprehensive network of functional associations between proteins. KEGG is a widely used resource for understanding biological pathways and the interactions among genes and proteins in various organisms. REACTOME is another valuable resource for pathway analysis, providing a curated knowledge base of biological pathways. Consolidating all the results and identifying the potential functional partner will be subjected to further computational analysis.

Protein–Protein Docking

The protein–protein docking between two or more proteins was performed using the HADDOCK 2.4 server, an advanced computational tool specifically designed for modeling macromolecular interactions.[17] The main objective of this docking study was to predict the potential binding modes and interface interactions between these two important proteins, which play crucial roles in signaling pathways and cellular processes. Parameters were set to define ERBB2 as the “receptor” and the identified functional protein as the “ligand,” given their respective roles in the interaction.

Active residues, crucial for the protein–protein interaction, were defined based on the known literature and biological context. These active residues were set to be unbound and flexible during the docking simulations. The initial docking runs were performed using the rigid body docking mode, allowing for a global search of possible binding orientations. HADDOCK generated an ensemble of docking solutions for further refinement. The top-scoring docking solutions were selected for the semiflexible refinement stage. During this step, the side chains of the active residues were allowed to optimize their positions, employing a simulated annealing protocol. The resulting docked complexes were analyzed to identify the most probable binding mode, interface residues, hydrogen bonding, and hydrophobic interactions between ERBB2 and functional proteins.[5] [18] The binding energy of the top-ranked solution was used as an indicator of the stability of the predicted complex.

Toxicity Prediction Analysis

The toxicity prediction of the identified lead compounds was performed using the ProTox-II server, a widely recognized computational tool specifically designed for predicting the potential toxicity of small molecules.[19] This step was crucial in the drug discovery process to assess the safety profile of the identified lead compounds before further experimental investigations. The top five potential lead compounds were selected based on their favorable binding energies and predicted binding modes from our structure-based virtual screening analysis. ProTox-II provided predictions for multiple toxicity endpoints, including mutagenicity, hepatotoxicity, carcinotoxicity, and others, using validated models. The predicted toxicity scores were interpreted, considering both the individual toxicity endpoints and the overall toxicity profile.[20] The toxicity predictions were critically analyzed in the context of the intended therapeutic application and the known safety standards for pharmaceutical compounds. This approach ensures that the lead compounds with the most promising efficacy and favorable safety profiles are advanced in the drug discovery pipeline, enhancing the overall success rate of the drug development process.

Ethics

No human participants/subjects were involved in this study.

Results

Virtual Screening Studies

The structure-based drug design study conducted using AutoDock Vina yielded promising results in the search for potential drug candidates from the Enamine and LifeChemicals databases. After rigorous filtering based on the Lipinski Rule of Five, which ensures drug-likeness and favorable pharmacokinetic properties, a total of 385,000 and 137,000 compounds were retained from the LifeChemicals and Enamine databases, respectively, for further analysis. The AutoDock Vina program automatically generates the grid map and presents clustered results to users in a transparent manner. Within Vina, diverse stochastic global optimization techniques, including genetic algorithms, simulated annealing, and particle swarm optimization, were employed. The active site cavity was carefully chosen, followed by postdocking steps involving energy minimization and H-bond optimization.

Upon thorough docking simulations, we identified 10 compounds that exhibited notably favorable binding energies, suggesting strong interactions with the receptor ([Table 1]). The binding energy calculated for the top compounds ranged from –8.346 to –6.296 kcal/mol. The interacting residues were Thr5, Arg412, Leu414, and Ser441 with the docked ligands ([Fig. 1]). These compounds demonstrated specific amino acid interactions within the binding site or active site of the receptor, reinforcing the potential for selective binding and biological activity. The identification of these 10 compounds with both promising binding energy and significant interactions with the receptor represents a significant outcome of our study.

Table 1 List of potential compounds identified through structure-based virtual screening

Compound ID	Binding energy (kcal/mol)	Interacting residues
LC_87763	–8.346	Thr5, Arg412
LC_33378	–7.858	Thr5, Arg412
LC_27122	–7.409	Thr5, Leu414, Ser441
Enamine_101102	–7.359	Thr5, Gly6, Gly411, Leu414
LC_87632	–7.221	Thr5, Arg412
Enamine_95473	–6.729	Thr5, Leu414
Enamine_68284	–6.489	Thr5, Gly6, Gly411
LC_48628	–6.385	Thr5, Leu414
LC_12121	–6.337	Thr5, Gly411
LC_34889	–6.296	Thr5

Fig. 1 Two-dimensional interaction diagram of the top five compounds in the active site of the protein. (A) LC_87763; (B) LC_33378; (C) LC_27122; (D) Enamine_101102; and (E) LC_87632.

From the results we observe, most of the potential compounds were identified from the LifeChemicals database. Out of the 10 compounds, 7 from LifeChemicals and 3 are from the Enamine database. The identified compounds were found to have hydrogen bond, salt-bridge, and pi–pi interactions.

Protein Functional Partners

The comprehensive analysis of functional partners using STRING, KEGG, and REACTOME has yielded crucial insights into the intricate network of interactions involving the ERBB2 protein. Through STRING, we uncovered a multitude of potential interaction partners, which were further enriched and contextualized within biological pathways and processes using the KEGG and REACTOME databases. The rigorous exploration highlighted a significant finding: among the various candidates, epidermal growth factor receptor (EGFR) emerged as the most prominent and compelling functional partner for the ERBB2 protein ([Fig. 2]). This outcome was supported by multiple lines of evidence, including high-confidence protein–protein interaction scores, shared pathways, and known biological relevance. The identification of EGFR as the best functional partner of ERBB2 underscores its central role in cellular signaling and its potential significance in various physiological and pathological contexts. Further experimental validation and functional studies will be crucial to decipher the specific mechanisms through which this interaction contributes to cellular processes and disease pathways, potentially opening new avenues for therapeutic interventions targeting the ERBB2–EGFR complex.

Fig. 2 (A) Protein–protein docking interaction and molecule binding pose representing the interaction site. (B) Interacting amino acid residues from chain A (ERBB2) and chain B (epidermal growth factor receptor).

Protein–Protein Docking

The protein–protein docking studies conducted using HADDOCK 4.2 server provided critical insights into the binding interactions between ERBB2 as the receptor protein and EGFR as the ligand. The docking simulations yielded a range of potential binding modes, allowing us to explore the diverse conformations in which these two proteins may interact. Through comprehensive analysis, we identified a highly favorable binding mode that demonstrated strong binding energy, indicative of stable and specific interactions between ERBB2 and EGFR ([Table 2]). The best cluster is observed as Cluster 1 with a Haddock score of –95.3, lowest root mean square deviation (RMSD) of 0.6 Å, –384 kcal/mol of Electrostatic energy, and –95.9 kcal/mol of Van der Walls energy. The structure selected from the initial cluster exhibits a Haddock score of –91.46, a minimal RMSD of 0.4 Å, and recorded energy values of –311 kcal/mol for Electrostatic forces and –90.2 kcal/mol for Van der Waals interactions. The important residues of EGFR found in/around the active site comprises nearly 30 amino acids from 353 to 359 and 448 to 464; specifically, Gln408, Lys463, Phe412, and Asp436 are vital residues. The detailed examination of the docked complex revealed key amino acid residues involved in the binding interface, highlighting the precise molecular interactions contributing to the formation of the ERBB2–EGFR complex ([Fig. 2]). These findings shed light on the potential functional implications of this protein–protein interaction, further underscoring the significance of the ERBB2–EGFR interaction in cellular signaling pathways and disease contexts.

Table 2 Haddock score for the docked complex of ERBB2 and EGFR of Cluster 1

Parameters	Values
Cluster size	53
Haddock score	–95.3 ± 12.5
Van der Waals energy	–95.9 ± 5.1
Electrostatic energy	–384.2 ± 42.6
RMSD from the overall lowest-energy structure	0.6 ± 0.4
Desolvation energy	0.0 ± 3.3
Restraints violation energy	774.2 ± 27.8
Buried surface area	3,259.9 ± 151.7
Z-score	–2.1

Abbreviations: EGFR, epidermal growth factor receptor; ERBB2, ---; MSD, root mean square deviation.

Toxicity Prediction

The toxicity analysis of the 10 compounds performed using the ProTox-II server has yielded encouraging results. The comprehensive evaluation encompassing multiple toxicity endpoints has revealed that the majority of the compounds demonstrated favorable profiles with no indication of significant toxicity concerns ([Table 3]). Remarkably, 6 out of the 10 compounds exhibited excellent prediction scores, indicating their potential safety in terms of the assessed toxicity endpoints. It is worth noting that, while the majority of compounds were in an acceptable stage in terms of predicted toxicity, four compounds did exhibit relatively lower prediction scores, suggesting the need for cautious consideration or further assessment before advancing them for experimental testing.

Table 3 Toxicity prediction of the identified compounds using ProTox-II server

Compounds	Predicted toxicity class	Prediction accuracy	Hepatotoxicity	Carcinogenicity	Immunotoxicity	Mutagenicity	Cytotoxicity
LC_87763	CLS 4	92.0%	INACT	INACT	INACT	INACT	INACT
LC_33378	CLS 5	100%	INACT	INACT	INACT	INACT	INACT
LC_27122	CLS 4	88.54%	INACT	INACT	INACT	INACT	INACT
Enamine_101102	CLS 6	63.70%	INACT	INACT	INACT	INACT	INACT
LC_87632	CLS 5	70.97%	INACT	INACT	INACT	INACT	INACT
Enamine_95473	CLS 5	77.80%	ACT	INACT	INACT	INACT	INACT
Enamine_68284	CLS 4	100%	INACT	INACT	INACT	INACT	INACT
LC_48628	CLS 6	92.98%	INACT	INACT	INACT	INACT	INACT
LC_12129	CLS 4	71.90%	ACT	INACT	INACT	INACT	INACT
LC_34889	CLS 5	98.67%	INACT	INACT	INACT	INACT	INACT

Abbreviations: ACT, active; CLS, class; INACT, inactive.

Discussion

Virtual screening's ability to expedite lead discovery, optimize resources, and streamline the drug development pipeline underscores its vital role in modern pharmaceutical research, offering a critical bridge between computational analysis and experimental validation, ultimately driving the development of innovative therapeutic agents that hold the potential to address unmet medical needs and improve patient outcomes. Recently, a study investigated similar workflow using Zinc database with 300 natural compounds. The successful docking of ERBB2 and EGFR using HADDOCK 4.2 demonstrates the utility of this computational tool in deciphering protein–protein interactions, paving the way for future experimental validation and the design of targeted interventions aimed at modulating this critical interaction for therapeutic purposes. These results underscore the importance of such computational toxicity analysis in early stage drug discovery, enabling the identification of compounds with favorable safety profiles and highlighting those that may require additional scrutiny. Further validation through in vitro and in vivo toxicity studies is essential to confirm these predictions and refine our understanding of the overall safety profile of the identified compounds. Nonetheless, the majority of the compounds showing promising results in the ProTox-II analysis represent encouraging candidates for further investigation in drug development efforts.

Conclusion

The results obtained from the combination of various bioinformatics tools and computational methodologies in our study provide a comprehensive understanding of the molecular interactions, binding affinities, functional partnerships, and toxicity profiles of the identified lead compounds. Through protein–protein docking studies utilizing HADDOCK 4.2, we elucidated the binding modes between ERBB2 and EGFR, shedding light on potential interaction mechanisms crucial for cellular processes. Our analysis of functional partners using STRING, KEGG, and REACTOME revealed the pivotal role of EGFR as a functional partner of ERBB2, enhancing our understanding of their interplay in biological pathways. The subsequent toxicity prediction using ProTox-II enabled the early assessment of potential safety concerns, and the identification of compounds demonstrating acceptable toxicity profiles highlights the importance of computational tools in prioritizing lead compounds for further experimental validation. This integrated approach, encompassing molecular docking, pathway analysis, and toxicity assessment, serves as a robust framework for efficient and informed decision-making in drug discovery, accelerating the identification and development of promising candidates while minimizing the risk of adverse effects in the later stages of drug development.

• References

• 1 Ashraf SQ, Nicholls AM, Wilding JL, Ntouroupi TG, Mortensen NJ, Bodmer WF. Direct and immune mediated antibody targeting of ERBB receptors in a colorectal cancer cell-line panel. Proc Natl Acad Sci U S A 2012; 109 (51) 21046-21051
• 2 Pectasides E, Bass AJ. ERBB2 emerges as a new target for colorectal cancer. Cancer Discov 2015; 5 (08) 799-801
• 3 Gilbertson RJ. ERBB2 in pediatric cancer: innocent until proven guilty. Oncologist 2005; 10 (07) 508-517
• 4 Hynes NE. ErbB2: from an EGFR relative to a central target for cancer therapy. Cancer Res 2016; 76 (13) 3659-3662
• 5 Sivaiah G, Raveesha R, Prasad SBB. et al. Synthesis, anticancer activity and molecular docking of new pyrazolo[1,5-a]pyrimidine derivatives as EGFR/HER2 dual kinase inhibitors. J Mol Struct 2023; 1289
• 6 Bateman A, Martin MJ, Orchard S. et al; UniProt Consortium. UniProt: the universal protein knowledgebase in 2021. Nucleic Acids Res 2021; 49 (D1): D480-D489
• 7 Loganathan L, Natarajan K, Muthusamy K. Computational study on cross-talking cancer signalling mechanism of ring finger protein 146, AXIN and Tankyrase protein complex. J Biomol Struct Dyn 2020; 38 (17) 5173-5185
• 8 Loganathan L, Briget Kuriakose B, Lobelle Sampayan E, Muthusamy K. Targeting renin receptor for the inhibition of renin angiotensin aldosterone system: an alternative approach through in silico drug discovery. Comput Theor Chem 2022; 1208: 113541
• 9 Loganathan L, Gopinath K, Sankaranarayanan VM. et al. Computational and pharmacogenomic insights on hypertension treatment: rational drug design and optimization strategies. Curr Drug Targets 2020; 21 (01) 18-33
• 10 Seeliger D, de Groot BL. Ligand docking and binding site analysis with PyMOL and Autodock/Vina. J Comput Aided Mol Des 2010; 24 (05) 417-422
• 11 Li Q, Shah S. Structure-based virtual screening. Methods Mol Biol 2017; 1558: 111-124
• 12 Shaikh N, Linthoi RK, Swamy KV, Karthikeyan M, Vyas R. Comprehensive molecular docking and dynamic simulations for drug repurposing of clinical drugs against multiple cancer kinase targets. J Biomol Struct Dyn 2023; 41 (16) 7735-7743
• 13 Loganathan L, Muthusamy K. Investigation of drug interaction potentials and binding modes on direct renin inhibitors. A computational modeling studies. Lett Drug Des Discov 2018; 15
• 14 Snel B, Lehmann G, Bork P, Huynen MA. STRING: a web-server to retrieve and display the repeatedly occurring neighbourhood of a gene. Nucleic Acids Res 2000; 28 (18) 3442-3444
• 15 Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res 1999; 27 (01) 29-34
• 16 Croft D, Mundo AF, Haw R. et al. The Reactome pathway knowledgebase. Nucleic Acids Res 2014; 42 (Database issue, D1): D472-D477
• 17 van Zundert GCP, Rodrigues JPGLM, Trellet M. et al. The HADDOCK2.2 web server: user-friendly integrative modeling of biomolecular complexes. J Mol Biol 2016; 428 (04) 720-725
• 18 Tai W, Mahato R, Cheng K. The role of HER2 in cancer therapy and targeted drug delivery. J Control Release 2010; 146 (03) 264-275
• 19 Banerjee P, Eckert AO, Schrey AK, Preissner R. ProTox-II: a webserver for the prediction of toxicity of chemicals. Nucleic Acids Res 2018; 46 (W1): W257-W263
• 20 Raies AB, Bajic VB. In silico toxicology: computational methods for the prediction of chemical toxicity. Wiley Interdiscip Rev Comput Mol Sci 2016; 6 (02) 147-172