Discovery of drug candidate to inhibit bronchogenic carcinoma genes biomarkers based on drug repurposing

Abstract

Objective

Lung cancer is one of the most common and challenging cancers to treat. Advances in research have led to the development of targeted drug therapies that provide suitable options for patients with bronchogenic lung cancer. The aim of the current study is to inhibit genes that have been introduced as biomarkers for the diagnosis of bronchogenic lung cancer.

Methods

This study leverages computational methods to identify repurposable drugs targeting key biomarkers—MLKL, YWHAG, OAS3, TFRC, NXA2, CDK6, NTN1, CD59, RRAS2, and CYP51A1—for cancer treatment. As a primary screening tool, AutoDock Vina was employed for molecular docking to evaluate the binding affinity of existing drugs against these targets. Following the initial screening, molecular dynamics simulations were utilized to select the most stable and ideal drug candidates with specific inhibitory therapeutic properties from the pool identified by docking. This integrated workflow demonstrates an efficient path for discovering new therapeutic uses for existing drugs against a defined panel of cancer biomarkers.

Results

For the ten genes under investigation, structurally reliable PDB entries were chosen as the basis for subsequent analyses. The herbal compound Dracorubin, along with 2,299 ligands collected from the PubChem database, was assessed. Molecular dynamics simulations indicated that Dracorubin sustained stable and meaningful inhibitory interactions with the proposed biomarkers.

Conclusion

The study aimed to find new uses for existing drugs for Bronchogenic lung cancer. Using computational methods like molecular docking and dynamics, researchers prioritized compounds that bind strongly to target proteins. These candidate drugs, prioritize a computational lead requiring experimental validation of Bronchogenic lung cancer progression.

Introduction

Lung cancer is the second most common cancer worldwide after breast cancer. In 2025, it is estimated that 226,650 new cases of lung cancer will be diagnosed in the United States, with 124,730 deaths attributed to this disease [1, 2].

Despite significant advances in the early diagnosis and treatment of lung cancer, the 5-year overall survival (OS) rate remains disappointingly low, at only 25% [3]. Survival outcomes can be improved through improvements in diagnostic and therapeutic strategies, such as disease staging and identification of stage-specific biomarkers. Despite the availability of various treatment regimens, the 5-year overall survival rate ranges from 4% to 17%, depending on the stage and anatomical location of the tumor [4]. Therefore, the development and discovery of drugs tailored to disease stage is of particular importance, underscoring the urgent need for discovery and development of novel drugs for lung cancer. In recent years, using in silico approaches has attracted considerable attention from pharmaceutical companies due to their ability to significantly reduce the time and cost of drug development [5, 6]. Drug discovery and development remain lengthy and expensive processes, partly due to increasingly stringent regulatory expectations that raise trial failure rates and overall development costs [7, 8]. These financial and temporal burdens continue to pose significant obstacles for the pharmaceutical sector [9–11].

Drug repurposing, which aims to identify new therapeutic uses for existing drugs based on data-driven analyses, has been widely employed in drug discovery to develop treatments for various human diseases. Drug repurposing aims to uncover new uses for known compounds through integrative data analysis [12]. By building on existing clinical and safety information, this approach can shorten development timelines and lower overall risk [13]. Because these agents have already been evaluated in humans, their reuse generally carries a reduced chance of toxicity-related failure and may expedite regulatory progression [14].

Drug repurposing and drug reuse are similar concepts, often used to describe the strategy of applying an approved drug candidate to a completely new indication [15]. Pharmaceutical companies, based on the drug repurposing hypothesis, which suggests that reusing drugs that have already undergone clinical trials reduces the risk of failure in future trials due to lower toxicity and can therefore lead to faster drug approval, actively pursue drug repurposing [16, 17]. In this way, they can address a disease condition in patients different from the originally approved indication, engaging in drug repurposing or rediscovery as an alternative to traditional drug development. Among the 1,248 FDA-approved drugs, over 270 are classified as anti-neoplastic agents [17]. Lung cancer remains one of the most prevalent malignancies worldwide; according to the 2020 GLOBOCAN report, it accounts for roughly 2.2 million new cases and over 1.7 million deaths each year [18].

If a drug intended for repurposing has already been reviewed by the FDA and is administered through the same route at a similar or lower dose compared to the approved product, its repositioning is considered relatively straightforward, reliable, safe, and cost-effective [19].

People have long gravitated toward natural remedies, and growing worries about side effects from synthetic drugs are pushing that interest even higher. As a result, researchers are taking a closer look at plant-based treatments for many diseases, making it more important than ever to find the compounds that truly work. The problem is that traditional discovery is time-consuming and expensive. That’s where in silico tools help. They can quickly scan and rank plant compounds, reducing lab work and accelerating the search for promising therapies.

The present study builds upon previous research [20], entitled “Identifying Candidate Biomarkers for Detecting Bronchogenic lung cancer Stages Using Metaheuristic Algorithms Based on Information Fusion Theory”. This study sought small-molecule inhibitors for ten key bronchogenic lung cancer biomarkers. A library of 2,299 plant-derived ligands was screened using AutoDock Vina, and molecular dynamics simulations assessed the stability and persistence of protein-ligand interactions.

Literature review

Several studies have focused on identifying biomarker genes through bioinformatics analyses of transcriptional profiles in lung cancer patients. These studies have confirmed the significance of the identified biomarkers by demonstrating their prognostic value or associating these genes with relevant tumor indicators. However, few studies have been conducted to identify drugs specifically targeting the stage-specific biomarker genes.

In the study by Li et al., a novel drug repurposing approach was proposed based on transcriptional data and chemical structures using deep learning techniques [21]. Pimozide is an anti-dyskinesia agent used to suppress motor and phonic tics in patients with Tourette’s disorder [22]. They found that Pimozide is a strong candidate for treating non-small cell lung cancer, and its cytotoxicity against A549 cell lines was confirmed.

Jain et al. provided a comprehensive review of drug repositioning for anticancer applications, with a focus on the targeted repurposing of non-cancer drugs [23]. According to the information presented in this review, targeting the Ras/Raf/MEK/ERK pathway represents a promising and alternative strategy for NSCLC treatment. Additionally, the MAPK signaling pathways were highlighted as a critical aspect of NSCLC, contributing to advancements in therapeutic approaches for this cancer.

Sultana et al. investigated six genes involved in NSCLC [24]. They analyzed the functions, pathways, prognostic significance, and therapeutic implications of these genes using integrated bioinformatics and statistical approaches, and demonstrated the effectiveness of their approach in producing meaningful results. Seven candidate drugs were proposed using knowledge graphs through network pharmacology and molecular docking analysis [25].

Inci et al. found that resistance mechanisms can develop rapidly in the NSCLC microenvironment and primarily affect specific detectable genes such as EGFR and ALK [26, 27], highlighting the need for personalized therapeutic approaches. While cancer nanomedicine facilitates targeted treatment, one of the main challenges remains achieving drug release at the desired site with the appropriate dosage and timing [28].

Today, significant advances in drugs, diseases, and bioinformatics provide substantial opportunities for developing novel drug repurposing approaches through a comprehensive understanding of pharmacological data. For example, sildenafil is used to treat erectile dysfunction. It is a phosphodiesterase type 5 inhibitor of cyclic guanosine monophosphate and was originally developed by Pfizer to treat coronary artery disease [29].

Minoxidil was originally developed as an oral medication for the treatment of hypertension. Researchers observed that some patients taking the drug experienced increased hair growth as a side effect [30].As another example, isoniazid is an antibiotic commonly used for the treatment and prevention of tuberculosis and other diseases. In multiple sclerosis, isoniazid is prescribed for certain types of tremors [31]. The approach developed in this paper can be extended to other diseases, offering promising prospects for drug discovery targeting disease stages.

Proposed method

Flowchart of proposed research method

This study focuses on biomarkers used to identify the stages of bronchogenic lung cancer, including Stage I, Stage II, Stage III, and Stage IV. Through the repurposing of existing drugs, the ultimate goal is to identify an optimal drug that provides maximal coverage across all disease stages. In other words, the best prioritized a computational lead requiring experimental validation as one that can treat and prevent the progression of all four stages of the disease. The study is based on the strategy of inhibiting highly expressed genes, and the procedures are carried out according to the proposed flowchart (Fig. 1).

Fig. 1.

Schematic view of the study

Materials and methods

Protein 3D structure search and download

For the genes validated in the previous study, the corresponding target proteins were identified, and their 3D structures were retrieved and stored. Structure-based drug design was performed using virtual screening of 3D structures obtained from the Protein Data Bank (PDB, https://www.rcsb.org/), as presented in Table 1.

Table 1.

PDB selection criterion values and UniProt IDs, PDB IDs, and biological role in lung cancer

Target	Uniprot ID	Representative PDB ID	Biological rationale in lung
MLKL (Mixed lineage kinase domain-like protein)	Q8NB16	6ZVO	Necroptosis effector; its dysregulation promotes tumor cell survival and resistance to cell death; cytoplasmic localization makes it accessible to small molecules
YWHAG (14-3-3 protein gamma)	P61981	6F0Y	Adapter protein regulating proliferation and apoptosis via PI3K/AKT and MAPK signaling; overexpressed in NSCLC; cytoplasmic/nuclear
OAS3 (2′–5′-oligoadenylate synthetase 3)	Q9Y6K5	4S3N	Interferon-induced enzyme modulating antiviral and immune pathways; affects tumor immune microenvironment and cell survival
TFRC (Transferrin receptor protein 1)	P02786	3S9L	Mediates cellular iron uptake; overexpressed in proliferating lung tumor cells; cell surface receptor suitable for targeted inhibition.
NXA2 (Annexin A2)	P07355	2HYW	Involved in cytoskeletal reorganization, angiogenesis, and metastasis; high ANXA2 correlates with EMT and poor prognosis
CDK6 (Cyclin-dependent kinase 6)	Q00534	1XO2	Governs G1–S phase transition; overactivation drives uncontrolled proliferation; validated drug target with known inhibitors
NTN1 (Netrin-1)	O95631	4DNX	Secreted guidance cue supporting tumor cell migration, survival, and angiogenesis; implicated in lung tumor metastasis
CD59 (Protectin)	P13987	2J8B	Complement regulatory protein overexpressed in tumor cells to evade immune destruction; membrane-bound and druggable via antibody or small-molecule binding
RRAS2 (Related RAS viral oncogene homolog 2)	P62070	3KKO	Small GTPase that activates MAPK/PI3K pathways; mutations or overexpression promote oncogenic signaling and proliferation
CYP51A1 (Lanosterol 14-α-demethylase)	Q16850	3LD6	A cytochrome P450 enzyme involved in cholesterol biosynthesis; upregulated cholesterol metabolism supports membrane synthesis and oncogenic signaling in lung cancer. Inhibition of CYP51 may suppress tumor growth and proliferation

Molecular docking

Ten lung cancer biomarkers were selected for virtual screening (Table 1). Plant-derived ligands (2,299 compounds) were processed with OpenBabel to add hydrogens and generate 3D structures, while protein targets were prepared and converted to PDBQT format (Supplementary Data S1). Docking was performed using AutoDock Vina v1.2.5 with defined grid centers, box sizes, and binding site residues (Table 2), followed by ADMET and drug-likeness evaluation via OSIRIS (https://www.cheminfo.org/flavor/cheminformatics/Utility/Property_explorer/index.html) and pkCSM (https://biosig.lab.uq.edu.au/pkcsm/). Protein-ligand interactions were analyzed with PLIP and visualized in PyMOL.

Table 2.

Grid box size surrounding binding site of proteins

Protein names	Grid box size	Binding site
2’-5’-oligoadenylate synthase 3	Size_x = 19.500 Size_y = 20.250 Size_z = 33.000 Center_x = 9.979 Center_y = 19.415 Center_z = 6.577	ASP12,ASP82,ASP244 GLU27,GLU42,GLU186 LYS23, LYS56, LYS59, LYS198, LYS200
14-3-3 protein gamma	Size_x = 21.750 Size_y = 15.000 Size_z = 22.500 Center_x = 9.979 Center_y = 19.415 Center_z = 6.577	ASP228
Annexin A2	Size_x = 14.250 Size_y = 12.750 Size_z = 13.500 Center_x = 86.427 Center_y= −9.366 Center_z = 44.272	GLU678 ASP721, ASP726, ASP824 HIS763, HIS633 CYS801 LYS805, LYS825 TYR866, TYR648
CD59 glycoprotein	Size_x = 35.500 Size_y = 34.000 Size_z = 22.500 Center_x = 0.521 Center_y = 0.435 Center_z= −4.569	GLU439 LYS473
Cyclin-dependent kinase 6	Size_x = 16.500 Size_y = 21.750 Size_z = 21.750 Center_x = 22.198 Center_y = 38.486 Center_z= −8.657	GLU61, GLU99, GLU18 HIS100 ASP104, ASP163 TYR108 LYS43, LYS147, LYS160, LYS29
human lanosterol 14alpha-demethylase (CYP51)	Size_x = 15.000 Size_y = 14.250 Size_z = 19.500 Center_x= −4.871 Center_y= −5.476 Center_z= −6.728	LYS43, LYS156, LYS160, LYS29, LYS147 GLU61, GLU99, GLU18 ASP104, ASP231, ASP163 TYR108, TYR63, TYR92, TYR131, TYR145 HIS236, HIS314, HIS100 IS447 CYS449, CYS452 HEM601
Mixed lineage kinase domain-like protein	Size_x = 16.500 Size_y = 18.750 Size_z = 15.750	LYS43, LYS147, LYS160, LYS156, LYS29 GLU61, GLU99, GLU18, GLU452, GLU210 ASP102, ASP163, ASP104, ASP231 TYR108, TYR63, TYR92, TYR131, TYR145, TYR211, TYR307, TYR376 CYS449 HIS190, HIS212, HIS236, HIS314, HIS447, HIS100
Netrin-1	Size_x = 15.000 Size_y = 21.750 Size_z = 24.750 Center_x= −142.990 Center_y= −98.977 Center_z = 83.546	ASP104, ASP163, ASP98, ASP231 LYS29, LYS147, LYS160, LYS43, LYS100, LYS156 TYR63, TYR92, TYR307, TYR108, TYR376, TYR940, TYR131, TYR145, TYR211 CYS449, GLU61, GLU99, GLU452, GLU18, GLU210 HIS212, HIS936, HIS190, HIS236, HIS314, HIS447
Ras-related protein R-Ras2	Size_x = 18.000 Size_y = 17.250 Size_z = 15.750 Center_x= −13.458 Center_y= −13.168 Center_z= −15.128	GLU61, GLU9, GLU452, GLU18, GLU210 TYR108, TYR63, TYR92, TYR131, TYR145, TYR307, TYR376, TYR940, TYR43, TYR211 LYS147, LYS160, LYS29, LYS156, LYS160, LYS43, LYS100, LYS128, LYS27, LYS159 ASP163, ASP104, ASP231, ASP68, ASP42, ASP130, ASP98 HIS236, HIS100, HIS314, HIS447, HIS190, HIS212, HIS936 CYS449
Transferrin receptor protein 1	Size_x = 15.000 Size_y = 14.250 Size_z = 19.500 Center_x = 42.397 Center_y = 5.493 Center_z = 52.002	ASP68, ASP98, ASP130, ASP667, ASP42, ASP104, ASP163, ASP231 GLU61, GLU99, GLU18, GLU452, GLU759, CYS449, GLU210 TYR108, TYR63, TYR92, TYR131, TYR43, TYR376, TYR940, TYR211 LYS147, LYS160, LYS156, LYS43, LYS29, LYS27, LYS128, LYS159, LYS15, LYS385, LYS100 HIS236, HIS314, HIS447, HIS190, HIS212, HIS936

Molecular simulation

Molecular dynamics simulations were performed in GROMACS-2022 to study ligand-receptor interactions. Ligands were parameterized via LigParGen (https://traken.chem.yale.edu/ligpargen/) with the OPLS-AA force field, and proteins with OPLS-AA/M. Cofactors used validated OPLS-AA-compatible parameters. Complexes were solvated in TIP3P water, neutralized, energy-minimized, and equilibrated under NVT and NPT ensembles at 310 K and 1 atm. Production runs of 150 ns were performed with standard periodic boundary conditions and PME for electrostatics. Trajectories were analyzed for RMSD, RMSF, radius of gyration, hydrogen bonds, and radial distribution functions.MM-PBSA calculations using g_mmpbsa were performed to estimate protein-ligand binding free energies (ΔGbind). This approach decomposes ΔGbind into molecular mechanics energies (van der Waals and electrostatic) and solvation contributions (polar and nonpolar), providing a detailed assessment of the energetic factors driving receptor-ligand interactions.

In GROMACS molecular dynamics simulations, RMSD, RMSF, and radius of gyration (Rg) are key metrics for evaluating protein stability, flexibility, and compactness. RMSD tracks overall structural deviation from a reference, RMSF assesses local residue flexibility, and Rg monitors protein folding and structural compactness. These parameters help interpret conformational changes and the impact of ligand binding on protein stability.

Results

Identification of Dracorubin as a broad-spectrum natural inhibitor

In the initial phase of this study, a structure-guided virtual screening campaign was conducted on a curated library of 2,299 plant-derived metabolites obtained from PubChem. The goal was to identify natural compounds with the capacity to modulate multiple signaling and survival pathways associated with bronchogenic non-small cell lung carcinoma (NSCLC). Ten proteins representing diverse facets of NSCLC pathology were selected as targets: OAS3, YWHAG (14-3-3γ), Annexin A2, CD59, CDK6, CYP51A1, MLKL, Netrin-1, RRAS2, and TFRC. These proteins collectively cover mechanisms ranging from dysregulated kinase signaling to immune evasion, metabolic rewiring, ferroptosis, and necroptotic pathways. Among the screened phytochemicals, Dracorubin (PubChem CID: 160270; C₃₂H₂₄O₅; MW 488.54 Da), an anthraquinone derivative, emerged conspicuously as the only compound capable of achieving favorable binding across all ten targets. Docking scores spanned − 47.9 kJ/mol to − 18.9 kJ/mol (Table 3), with the majority clustering below − 29.4 kJ/mol, indicating a consistent tendency of the compound to adopt energetically preferred poses across structurally distinct proteins. This rare polypharmacological footprint led to a comprehensive interrogation of Dracorubin using extensive molecular dynamics (MD) simulations and binding free-energy analyses to assess whether the favorable docking outcomes translate into physically realistic and stable interactions.

Table 3.

Docking results and binding coordinates of protein names with Dracorubin

Protein names	Dracorubin|C32H24O5|CID 160,270	Grid box size
2’-5’-oligoadenylate synthase 3	−32.3	Size_x = 19.500, size_y = 20.250, size_z = 33.000
14-3-3 protein gamma	−26.9	Size_x = 21.750, size_y = 15.000, size_z = 22.500
Annexin A2	−18.9	Size_x = 14.250, size_y = 12.750, size_z = 13.500
CD59 glycoprotein	−33.2	Size_x = 35.500, size_y = 34.000, size_z = 22.500
Cyclin-dependent kinase 6	−38.2	Size_x = 16.500, size_y = 21.750, size_z = 21.750
Human lanosterol 14alpha-demethylase (CYP51)	−47.9	Size_x = 15.000, size_y = 14.250, size_z = 19.500
Mixed lineage kinase domain-like protein	−37.4	Size_x = 16.500, size_y = 18.750, size_z = 15.750
Netrin-1	−38.2	Size_x = 15.000, size_y = 21.750, size_z = 24.750
Ras-related protein R-Ras2	−37.0	Size_x = 18.000, size_y = 17.250, size_z = 15.750
Transferrin receptor protein 1	−36.1	Size_x = 15.000, size_y = 14.250, size_z = 19.500

Molecular dynamics simulations demonstrate stable complex formation

To evaluate conformational stability and dynamic behavior, each Dracorubin–protein complex underwent a 150-ns all-atom MD simulation in explicit solvent. Across all systems, the backbone Cα root-mean-square deviation (RMSD) profiles displayed an early stabilization phase within the first 20–40 ns (Supplementary Figures 1–10). Average RMSD values for the equilibrated trajectories remained below 0.30 nm, suggesting that ligand binding did not induce undue structural distortion. Particularly low RMSD readings were observed for the complexes formed with RRAS2 (~ 0.15 ± 0.02 nm) (Supplementary Figure 9 A), CD59 (~ 0.2 ± 0.01 nm) (Supplementary Figure 4 A), CYP51 (~ 0.24 ± 0.02 nm) (Supplementary Figure 6 A) and CDK6 (~ 0.25 ± 0.03 nm) (Supplementary Figure 5 A), reflecting exceptional rigidity in these interactions. Residue-level fluctuations (RMSF) further reinforced this observation (Supplementary Figures.1B-10B). Binding-site residues in proximity to Dracorubin exhibited restrained motions, typically < 0.10 nm, while larger fluctuations were limited to flexible loops or terminal regions—areas naturally prone to mobility regardless of ligand presence. The radius of gyration (Rg) remained unaltered across simulations, indicating preservation of global protein compactness (Supplementary Figures.1–10 C). A critical indicator of true binding stability-the heavy-atom RMSD of the ligand-also showed highly favorable behavior (Supplementary Figures.1D-10D). In all systems, Dracorubin rapidly adopted a stable conformation and remained confined within the binding pocket, consistently maintaining RMSD values < 0.10 nm without any dissociation or notable drift during the entire 150-ns trajectory (Supplementary Figure 1D-10D). Such uniform stability across ten unrelated proteins strongly suggests that the interactions are not artifacts of docking but reflect genuine energetic complementarity.

Binding free-energy profile and driving forces

To quantify the thermodynamic favorability of binding, MM/GBSA calculations were performed using snapshots extracted from the final 50 ns of each trajectory. All ten protein complexes exhibited significantly negative binding free energies, confirming that Dracorubin engages each target in a spontaneously favorable manner. The strongest interaction was recorded for CYP51 (ΔGbind − 157.67 ± 7.05 kJ/mol), followed by MLKL, CDK6, and TFRC (Table 4). This ranking correlates well with the RMSD and pocket-depth observations from the simulations. Decomposition of energetic contributions provided insight into the underlying forces. van der Waals interactions were consistently the largest favorable component, ranging from − 252.06 kJ/mol in CYP51 to − 84.32 kJ/mol in CD59. These values underscore the strong hydrophobic complementarity between the rigid, planar anthraquinone framework of Dracorubin and the predominantly apolar microenvironments of the binding sites. Electrostatic contributions varied among targets and were particularly prominent in CDK6 and CYP51, where the ligand forms well-positioned polar contacts in structurally conserved regions such as kinase hinge residues or the narrow heme-proximal channel of CYP51. Although polar solvation penalties naturally oppose binding, they were more than compensated by favorable gas-phase electrostatics and non-polar solvation terms, ultimately producing consistently negative net free energies.

Table 4.

Free binding energy and the individual energy contributions result of proteins complexes (kJ/mol)

	2′-5′-oligoadenylate synthase 3	14-3-3 GAMMA	Annexin A2	CD59 glycoprotein	Cyclin-dependent kinase 6	CYP51	Mixed lineage kinase domain-like protein	Netrin-1	Ras-related protein R-Ras2	Transferrin receptor protein 1
ΔEele	−13.1 ± 3.1	−3.03 ± 4.75	−49.03 ± 1.93	−23.76 ± 4.42	−115.46 ± 3.91	−53.81 ± 1.62	−25.26 ± 5.33	−17.35 ± 1.58	−25.7 ± 2.02	−8.85 ± 1.86
ΔEvdW	−119.44 ± 10.58	−93.37± 3.89	−122.23 ± 6.12	−84.32 ± 16.01	−161.25 ± 7.04	−252.06 ± 5.37	−181.83 ± 5.81	−107.49 ± 1.81	−112.82 ± 5.28	−130.97 ± 5.1
ΔGGB	109.65 ± 14.83	33.23 ± 4.62	114.94 ± 2.89	49.47 ± 4.63	201.26 ± 5.55	174.18 ± 4.15	89.95 ± 6.64	63.85 ± 4.79	66.26 ± 3.11	65.79 ± 3.06
ΔGSA	−13.41 ± 0.95	−11.5 ± 0.27	−16.17 ± 0.46	−11.31 ± 1.60	−19.68 ± 0.23	−25.98 ± 0.19	−20.7 ± 0.62	−12.51 ± 0.8	−13.34 ± 0.47	−14.62 ± 0.2
ΔEnon-polara	−132.85±8.92	−104.87 ± 3.89	−138.4 ± 4.85	−95.63 ± 16.01	−180.93 ± 0.21	−278.04 ± 5.37	−202.53 ± 5.81	−120 ± 1.81	−126.156 ± 5.28	−145.59 ± 5.1
ΔEpolarb	96.55±3.1	30.2 ± 4.62	65.91 ± 2.89	25.7 ± 4.42	85.8 ± 3.91	120.37 ± 4.15	64.69 ± 5.33	46.5 ± 4.79	40.6 ± 3.11	56.92 ± 3.06
ΔGbind	−36.3 ± 5.88	−74.67 ± 3.45	−72.49 ± 4.99	−69.93 ± 18.49	−95.13 ± 6.59	−157.67 ± 7.05	−137.84 ± 5.37	−73.5 ± 3.28	−85.55 ± 5.28	−88.67 ± 5.

Molecular recognition patterns across targets

A residue-level examination of the equilibrated structures (Fig. 2A–J) demonstrates that Dracorubin interacts predominantly via hydrophobic contacts in all ten proteins, forming a total of seventy-two such interactions (7.2 ± 1.4 per complex). These contacts span aliphatic side chains (Ile, Val, Leu, Ala, Met), aromatic rings (Phe, Tyr) cumulatively sequestering more than 85% of the ligand’s solvent-accessible surface. The most extensive hydrophobic encapsulation was observed in CYP51, where residues L134, T135, Y145, F234, H236, H314, T315, and I379 form a densely packed apolar corridor that envelopes the planar anthraquinone chromophore. Even in more solvent-accessible systems such as CDK6 and OAS3, the ligand retained 3 to 4 well-defined nonpolar interactions, highlighting a conserved interaction motif across structurally unrelated proteins.

Fig. 2.

Dracorubin complex with (A) 2’−5’-oligoadenylate synthase 3, (B) 14-3−3 GAMMA, (C) Annexin A2, (D) CD59 glycoprotein, (E) Cyclin-dependent kinase 6, (F) CYP51, (G) Mixed lineage kinase domain-like protein, (H) Netrin-1, (I) Ras-related protein R-Ras2, (J) Transferrin receptor protein 1. The highlighted residues are marked with light blue sticks. Blue lines and red dashed lines indicate hydrogen bonds and hydrophobic interactions respectively

In contrast, hydrogen bonds were remarkably sparse and restricted to only three complexes:

OAS3: a single persistent hydrogen bond between R836 and a quinone carbonyl (Fig. 2A).

CDK6: two interactions formed by H100 and K147, each donating to separate carbonyl groups (Fig. 2E).

RRAS2: hydrogen bonds donated by Y105 and Q110 (Fig. 2I).

Seven remaining targets-14-3-3γ, ANXA2, CD59, CYP51A1, MLKL, Netrin-1, and TFRC—showed no stable hydrogen bonds over the production phase. On average, this yields fewer than one hydrogen bond per complex compared to more than seven hydrophobic interactions, demonstrating an overwhelming reliance on nonpolar forces.

The ligand’s physicochemical properties—planar, conjugated, hydrophobic (logP ~ 5.7), and with minimal hydrogen-donor capacity—predispose it to exploit nonpolar clefts. Hydrogen bonds, when present, function mainly as steric anchors rather than primary determinants of affinity. These structural features explain the consistent dominance of ΔE_vdW in the MM/GBSA profiles and the pronounced desolvation-driven entropic gains accompanying ligand binding.

In silico pharmacokinetic and toxicological characteristics

The predictive ADME-tox assessment indicated a profile compatible with early-stage drug discovery. Dracorubin demonstrated excellent intestinal absorption (> 98%) and high Caco-2 permeability, suggesting favorable oral bioavailability. The compound is predicted to inhibit major P-glycoprotein isoforms, potentially reducing efflux and enhancing intracellular retention—an advantageous feature for anticancer agents that must accumulate within tumor cells (Table 5).

Table 5.

ADME analysis and toxicity profiling of Dracorubin obtained from OSIRIS property explorer and PkCSM

	Dracorubin (160270)
Water solubility	−5.7
TPSA	53.99
H bond acceptor	5
H bond donor	0
Nb rotatable bonds	3
Molecular weight	488.539
Caco-2 permeability	0.99
P-glycoprotein I inhibitor	Yes
P-glycoprotein II inhibitor	Yes
Intestinal absorption (human)	98.9
BBB permeability	0.01
CNS permeability	−0.97
Total Clearance	0.60
Renal OCT2 substrate	No
AMES toxicity	No
Oral Rat Acute Toxicity (LD50)	2.38
Oral Rat Chronic Toxicity (LOAEL)	−0.61
Skin Sensitisation	No

Importantly, Dracorubin displayed no predicted AMES mutagenicity, lacked skin-sensitization potential, and exhibited acceptable acute and chronic toxicity parameters. Its low predicted CNS penetration may reduce neurological side effects, which is desirable for oncology drug candidates not intended to target the brain (Table 5).

Toxicity profiling is favorable: AMES-negative, non-sensitizing, with oral LD50 2.38 mol/kg and negative LOAEL (− 0.61), indicating low acute and chronic risk. In summary, Dracorubin satisfies quantitative drug-likeness metrics, displays high oral bioavailability potential, and presents a safe peripheral profile, positioning it as a promising lead for non-CNS indications.

Discussion

This study provides the first comprehensive evidence that the natural compound Dracorubin can simultaneously target ten proteins implicated in NSCLC, spanning multiple oncogenic mechanisms. The selected targets include regulators of cell-cycle progression (CDK6), sterol metabolism (CYP51), necroptotic signaling (MLKL), immune evasion (CD59), and iron homeostasis (TFRC). The engagement of these diverse pathways positions Dracorubin as a multi-pathway modulator, potentially offering therapeutic benefits beyond conventional single-target inhibitors. Iron metabolism, in particular, represents a critical vulnerability in lung adenocarcinoma, as COPZ1 has been shown to regulate NCOA4-mediated ferritinophagy, thereby linking iron homeostasis to tumor proliferation and survival [32]. Targeting such pathways with a single bioactive compound reinforces the biological relevance and translational potential of this approach. Molecular dynamics simulations revealed remarkable stability across all protein-ligand complexes, with protein backbone RMSDs below 0.30 nm and ligand RMSDs below 0.10 nm. These findings indicate that Dracorubin maintains structurally coherent interactions under physiological conditions. MM/GBSA analyses further elucidated the molecular determinants of binding, highlighting deep hydrophobic burial complemented by strategically positioned hydrogen bonds. This combination underpins both affinity and specificity, suggesting that Dracorubin can exploit conserved hydrophobic architectures across structurally diverse protein folds. The ability to engage multiple targets with coherent energetic and structural features supports its designation as a privileged natural scaffold with favorable pharmacokinetic attributes. The multi-target nature of Dracorubin aligns with emerging network-based strategies in biomarker-driven therapy. Utilizing approaches such as the NetAUC framework, multi-biomarker optimization can maximize predictive accuracy and therapeutic relevance, guiding compound prioritization across complex disease networks [33]. Additionally, Dracorubin’s engagement with signaling pathways intersects with current mechanistic insights into tumor biology. For instance, TMEM64-mediated activation of the Wnt/β-catenin cascade demonstrates how aberrant signaling contributes to malignant progression, paralleling the pathways modulated by Dracorubin [34]. Moreover, chronic inflammatory signals, such as smoke-induced macrophage polarization through exosomal lncRNA MEG3 → TREM-1, underscore the importance of integrating immune and inflammatory axes into therapeutic considerations [35]. Comparative evaluation of therapeutic modalities indicates that small molecules like Dracorubin offer advantages in terms of tissue penetration, pharmacokinetics, and ease of chemical modification, complementing the safety and specificity profiles of biologics such as monoclonal antibodies [36, 37]. In silico screening strategies employed here are consistent with current ensemble and machine-learning approaches for anticancer drug-response prediction, providing an efficient and robust framework for prioritizing multi-target candidates [38]. Given the compelling computational evidence, experimental validation is warranted. Priority assays should include binding confirmation via surface plasmon resonance or microscale thermophoresis, cellular engagement profiling such as CETSA, kinase inhibition assays for CDK6, sterol pathway interference tests for CYP51, and functional studies in NSCLC models. Collectively, these findings support the positioning of Dracorubin as a promising multi-target therapeutic candidate capable of mitigating pathway redundancy and drug resistance in lung cancer, and exemplify how natural compounds can serve as versatile scaffolds in precision oncology.

Conclusion

In recent years, drug design has significantly advanced through drug repurposing and repositioning approaches, leading to the rapid and cost-effective development of new inhibitors for various diseases. Complex lung cancers, particularly those detected at advanced stages, have attracted considerable attention. Researchers are increasingly focusing on designing and developing novel therapeutic agents to detect and target lung cancer in its early stages.

Although numerous drugs with therapeutic potential for this type of cancer have been identified, a significant gap remains in the literature regarding the discovery of compounds capable of simultaneously targeting multiple proteins involved in disease pathogenesis. In this study, a multi-target strategy was adopted to evaluate the inhibitory potential of dracorubin. Research has shown that dracorubin exhibits potential inhibitory effects against all ten targets. These compounds demonstrated higher binding affinity and stability in their interactions compared to other reference ligands.

The results of this study, along with future experimental investigations of the identified prioritized a computational lead requiring experimental validation, could pave the way for the development of a novel therapeutic approach to managing bronchogenic lung cancer. These compounds can be further evaluated in in vitro and in vivo studies. Additionally, evaluating their effects in subsequent clinical trials could contribute to the discovery and development of effective therapeutic agents. This study not only highlighted the potential for designing an effective drug targeting multiple stages of the disease, but also emphasized the importance of a stage-specific therapeutic strategy for bronchogenic lung cancer.

Limitations

This study faced several limitations. The proposed drug was selected as a candidate to cover all four stages of the disease and was chosen from plant-derived compounds. Additionally, the study focused on inhibiting genes with high expression levels. Furthermore, clinical trials will be required to fully evaluate the efficacy and safety of the therapeutic, preventive, or diagnostic interventions in humans.

This study outlines the limitations of a computational drug design study, noting that the purely in silico approach requires experimental validation in vitro and in vivo to confirm the predicted biological activity. The targeting protein-protein interfaces carry a risk of false positives due to the challenging nature of these sites. Furthermore, the study does not assess critical pharmacokinetic properties like permeability and metabolic stability, which are essential for determining if a compound can reach its target in a living system. Also, we acknowledge that while the computational study offers valuable insights into potential phytochemical inhibitors for lung cancer targets, its limitations must be considered. These include inherent in silico biases from scoring functions, simplified models, and static protein structures that may not reflect biological reality.

Author contributions

Conceptualization: KK, AK, and MA Methodology: MA, BK, and EBInvestigation: KKVisualization: BK, EBProject administration and Supervision: AK, MA, and KK Writing – original draft: BK, EB, and MAWriting – review & editing: MA, AK, EB, and KK.

Funding

Not applicable.

Data availability

The datasets analyzed during the current study are available in RCSB Protein Data Bank (RCSB PDB) and open chemistry database at the National Institutes of Health database, accessible at https://pubchem.ncbi.nlm.nih.gov/ and https://www.rcsb.org/.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.