Network Pharmacology and Preliminary In-Vitro Evaluation of Piper Longum (Indian Long Pepper) in Breast Cancer

Background

Breast cancer continues to be a significant challenge in global health, as its complex development and varied clinical presentations pose substantial difficulties for effective treatment and prevention.^[1] With the worldwide occurrence of this disease on the rise, it is crucial to explore the intricate nature of breast cancer to create successful therapeutic approaches.^[2]

Network pharmacology is widely used to prioritize compound-target relationships in complex mixtures, although its predictions require experimental validation. Unlike conventional single-target approaches, it enables the identification of multi-target mechanisms underlying the therapeutic effects of complex herbal medicines. In the present study, network pharmacology was employed to predict the potential molecular targets and biological pathways through which Piper longum phytochemicals may exert anti-breast cancer activity.

Piper longum (Indian long pepper) is a medicinal plant of the family Piperaceae that has long been used in traditional medicine and has attracted increasing attention for its anticancer potential.^[3] Experimental studies have shown that P. longum extracts and several of its bioactive constituents, including piperine and piperlongumine, inhibit breast cancer cell proliferation through mechanisms involving apoptosis, cell-cycle arrest, oxidative stress, and modulation of multiple oncogenic signaling pathways.^[4,5] However, the comprehensive molecular targets and interconnected signaling networks through which the diverse phytochemicals of P. longum exert their anti-breast cancer effects remain poorly understood. Therefore, this study used an integrated network pharmacology, molecular docking, and experimental approach to systematically identify the anti-breast cancer activity of P. longum.

Objectives

This study used network analysis to prioritize P. longum phytochemicals and breast-cancer-associated targets for further testing. The process involved several key steps, including gathering and organizing data on the chemical compounds present in P. longum, constructing a network using analytical methods, identifying key nodes through network analysis, computationally predicting results using molecular docking, and interpreting the findings to predict potential drug targets and therapeutic uses. In addition, the effect of P. longum aqueous extract on cancer cell proliferation was also assessed in the MCF-7 breast cancer cell line.

Aim

The study aimed to investigate different compounds found in P. longum as potential phytochemicals against breast cancer and to explore the effect of the aqueous extract of P. longum fruits on breast cancer cell proliferation.

P. longum Bioactive Compounds Identification

Bioactive phytochemical constituents reported from the fruits of Piper longum were compiled using two independent sources: (i) the Indian Medicinal Plants, Phytochemistry and Therapeutics (IMPPAT) database (https://cb.imsc.res.in/imppat/; accessed on 25 April 2025), and (ii) a systematic literature survey of published phytochemical studies.^[6] The IMPPAT database was searched using the keyword “Piper longum” under the medicinal plant search option. All phytochemicals specifically annotated for the fruit of P. longum were retrieved. Literature searches were performed using combinations of the keywords “Piper longum,” “fruit,” “phytochemical,” “constituents,” “secondary metabolites,” and “chemical composition.”

The lists obtained from both sources were merged, and duplicate compounds were removed using compound names and PubChem Compound Identifiers (CID). Only compounds with a defined chemical structure and an available PubChem entry were retained for further analysis. Canonical SMILES and molecular structures were downloaded from PubChem and used as input for subsequent analyses (https://pubchem.ncbi.nlm.nih.gov/).^[7] All databases were accessed during 2024.

Screening of Potential Bioactive Compounds Using Absorption, Distribution, Metabolism, and Excretion (ADME) Analysis

The collected phytochemical compounds were subjected to pharmacokinetic evaluation using the SwissADME web server (http://www.swissadme.ch/) to assess their drug-likeness and suitability for oral administration.^[8] The canonical SMILES notation for each compound, obtained from the PubChem database, was uploaded to the SwissADME server for analysis. Several physicochemical and pharmacokinetic parameters relevant to drug discovery were evaluated, including molecular weight, hydrogen bond donor and acceptor counts, topological polar surface area (TPSA), XLOGP3 (octanol/water partition coefficient), predicted gastrointestinal (GI) absorption, Lipinski’s Rule of Five compliance, and the predicted bioavailability score.^[4,9] These parameters were selected to assess the oral drug-likeness, pharmacokinetic behavior, and medicinal chemistry suitability of each phytochemical.

Compounds were screened using predefined inclusion criteria to select candidates with favorable pharmacokinetic properties. Only compounds with complete physicochemical information in SwissADME, a canonical chemical structure available in the PubChem database, no violations of Lipinski’s Rule of Five, a predicted bioavailability score ≥ 0.30, and acceptable GI absorption were considered eligible for further analysis. The Lipinski rule is based on specific physicochemical properties, such as absorption, distribution, metabolism, and excretion, to predict drug pharmacokinetic properties such as (1) a maximum of five hydrogen bond donors, (2) a maximum of 10 hydrogen bond acceptors, (3) a molecular mass less than 500 Da, and (4) the partition coefficient, or XLOGP3, must not exceed 5 (Supplementary Table 2). The partition coefficient quantifies the relationship between a compound’s lipophilicity (fat solubility) and hydrophilicity (water solubility). A compound with XLOGP3 > 5 is considered excessively lipophilic. Such compounds often suffer from several pharmacokinetic problems, including poor aqueous solubility in fluids such as GI fluids, reduced intestinal absorption, excessive tissue accumulation, and more. Conversely, compounds were excluded if they lacked complete structural or physicochemical information, were unavailable in the PubChem database, violated one or more of Lipinski’s criteria, exhibited poor predicted GI absorption, or had a predicted bioavailability score below 0.30 because in early drug discovery, compounds with a bioavailability score of ≥0.3 are generally considered acceptable, whereas values >0.5 are preferred for improved therapeutic efficacy and dose optimization.^[10] After applying these screening criteria, 19 phytochemical compounds were identified as potential bioactive constituents of Piper longum and were subsequently selected for downstream target prediction and network pharmacology analyses. Supplementary Figure 1 shows a brief flow diagram of the screening pipeline.

P. longum Potential Targets Prediction

The potential human protein targets of the 19 phytochemicals that passed the predefined ADME and drug-likeness criteria were predicted using three complementary in silico platforms: SuperPred (https://prediction.charite.de/; accessed on 10^th May 2025), SwissTargetPrediction (http://www.swisstargetprediction.ch/; accessed on 10^th May 2025), and DIGEP-Pred (http://www.way2drug.com/ge/; accessed on 10^th May 2025).^[11–13] Canonical SMILES structures obtained from PubChem were used as input for target prediction. Targets meeting the predefined confidence criteria were retained for subsequent analysis. Specifically, predicted targets with a probability score >0.80 in SuperPred, >0.40 in SwissTargetPrediction, and >0.50 in DIGEP-Pred were selected.^[14] The search was restricted to Homo sapiens targets. Predicted target names, UniProt identifiers, gene symbols, and probability scores were retrieved. Duplicate targets predicted for multiple compounds were merged before downstream analyses. The resulting target lists from the three platforms were integrated to generate a comprehensive set of candidate targets associated with the selected P. longum phytochemicals. These targets represent computational predictions and database-derived associations and were not considered experimentally validated molecular targets.

Breast Cancer Screening Targets

Breast cancer-associated genes were retrieved from the Online Mendelian Inheritance in Man (OMIM) (https://www.omim.org/; accessed on 12^th May 2025), Therapeutic Target Database (TTD) (https://db.idrblab.net/ttd/; accessed on 12^th May 2025), and GeneCards (https://www.genecards.org/; accessed on 13^th May 2025) (with a relevance score exceeding 30) using the keyword “Breast Cancer.”^[15–17] The predefined database-specific criteria for human protein-coding genes were applied. Database-specific inclusion criteria were applied, gene symbols were standardized, and duplicate entries were removed. The resulting disease-associated gene set was then compared with the predicted phytochemical target set to identify shared candidate targets.

Shared Target Identification

The predicted P. longum phytochemical targets and breast cancer-associated genes were compared using Venny version 2.1.0 (https://bioinfogp.cnb.csic.es/tools/venny/; on 15^th May 2025). The overlapping genes were considered shared candidate targets linking the selected phytochemicals with breast cancer. They were subsequently subjected to PPI-network construction, enrichment analysis, compound–target–disease network analysis, and molecular docking.

Construction and Analysis of Protein–Protein Interaction (PPI) Network

Protein-protein interactions (PPIs) play a crucial role in regulating cellular activities and biological functions across different organisms. Gaining insights into protein interactions can enhance our understanding of infection mechanisms and inform the development of effective drugs and treatment strategies. A PPI network was constructed using the STRING database (https://string-db.org/; 16th May 2025), focusing on Homo sapiens genes with high-confidence interaction scores of ≥0.7, with disconnected nodes hidden.^[18] Next, network data from the STRING database were exported as a tab-separated values file to build the PPI network diagram and perform enrichment analysis using Cytoscape v3.9.1.^[19] Network topology was analyzed using the CytoNCA plugin by calculating Degree Centrality (DC), Betweenness Centrality (BC), and Closeness Centrality (CC) for each node. Degree centrality was used to identify highly connected proteins, while betweenness and closeness centrality were used to evaluate each node’s relative influence and connectivity within the network.

Construction and Analysis of Compound, Target, and Disease (C-T-D) Network

A C-T-D network was developed using Cytoscape v3.9.1 (https://cytoscape.org/) to visualize interactions among the selected phytochemicals, predicted target genes, and breast cancer. We extracted nodes representing core genes, drugs, and diseases. By analyzing the nodes, we identified several targets based on their connectivity degree, which reflects their importance. We used the CytoNCA plugin to perform network topology analysis and loaded the C-T-D network map into Cytoscape version 3.9.1. Nodes were ranked by degree centrality. Degree was used as a network prioritization measure, not to test statistical significance.^[20] Nodes with higher degree values were considered more influential within the network, and the top-ranked phytochemicals were identified based on their connectivity to predicted target genes.

Gene Ontology (GO) Analysis for Enrichment and Pathway Analysis

ShinyGO 0.7 (http://bioinformatics.sdstate.edu/go/; accessed on 20^th May 2025) was utilized for the analysis of pathway and process enrichment.^[21] This tool encompasses a standard collection of ontologies for enrichment analysis, including Gene Ontology (GO) processes, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, Reactome gene sets, canonical pathways, and a thorough resource for mammalian protein complexes. GO and KEGG enrichment analyses were performed on the 23 top-ranked genes selected from the PPI network.^[22]

Analysis of Gene Expression

Gene expression analysis was performed using the Gene Expression Profiling Interactive Analysis 2 (GEPIA2, http://gepia.cancer-pku.cn/; accessed on 22^nd May 2025), which integrates RNA-sequencing data from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) projects.^[23] Breast Invasive Carcinoma (BRCA) tumor samples were compared with normal breast tissues using the GEPIA2 “Expression DIY” module. The differential expression cut-off parameters were set to |log2FC| ≥ 1 and p < 0.01, consistent with previous reports.^[24] At the time of analysis, the BRCA dataset comprised 1,085 tumor samples and 291 normal samples (TCGA and GTEx combined). Differential expression analysis was performed using the statistical methods implemented in the GEPIA2 platform. GEPIA2 uses ANOVA for differential expression between tumor and normal tissues and the log-rank test for survival analysis.

Analysis of Overall Survival and Heat Map of Survival

Overall survival analysis of the six hub genes was performed using the GEPIA2 survival module with the default parameters.^[25] Patients were automatically divided into high- and low-expression groups according to the default GEPIA2 expression cut-off (median expression), and survival differences were evaluated using Kaplan–Meier survival curves and the log-rank test. The GEPIA2 server generated hazard ratios with 95% confidence intervals. Because survival analysis was performed only for six predefined hub genes identified through the network analysis, no additional multiple-testing correction was applied.

Molecular Docking

The PDB structural data of the six hub genes were obtained from the Protein Data Bank (PDB) (https://www.rcsb.org/; accessed on 25th May 2025).^[26] The PyMOL Molecular Graphics System, Version 2.3.2, developed by Schrödinger, LLC (http://www.pymol.org/pymol), was then used to view the downloaded target structure files in .pdb format.^[27] Using PyMOL, water molecules and other heteroatoms, such as Ca, Cl, and Na, were removed from the structures before energy minimization. Then, using PubChem (https://pubchem.ncbi.nlm.nih.gov/), the top five bioactive compounds (ligands) from P. longum were converted into their corresponding 3D conformations in the .sdf file format.^[2] Subsequently, the target molecules in .pdb format and ligand molecules in .sdf format were loaded into the cavity-detection guided blind docking program CB-Dock2 (http://cadd.labshare.cn/cb-dock2/; accessed on 25^th May 2025).^[28] CB-Dock2 automatically identifies potential ligand-binding cavities within the protein structure and ranks the predicted cavities according to their size. Based on the predicted cavities, the docking center and docking box dimensions are automatically determined according to the ligand structure. Molecular docking is subsequently performed using AutoDock Vina (https://github.com/ccsb-scripps/AutoDock-Vina). The resulting ligand-binding poses are ranked according to their predicted Vina scores, expressed in kcal/mol. These values represent the predicted binding affinity/docking scores of the corresponding protein–ligand complexes, with more negative values indicating a more favorable predicted interaction.

The top-ranked conformation for each protein–ligand complex was considered the optimal predicted binding pose. Finally, 2D schematic representations of predicted protein-ligand interactions were further visualized using two-dimensional interaction diagrams generated with the LIGPLOT tool (https://www.ebi.ac.uk/thornton-srv/software/LigPlus/; accessed on 25^th May 2025) with standard Protein Data Bank file input.^[29] The result is a PostScript file, in color or black and white, that identifies hydrogen bonds, hydrogen-bond distances, interacting amino acid residues, and hydrophobic contacts between the ligands and the respective target proteins.

Extract Preparation

Dried fruits of Piper longum Linn. were procured from the pharmacy of the All India Institute of Ayurveda (AIIA), New Delhi, India. The plant material was authenticated in the Department of Dravyaguna, AIIA, before use in accordance with institutional quality-control procedures and the standards of the Ayurvedic Pharmacopoeia of India. The fruits were thoroughly washed with distilled water to remove adhering dust, foreign particles, and damaged plant material, then shade-dried at ambient temperature for approximately 20 days until a constant weight was achieved. The dried fruits were then pulverized into a coarse powder using an electric grinder and stored in airtight containers under dry conditions until extraction.

To compare different extraction approaches, aqueous and hydroalcoholic extracts were prepared using both cold maceration and Soxhlet extraction techniques at the Department of Rog Nidan and Vikriti Vigyan, AIIA. For cold maceration, 400 g of the powdered fruit material was suspended in 5 L of sterile double-distilled water (solid-to-solvent ratio, 1:12.5 w/v) and continuously stirred at room temperature (25 ± 2°C) for 12 hours to facilitate efficient extraction of water-soluble phytoconstituents. The extract was subsequently filtered through Whatman No. 1 filter paper to remove insoluble plant debris.

For Soxhlet extraction, coarsely powdered P. longum fruits were extracted separately with distilled water and a hydroalcoholic solvent (water, 1:1, v/v) until the extraction solvent in the siphon tube became colorless, indicating exhaustive extraction of soluble constituents. The resulting extracts were filtered through Whatman No. 1 filter paper to remove particulate matter.

The filtrates from all extraction methods were concentrated by evaporation in a hot-air oven at 55°C for 24 hours until a semisolid crude extract was obtained. The dried extracts were transferred to sterile airtight containers and stored at 4°C until further use. The extraction yield percentage was calculated as follows:

The extraction yields are given in (Supplementary Table 1).

Cell Culture

The human breast adenocarcinoma cell line MCF-7, obtained from the National Center for Cell Science (NCCS), Pune, India, was maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin. Cells were cultured in a humidified incubator at 37°C with 5% CO₂ and 95% air. The culture medium was replaced every 2–3 days, and cells were passaged at approximately 70–80% confluency using 0.25% trypsin–EDTA. Cell viability was routinely assessed by the trypan blue exclusion assay, and only cultures with>95% viability and no visible contamination were used for experiments. For treatment studies, cells were seeded at the required density and allowed to adhere for 24 hours before exposure to the test extracts.

Preparation of Extract Dilutions

A stock solution of the aqueous P. longum extract was prepared at 1 mg/mL in phosphate-buffered saline (pH 7.4) and filter-sterilized before use. Immediately before each experiment, serial working dilutions were freshly prepared from the stock solution using complete culture medium to obtain final treatment concentrations of 0.25, 0.5, 1.0, 2.5, 3.75, 5.0, 7.5, 10, 12.5, 15, 20, 25, 30, and 50 µg/mL. Cells treated with complete culture medium alone served as the untreated control.

MTT Assay

The cytotoxic effect of Piper longum aqueous extract on MCF-7 human breast cancer cells was evaluated using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Briefly, 1 × 10⁴ MCF-7 cells were seeded into 96-well flat-bottom tissue culture plates at an appropriate density and allowed to attach overnight under standard culture conditions. The next day, the culture medium was replaced with fresh complete medium containing different concentrations of P. longum aqueous extract. Untreated cells maintained in complete culture medium served as the negative control.

After 24 hours of treatment, 20 μL of MTT solution (5 mg/mL prepared in phosphate-buffered saline) was added to each well, followed by incubation for an additional 4 hours at 37°C in a humidified atmosphere containing 5% CO₂. During this period, metabolically active cells reduced the yellow tetrazolium salt to insoluble purple formazan crystals. At the end of the incubation period, the culture medium containing excess MTT reagent was carefully aspirated without disturbing the crystals, and 100 μL of dimethyl sulfoxide (DMSO) was added to each well to completely dissolve the formazan crystals. The plates were gently agitated for 10–15 minutes to ensure complete solubilization before measuring the absorbance at 570 nm using a microplate reader. Cell viability was expressed as the percentage of viable cells relative to the untreated control using the following equation:

The concentration of P. longum aqueous extract required to inhibit 50% of cell viability (IC₅₀) was determined from the dose-response curve in GraphPad Prism version 10. Each concentration was tested in [number] technical wells in each of [number] independent experiments. Data are presented as mean ± SD.

Statistical Analysis

All experiments were performed in three independent biological replicates, with each treatment and control analyzed in triplicate wells. Data are expressed as the mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism version 10.0 (GraphPad Software, San Diego, CA, USA). The half-maximal inhibitory concentration (IC₅₀) of P. longum aqueous extract was determined by nonlinear regression analysis using a log(inhibitor) vs. normalized response (variable slope) dose-response curve model. A p-value of <0.05 was considered statistically significant.

P. longum Active Compounds

Literature review and IMPPAT database searches identified 82 bioactive compounds in P. longum fruit (Supplementary Table 2). All compounds were evaluated using SwissADME analysis to assess drug-likeness based on bioavailability score ≥0.3 and Lipinski’s rule of five, which is used to predict the drug pharmacokinetic properties based on the facts such as (1) a maximum of five hydrogen bond donors, (2) a maximum of 10 hydrogen bond acceptors, (3) a molecular mass less than 500 Da, and (4) the partition coefficient, or XLOGP3, must not exceed 5 (Supplementary Table 3). This screening yielded 19 promising ligands (Supplementary Table 2). The selected compounds had calculated molecular weights of approximately 222-370 Da and XLOGP3 values of 1.23–4.70. Predicted GI absorption and blood-brain barrier penetration have also been reported separately for each compound in Supplementary Table 3. Additionally, most compounds were not P-glycoprotein (PGP) substrates, suggesting reduced efflux and enhanced bioavailability and reinforcing their potential as effective therapeutic candidates.

Breast Cancer and P. longum Potential Targets

Following the ADME analysis, the bioactive compounds underwent target prediction using the SuperPred, SwissTargetPrediction, and DIGEP-Pred databases, yielding 11,812 predicted targets. A thorough search using the keyword “Breast Cancer” in OMIM, the TTD, and GeneCards identified 954 target genes. After reviewing the target genes linked to P. longum and breast cancer, we found an overlap of 600 targets, as shown in Figure 1b and listed in Supplementary Table 4.

Figure 1.

(a) Predicted BOILED-Egg diagram of the selected compounds. BBB - Blood Brain Barrier, HIA - Human Intestinal Absorption, PGP+ - substrate of P-glycoprotein and PGP- - non-substrate of P-glycoprotein. (b) Intersecting targets between the active phytochemical potential protein targets in P. longum and Breast cancer-related genes. (c) The PPI network of the 600 intersecting targets, (d) The PPI network of the top 23 target genes, and (e) Compound – Target – Disease network of breast cancer and P. longum.

Protein–Protein Interaction (PPI) Network

Of the 600 overlapping targets, 516 were included in the STRING interaction network under the predefined high-confidence interaction threshold, whereas the remaining targets were excluded because they lacked qualifying interactions. To refine the analysis, we ranked nodes by degree and generated a focused PPI network of the top 23 genes using Cytoscape (Figure 1d). These genes included TP53, CTNNB1, STAT3, EP300, BRCA1, JUN, HDAC1, SMAD3, FOXO3, EGFR, HIF1A, CREBBP, HSP90AA1, CDK2, SMAD2, CDK4, SMARCA4, ATM, GRB2, CDKN1B, PTPN11, CD44, and SHC1.

The refined network consisted of 23 nodes and 139 edges, with an average of 12.1 neighbors per node. Degree centrality (DC), betweenness centrality (BC), and closeness centrality (CC) values are provided in Supplementary Table 5. DC quantifies the total number of direct connections a node possesses. DC reflects the number of direct connections, BC indicates a node’s role in shortest paths, and CC measures proximity to all nodes.^[30] Six genes—TP53, CTNNB1, STAT3, SMAD3, EGFR, and HSP90AA1—showed above-average DC, BC, and CC, identifying them as hub genes (Supplementary Table 5). The network hierarchy for hub gene selection is as follows:

600 overlapping targets → 516 STRING-network nodes → Cytoscape → CytoNCA → Degree, Betweenness Centrality (BC), and Closeness Centrality (CC) → 23 top-ranked network genes → 6 final hub genes.

Construction and Analysis of C-T-D Network

The Cytoscape application was used to create a compound-target-disease network, enabling exploration of the complex interactions among compounds, targets, and breast cancer. Figure 1e depicts the C-T-D network, which encompasses 200 nodes and 400 edges, with an average of 4 neighboring nodes. Supplementary Table 6 presents the degree, betweenness, and closeness centralities for the top 15 bioactive compounds associated with breast cancer. The five compounds with the highest degree centrality were Pellitorine (20), Aristolodione (18), Aristolactam AII (18), Norcepharadione B (16), and N-isobutyl decadienamide (15).

Gene Ontology (GO) Enrichment and KEGG Pathway Analysis

Targets of P. longum were analyzed using functional annotation and enrichment analyses, revealing key biological roles associated with its phytochemicals. Gene Ontology (GO) and KEGG enrichment analyses were performed on the 23 top-ranked genes selected from the PPI network. Gene Ontology (GO) analysis showed involvement in biological processes (BP) such as responses to growth factors, regulation of signal transduction, and responses to chemical and oxygen-containing stimuli. Cellular components (CC) included SMAD complexes, cyclin-dependent kinase complexes, transcription repressor complexes, euchromatin, DNA repair complexes, and chromosomes. Molecular functions (MF) comprised p53 binding, transcription coactivator binding, kinase and phosphatase binding, and transcription regulation. Using an FDR cutoff of 0.05, 631 BP, 19 CC, and 74 MF terms were identified across 23 genes (Supplementary Figure 2). KEGG pathway analysis revealed 101 associated pathways, including cancer-related pathways, breast cancer, chronic myeloid leukemia, microRNAs in cancer, Forkhead box O (FOXO) signaling, PI3K-Akt signaling, cell cycle regulation, viral carcinogenesis, and hepatitis C. These findings highlight the therapeutic relevance of P. longum.

Unique Gene Expression

We assessed the expression levels of six breast cancer hub genes, including TP53, CTNNB1, STAT3, SMAD3, EGFR, and HSP90AA1, by comparing them with the BRCA dataset using the GEPIA2 web tool (Supplementary Figure 3). The Cancer Genome Atlas-BRCA (TCGA_BRCA) dataset comprises expression data from 1,085 tumor tissues and 291 corresponding normal tissues related to BRCA. In the box plot, tumor samples are represented in red, while normal samples are shown in grey. A red asterisk indicates a statistically significant p-value (below 0.01), implying a potentially important difference in gene expression between tumor and normal tissues. The y-axis represents the mean log2 (TPM + 1) values for gene expression levels. This transformation is commonly used in gene expression studies to stabilize variance across expression levels. At the stated cutoff, only EGFR and HSP90AA1 showed significant tumor-normal differences in the GEPIA2 analysis shown. HSP90AA1 was significantly upregulated and overexpressed in the TCGA_BRCA dataset. Additionally, TP53 was also overexpressed in the TCGA_BRCA dataset, but the difference was not significant (p > 0.05). In contrast, EGFR’s expression was significantly downregulated in this dataset. Moreover, STAT3 and SMAD3 expression was downregulated (p > 0.05) (Supplementary Figure 4). CTNNB1 showed a negligible change in expression in the TCGA_BRCA dataset (Supplementary Figure 3).

Heat Maps and Survival Analysis of Unique Genes

Using GEPIA2, a survival analysis was conducted by comparing gene expression levels identified in the survival heat map analysis with the TCGA_BRCA dataset (Supplementary Figure 4a). Among the six genes associated with breast cancer, HSP90AA1 and TP53 exhibited high expression levels in the BRCA dataset. Conversely, the remaining genes—CTNNB1, STAT3, SMAD3, and EGFR—demonstrated minimal expression in the BRCA dataset (Supplementary Figure 4a). A survival heatmap visually displays survival data, illustrating the relationship between specific variables (in this case, the six hub genes) and survival outcomes (such as overall survival time) across individuals or groups. The prognostic impact helps identify genes associated with improved or poorer survival, aiding the discovery of potential biomarkers for prognosis and treatment. Overall-survival analysis showed that the prognostic associations varied among the six hub genes. TP53 expression was significantly associated with poorer overall survival (HR = 1.4, p = 0.038), whereas the associations observed for HSP90AA1, CTNNB1, STAT3, SMAD3, and EGFR did not reach statistical significance under the analysis conditions used. Moreover, some genes exhibiting high expression levels and correlations with survival in BRCA patients included CTNNB1 (HR (Hazard ratio) = 0.94; p = 0.71), STAT3 (HR = 0.96; p = 0.82), SMAD3 (HR = 0.91; p = 0.58), and EGFR (HR = 0.9; p = 0.53). In contrast, the expression levels of TP53 (HR = 1.4, p = 0.038) and HSP90AA1 (HR = 1.3, p = 0.079) were associated with poorer prognoses and reduced survival rates among BRCA patients (Supplementary Figure 4b–g). The survival analysis suggested a potential association between higher TP53 expression and poorer overall survival (p = 0.038), whereas the association observed for HSP90AA1 was weaker and did not reach statistical significance. These findings should therefore be considered exploratory and require validation in independent cohorts (Supplementary Figure 4).

Molecular Docking and Visualization Analysis

Molecular docking was employed to predict the potential interactions between the selected P. longum phytochemicals and six breast cancer-associated hub proteins. The protein structures used for docking were obtained from the Protein Data Bank (PDB) and included 4MZI (TP53), 8Y0G (CTNNB1), 6NUQ (STAT3), 1MJS (SMAD3), 1M17 (EGFR), and 4BQG (HSP90AA1).^[31–36] The predicted ligand docking score values for every hub gene are shown in Supplementary Table 7, in which the corresponding values, expressed in kcal/mol, represent the AutoDock Vina scores generated during the docking process and were used to rank the predicted binding poses. More negative Vina scores indicate more favorable predicted docking scores. Among the evaluated complexes, eleven protein–ligand interactions showed predicted binding affinities below −7 kcal/mol. These include HSP90AA1_Norcepharadione B (−10.9 kcal/mol), HSP90AA1_Aristolactam AII (−10.7 kcal/mol), HSP90AA1_Aristolodione (−10.3 kcal/mol), EGFR_Aristolodione (−9.1 kcal/mol), HSP90AA1_N-isobutyl decadienamide (−9 kcal/mol), EGFR_Norcepharadione B (−8.9 kcal/mol), EGFR_Aristolactam AII (−8.4 kcal/mol), STAT3_Aristolodione (−7.6 kcal/mol), HSP90AA1_Pellitorine (−7.5 kcal/mol), STAT3_Norcepharadione B (−7.2 kcal/mol), and TP53_Aristolodione (−7.1 kcal/mol). Figure 2 illustrates representative 3D and 2D interaction plots of five selected protein–ligand complexes showing strong predicted binding affinities.

Figure 2.

Molecular docking results of key bioactive compounds of P. longum with breast cancer hub genes. (a) HSP90AA1_Norcepharadione B; (b) HSP90AA1_Aristolactam AII; (c) HSP90AA1_Aristolodione; (d) EGFR_Aristolodione; (e) HSP90AA1_N-isobutyl decadienamide.

Piperine Assessment for Validation of Extracts

To evaluate the validation of different extracts of P. longum, the extracts were analyzed for the presence of piperine, one of the most abundant alkaloids found in P. longum fruits, using HPTLC (Figure 3a–d). As shown in Figure 3b–d, the HPTLC chromatogram indicated the presence of piperine in hydro-alcoholic extracts prepared by the Soxhlet method (lanes 1 and 2), aqueous extracts obtained through cold maceration (lanes 5 and 6), and dried fruit extract (lanes 7 and 8), all compared to piperine standards with an R_F value of approximately 0.16. The aforementioned extracts were then tested for microbial contamination at the Department of Microbiology, AIIA. Subsequently, a 1 mg/mL stock solution of this extract was prepared in PBS (pH 7.2–7.4) and stored at −20°C for use in in vitro experiments (Figure 3e). Notably, piperine was used solely as a phytochemical marker for extract characterization and was not selected based on its network pharmacology ranking. In contrast, Norcepharadione B and Aristolactam AII were prioritized independently through the computational network pharmacology and molecular docking analyses based on their predicted interactions with breast cancer-associated targets. Thus, HPTLC detection of piperine provides chemical evidence for the characteristic composition of the extract but should not be interpreted as experimental validation of the predicted activities of Norcepharadione B, Aristolactam AII, or other computationally prioritized compounds.

Figure 3.

(a) Extract preparation. (b–d) HPTLC profile of standard and isolated piperine from different extraction methods and extract of P. longum dried fruits at 254nm (b), 366nm (c) and after derivatization (d) [1–2: Isolated Piperine from hydro-alcoholic Soxhlet method; 3–4: aqueous Soxhlet method; 5–6: cold maceration method; 7–8: P. longum dried fruit extract; 9–15: Piperine standard]. (e) Anti-microbial assessment of different extracts (A-Hydro-alcoholic extract by the Soxhlet method; B-aqueous extract by the Soxhlet method, and C-aqueous extract by the cold maceration method) on a Macconkey agar plate. (f) Line graph showing the effect of aqueous extract of P. longum on the growth of MCF-7 cell line using MTT assay.

Since we needed to treat the MCF-7 breast cancer cells with these extracts for subsequent experiments, we preferred to use the aqueous extract of P. longum prepared by cold maceration to minimize cell death caused by alcohol in the hydro-alcoholic extract.

Effect of P. longum Extract on MCF-7 MTT Signal

The effect of P. longum extract on MCF-7 MTT signal was analyzed after 24 hours of treatment using an MTT assay. The assay was conducted in both the extract-treated and untreated (control) groups. IC₅₀ was reached at an extract concentration of 3.75 μg/mL (Figure 3f) after 24 hours post-treatment. Interestingly, increasing P. longum extract concentration was associated with a decrease in MCF-7 cell MTT signal; data are presented as mean ± standard deviation (SD) (Supplementary Table 8).

Strengths and Limitations

Taken together, the findings support a model in which the aqueous P. longum extract might exert anticancer activity through a multi-level mechanism. Network pharmacology predicted several cancer-associated molecular targets and pathways, while molecular docking prioritized putative compound-target interactions for further investigation. An in vitro study in MCF-7 cells demonstrated that the aqueous extract of P. longum significantly inhibits the MTT signal of the breast cancer cell line. These results suggest that the anticancer effects of P. longum may arise from the coordinated modulation of multiple signaling pathways involved in proliferation and survival.

The present study has several limitations that should be acknowledged. First, although the aqueous P. longum extract was characterized using HPTLC with piperine as a marker compound, comprehensive chemical standardization and quantitative estimation of the prioritized phytochemicals identified through network pharmacology, including Norcepharadione B and Aristolactam AII, were not performed. Therefore, the biological activity observed in the present study cannot be directly attributed to any individual constituent of the extract. Second, the in vitro experiments used only a single breast cancer cell line (MCF-7). Validation in additional breast cancer subtypes, including triple-negative and HER2-positive cell lines, would improve the generalizability of the findings. Third, the present study did not include a non-tumorigenic human breast epithelial cell line as a normal-cell comparator. Consequently, the selectivity of the extract toward cancer cells relative to normal cells could not be assessed. Finally, although the MTT assay demonstrated a significant reduction in cellular metabolic activity following treatment with the aqueous P. longum extract, plant-derived compounds may occasionally interfere with tetrazolium-based colorimetric assays through direct chemical reduction or intrinsic color. To minimize this possibility, the MTT findings should be complemented with independent cell-cycle and Annexin V/PI apoptosis assays, which consistently demonstrated cell-cycle arrest and increased apoptotic cell death. Nevertheless, additional orthogonal viability assays (such as ATP-based luminescence or live-cell imaging assays) would further strengthen these observations. An additional limitation of the present network pharmacology approach is that hub-gene prioritization was based on network topology analysis. Centrality measures, including degree, betweenness, and closeness centrality, preferentially identify highly connected and extensively studied genes within biological interaction networks. Consequently, these metrics do not necessarily identify the most disease-specific, biologically causal, or therapeutically druggable targets. Therefore, the hub genes identified in this study should be regarded as computationally prioritized candidates for further investigation rather than experimentally validated therapeutic targets. Future studies integrating functional genomics, target validation, and pharmacological experiments are needed to establish their biological significance and clinical relevance.

In brief, the computational predictions should not be interpreted as experimentally confirmed molecular mechanisms. Direct validation of the predicted interactions and pathways through target-specific approaches, including quantitative PCR, Western blotting, assessment of protein phosphorylation, and functional target inhibition or rescue experiments, will be warranted and will be carried out in our lab in the near future.