Research report: Identification of potential breast cancer therapeutics from non-phenolic diarylheptanoids of Curcuma xanthorrhiza using network pharmacology and in silico methods

The work, led by Roney Miah and colleagues, is entirely predictive: no compounds were extracted from the plant and no laboratory experiments were performed, but the multi-layered in silico pipeline points to a specific molecular target worth experimental follow-up.

Breast cancer remains the most commonly diagnosed cancer in women worldwide. The study’s background cites International Agency for Research on Cancer figures of roughly 2.3 million new diagnoses and more than 670,000 deaths from breast cancer in 2022, when the disease had the highest global incidence of all cancers and ranked second in age-standardized cancer mortality. Although modern therapy has shifted toward precision approaches, including antibody-drug conjugates such as trastuzumab deruxtecan and the PI3K-alpha inhibitor alpelisib, many treatments still carry severe side effects and drug resistance continues to erode their long-term value, motivating the search for new lead molecules.

The team focused on three NPDs previously identified in C. xanthorrhiza and known for anti-inflammatory activity. First, they assessed drug-likeness and safety computationally: pkCSM predictions indicated the compounds comply with Lipinski’s rule of five and have favorable pharmacokinetic profiles, while the ProTox-II server suggested no discernible toxicity.

To find what the compounds might actually do in breast cancer, the researchers assembled target lists from public databases. SwissTargetPrediction and SuperPred yielded 564 predicted NPD targets, which collapsed to 301 unique proteins after de-duplication, while GeneCards supplied 460 unique breast cancer-related targets. Their intersection contained just 24 proteins, which the Panther classification system sorted into functional groups including transmembrane signal receptors (32 percent), metabolite interconversion enzymes (16 percent), transporters (8 percent), gene-specific transcriptional regulators (12 percent), and chromatin-associated proteins (4 percent).

A protein-protein interaction network built in the STRING database, using a high confidence threshold above 0.775, connected all 24 targets through 81 edges with an average node degree of 6.75. Degree-based ranking with the CytoHubba plugin in Cytoscape produced a top-ten hub list headed by ERBB2 (the gene encoding HER2), EGFR, and PIK3CA, followed by PPARG, AR, MDM2, EP300, ESR2, CYP19A1, and NTRK1. Enrichment analysis in DAVID generated 201 Gene Ontology terms, spanning 126 biological processes, 23 cellular components, and 52 molecular functions, and 71 KEGG signaling pathways.

The docking stage centered on the ERBB2 crystal structure deposited in the Protein Data Bank under accession 7PCD. Binding pockets were first predicted with PrankWeb, a machine-learning tool that identifies functionally relevant regions on a protein surface, and a docking grid was then generated around the predicted site. Using PyRx, all three NPDs docked into the pocket with scores between −8.2 and −8.5 kcal/mol. Compound 1 was the strongest binder at −8.5 kcal/mol, forming carbon-hydrogen bonds with LEU726, LEU796, and LEU852 and alkyl contacts with ALA751, LYS753, and ALA771. Compound 2, at −8.2 kcal/mol, made two hydrogen bonds with ALA751 and LEU796, alkyl interactions with VAL734, ALA751, and LYS753, and pi-sigma contacts with LEU726, LEU785, and LEU852. Compound 3 scored −8.3 kcal/mol, hydrogen-bonding with SER783 and THR862 while stacking hydrophobic alkyl and pi-alkyl contacts across LEU796, ALA771, LYS753, LEU726, ALA751, and LEU852. Neratinib, the approved ERBB2 inhibitor used as a reference, bound more tightly at −9.6 kcal/mol through alkyl contacts with LEU726, LEU852, ILE767, VAL734, and LYS753 plus two carbon-hydrogen bonds with ASP863 and ILE767.

To test whether the docked poses would hold up under dynamic conditions, the team ran 100-nanosecond molecular dynamics simulations in GROMACS 2023.2 using the CHARMM27 force field, a TIP3P water model, and a 2-femtosecond timestep, after energy minimization and separate NVT and NPT equilibration phases. The Cα atoms of ERBB2 fluctuated between 0.12 and 0.27 nanometers in root mean square deviation over the trajectories. The ligand-fit RMSD ranges were 0.12 to 0.95 nanometers for compound 1, 0.11 to 0.40 nanometers for compound 2, 0.10 to 0.40 nanometers for compound 3, and 0.13 to 0.52 nanometers for neratinib, indicating that the complexes with compounds 2 and 3 were slightly more stable than the control complex. Residue-level fluctuations measured by RMSF spanned 0.05 to 0.48 nanometers for the apo protein and up to 0.54, 0.30, 0.30, and 0.44 nanometers for the compound 1, 2, 3, and neratinib complexes respectively, with the tighter ranges for compounds 2 and 3 suggesting that ligand binding dampens flexibility in parts of the protein. Compactness, tracked by the radius of gyration, stayed below 2.01 nanometers in every system: 1.98 nanometers for ERBB2 alone and 1.93, 1.92, 1.94, and 2.01 nanometers for the four ligand complexes. Solvent-accessible surface area of the Cα atoms was about 149 square angstroms for the apo protein and dropped to 138, 134, 138, and 147 square angstroms upon binding of compounds 1, 2, 3, and neratinib, with the lower values for compounds 2 and 3 again pointing to a more compact folded state. Hydrogen-bond counts between ERBB2 and each NPD oscillated between two and three, with two bonds predominant for most of the simulation, whereas neratinib formed up to six, two of which persisted throughout.

Binding free energies calculated with the MM/GBSA method over the simulation trajectories reinforced this hierarchy. The total binding energies were −11.42, −29.76, −25.27, and −40.43 kcal/mol for compounds 1, 2, 3, and neratinib respectively, all favorable and thermodynamically stable. Van der Waals forces dominated, contributing −24.7, −37.11, −33.49, and −58.56 kcal/mol, while electrostatic terms added −7.28, −11.79, −12.44, and −32.46 kcal/mol. Polar solvation worked against binding, with unfavorable contributions of 24.52, 24.71, 26.19, and 58.87 kcal/mol, but nonpolar solvation energies of −3.97, −5.57, −5.54, and −8.28 kcal/mol partially compensated. Per-residue decomposition identified Leu726, Val734, Thr862, and Asp863 as the amino acids contributing most to stabilization across all four complexes, matching the residues contacted in the docking poses and suggesting a consistent binding mode.

The choice of ERBB2 as the lead target rests on well-established biology. ERBB2 overexpression occurs in approximately 15 to 20 percent of breast cancer cases and is frequently associated with aggressive tumor behavior and poor prognosis. The HER2-positive subtype, roughly 20 percent of cases, is driven by amplification and overexpression of the ERBB2 gene, which fuels uncontrolled proliferation through the PI3K/AKT/mTOR cascade. Clinically, HER2 status is determined by immunohistochemistry or in situ hybridization, with an IHC score of 3+ generally confirming positivity and equivocal 2+ scores requiring further molecular testing. The recent recognition of an ERBB2-low category has expanded treatment to patients once considered ineligible for HER2-directed drugs, and trastuzumab deruxtecan has shown efficacy in that subgroup. Crosstalk between ERBB2 and hormone receptors can also foster resistance to both endocrine and HER2-targeted therapies, which is one reason multi-target agents remain attractive.

The discussion situates the findings within earlier experimental work on the same plant. Xanthorrhizol, a major constituent of Curcuma xanthorrhiza, has been reported to induce apoptosis in MCF-7 breast cancer cells by activating p53, promoting PARP-1 cleavage, and downregulating Bcl-2, to suppress proliferation and metastasis in triple-negative breast cancer models through inhibition of NF-κB and matrix metalloproteinases MMP-2 and MMP-9, and to block VEGF-induced angiogenesis via PI3K/AKT/eNOS signaling. The authors note, however, that these studies examined isolated endpoints rather than pathway-level networks, that few compare effects across ER-positive, HER2-positive, and triple-negative subtypes despite differing pathway dependencies such as IGF-1R expression, and that mechanistic work on epigenetic regulation, including histone H2AXY142 and H3Y41 phosphorylation, remains scarce.

The authors are explicit about the limits of the work. The compounds were neither extracted nor chemically identified from the plant, and no biological assays of any kind were performed; every result derives from database mining and computational modeling and therefore remains predictive. Future directions they outline include subtype-specific in vitro and in vivo studies, longitudinal models to assess durable benefits such as reduced recurrence, biomarker-driven patient selection, combinations with receptor tyrosine kinase inhibitors, PI3K/AKT blockers, or endocrine therapy, and formulation advances such as nanoparticle delivery and prodrug strategies to improve bioavailability. Within those constraints, the study offers a ranked, testable hypothesis: compounds 2 and 3 of the non-phenolic diarylheptanoids deserve direct biochemical evaluation as ERBB2 ligands.

Subject of Research: Computational prediction of Curcuma xanthorrhiza diarylheptanoids as potential HER2-targeted breast cancer therapeutics

Article Title: Identification of potential breast cancer therapeutics from non-phenolic diarylheptanoids of Curcuma xanthorrhiza using network pharmacology and in silico methods

Article References: Miah, R., Uddin, M. N., Miah, M. M., Hossain, A., & Aluwi, M. F. F. M. (2026). Identification of potential breast cancer therapeutics from non-phenolic diarylheptanoids of Curcuma xanthorrhiza using network pharmacology and in silico methods. Future Journal of Pharmaceutical Sciences, 12(1), Article 93. https://doi.org/10.1186/s43094-026-01025-3

Image Credits: AI Generated

DOI: 10.1186/s43094-026-01025-3

Keywords: breast cancer, HER2, ERBB2, Curcuma xanthorrhiza, network pharmacology, molecular docking, molecular dynamics, MM/GBSA, natural products, drug discovery, in silico, temulawak