Investigating the therapeutic potential of Lepidium sativum-derived compounds in prostate cancer: an integrative in silico approach

Abstract

This study investigates the therapeutic potential of bioactive compounds from Lepidium sativum in the treatment of prostate cancer. Through a series of in silico analyses, three key compounds were identified and evaluated for their drug-likeness, pharmacokinetic properties, and safety profiles. These compounds demonstrated favorable drug-likeness according to Lipinski’s rule and other drug-likeness criteria, high gastrointestinal tract absorption, and non-inhibition of major cytochrome P450 enzymes. Protein–protein interaction network analysis identified ten hub genes, with AKT1 and PIK3CA emerging as prime targets for therapeutic intervention. Functional annotation and pathway analysis highlighted key biological processes and pathways associated with prostate cancer, emphasizing the significance of these targets. Molecular docking and dynamic simulation studies further confirmed the binding affinities and stability of the identified compounds with both mutated and non-mutated forms of the target genes. These findings suggest that compounds from Lepidium sativum hold promise as potential therapeutics for prostate cancer, warranting further investigation and development.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40203-025-00482-7.

Introduction

Cancer remains a global health challenge, with breast, cervical, prostate, colorectal, and liver cancers accounting for nearly half of all new cancer cases worldwide (Moeti 2023). Cancer-related mortality is expected to rise significantly, reaching nearly one million deaths annually by 2030. This projection highlights the urgent need for proactive intervention strategies (Grossman et al. 2018).

Some cancers progress slowly and exhibit low-grade biological behavior, which offers a therapeutic window of opportunity (Harvey et al. 2012). However, early detection remains challenging. Prostate cancer, in particular, often develops without clear symptoms in its initial stages. It is currently the most frequently diagnosed malignancy in more than half of all countries worldwide, emphasizing its significant public health impact (Grossman et al. 2018).

Dietary patterns are increasingly recognized as key modulators of cancer risk (Steck and Murphy 2020; Grosso et al. 2017; Ruiz and Hernández 2014). Certain dietary components demonstrate protective effects, while others may increase susceptibility to malignancy. Several studies have highlighted the anticancer potential of plant extracts, suggesting their relevance for prostate cancer prevention and therapy (Ma and Peng 2024; Raad et al. 2024). For instance, Salehi et al. (2019); Thomas-Charles and Fennell (2019) reported that plant-derived bioactive compounds can modulate prostate cancer cell proliferation. Similarly, Loeb et al. (2022) observed that higher intake of plant-based foods is associated with a reduced risk of aggressive prostate cancer, particularly in men under 65 years of age.

Lepidium sativum, a member of the Cruciferae or Brassicaceae family, has demonstrated anti-inflammatory, hypoglycemic, and antimicrobial properties (Al-Snafi 2019; Painuli et al. 2022; Jambor et al. 2022; CHETNA BAREGAMA and ANJU GOYAL 2019). Some studies have reported its potential in prostate cancer treatment (Taïbi et al. 2020), and ethnopharmacological evidence support its traditional use for prostate-related disorders (Ibrahim et al. 2023).

However, most available studies are limited to static docking predictions and do not investigate the dynamic behavior or mechanistic stability of ligand-target interactions. Additionally, no previous investigation has evaluated whether compounds from L. sativum can mimic the binding behavior of clinically relevant AKT1 and PIK3CA inhibitors. This is particularly important when considering oncogenic mutations such as AKT1-E17K and PIK3CA-H1047R, which are known to contribute to drug resistance and hyperactivation of the PI3K/AKT signaling pathway in prostate cancer (Pearson et al. 2018).

To address these gaps, this study combines target prediction, network pharmacology analysis, molecular docking, and TTClust-guided molecular dynamics simulations against both wild-type and mutant proteins. Residue-level interactions are further benchmarked against clinically investigated inhibitors. This integrated strategy aims to determine whether L. sativum-derived compounds can maintain clinically relevant interaction patterns within the AKT1/PIK3CA signaling axis.

Materials and methods

Screening of active compounds

A total of 398 compounds reported in Lepidium sativum were retrieved from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Their canonical SMILES notations were used to assess drug-likeness and pharmacokinetic properties. The SwissADME web server (http://www.swissadme.ch/) was used to evaluate Lipinski’s Rule of Five, Ghose, Veber, Egan, and Muegge criteria. Pharmacokinetic features such as gastrointestinal absorption, blood–brain barrier permeability, cytochrome P450 inhibition, and P-glycoprotein substrate status were also analyzed.

Toxicity profiles were predicted using the ProTox-III webserver (Banerjee et al. 2018; Drwal et al. 2014). Based on these filters, three compounds with favorable drug-like and safety properties were selected as potential active molecules (Supplementary File S1). The full compound filtering process is visually summarized in Fig. 1. The 2D structures of these bioactive constituents were depicted using ChemDraw Ultra version 14.0 (CambridgeSoft Corporation 2014).

Fig. 1.

Schematic workflow diagram of the computational pipeline for multitarget analysis

Screening of potential target genes

The three selected compounds were submitted to SwissTargetPrediction (Daina et al. 2019), PharmMapper (Ferrari et al. 2023; Liu et al. 2010; Wang et al. 2017), and SuperPred (Nickel et al. 2014) to predict potential targets. Only human (Homo sapiens) proteins were retained, and UniProt IDs were used for target standardization (Consortium 2019; Zaru et al. 2023). Three thresholds were applied to select the most significant targets: a Z-score of for PharmMapper, absolute values of scores (probabilities) of for SwissTargetPrediction (Guzmán-Flores et al. 2023), and a probability of that the input structure binds with the specific target for SuperPred. After removing duplicates, 266 compound-associated targets were obtained (Supplementary File S2).

Prostate cancer-related genes were then collected from GeneCards (8951 targets) (Safran et al. 2010), with the search limited to “prostate cancer”; DisGeNET (4388 targets) (Piiñero et al. 2020), with the search limited to Name: Prostate carcinoma; CUI: C0600139; and CTD (690 targets) (Davis et al. 2019), with the search limited to “prostate cancer”. A Venn diagram (https://www.interactivenn.net/) was used to identify overlapping targets between Lepidium sativum compounds and prostate cancer-associated genes. These intersecting genes were selected for downstream analysis.

Analysis of protein–protein interaction (PPI)

The intersecting target genes were analyzed using STRING Version 12.0 (https://string-db.org/) (Szklarczyk et al. 2021). The analysis was restricted to Homo sapiens. A confidence score threshold of 0.400 (medium confidence) was applied to ensure meaningful interactions. The resulting PPI network was imported into Cytoscape v3.10.1 (Shannon et al. 2003) for visualization. Hub genes were identified using the CytoHubba plugin (Chin et al. 2014), ranked by degree centrality, and the top 10 were selected for further analysis.

Functional annotation and pathway analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using DAVID (https://david.ncifcrf.gov/) (Huang et al. 2009). The analysis focused on the top 10 hub genes, with species set to Homo sapiens. GO terms were categorized into biological process (BP), molecular function (MF), and cellular component (CC). KEGG pathway analysis was used to identify dominant signaling mechanisms. Terms with a p-value of were considered significant, and the top 10 GO and KEGG outputs were retained for interpretation.

Therapeutic potential of hub genes

The therapeutic tractability of the top hub genes was evaluated using the Open Targets Platform (https://www.targetvalidation.org). For each gene, we retrieved association scores, druggability classification (e.g., small molecule or antibody tractable), and available clinical trial or drug information.

Mutation analysis

Mutation frequencies of the prostate cancer targets associated with Lepidium sativum compounds were examined using the cBioPortal database (https://www.cbioportal.org). Data were extracted from the Prostate Adenocarcinoma (MSK, Clin Cancer Res. 2022) cohort (Chakraborty et al. 2022), which includes 1417 samples. Genetic alterations were visualized using the OncoPrint module with z-score normalization. Pairwise hub-gene mutation co-occurrence was calculated using the odds ratio (OR).

Expression analysis of hub genes

The GEPIA2 web server (http://gepia2.cancer-pku.cn/) was used to compare hub gene expression between prostate cancer and normal tissues. A and were used as selection thresholds. Disease-free survival analysis was also conducted using GEPIA2 with a log-rank test, with significance set at .

Molecular docking analysis

Molecular docking was used to characterize ligand–target interactions. The 3D structures of the selected targets, including both mutant and wild-type forms of AKT1 (PDB IDs: mutant 3QKM and wild-type 4GV1) and PIK3CA (PDB IDs: mutant 4JPS and wild-type 4L23), were retrieved from the RCSB Protein Data Bank (PDB) in PDB format (https://www.rcsb.org/). To ensure translational relevance, we specifically selected 3QKM (AKT1-E17K) and 4JPS (PIK3CA-H1047R) mutant structures, as these hotspot mutations are known to drive androgen-independent PI3K/AKT signaling and reduced inhibitor sensitivity in prostate cancer, making them clinically relevant models for binding assessment.

The initial steps involved preparing the protein structure by eliminating surrounding water molecules near the protein. Grid coordinates associated with co-crystallized ligands were recorded, and the existing ligand was removed to clear the active site. Gasteiger charges were computed, and polar hydrogen atoms were added, while non-polar hydrogen atoms were merged, maintaining the protein structure’s rigidity. Conversely, the ligand was treated as flexible to explore favorable binding conformations.

Molecular docking was conducted using AutoDock Vina software (Trott and Olson 2010). Docking grid boxes were centered on the coordinates of the co-crystallized ligand in each respective structure. The detailed center coordinates, box dimensions, and calculated binding affinities (Vina scores) for all protein–ligand complexes are summarized in Table 3. All docking runs, including those for the clinical reference inhibitors (Afuresertib, Copanlisib, Dactolisib, Apitolisib), were performed using identical Vina parameters (exhaustiveness = 8, energy range = 3 kcal/mol) to ensure benchmarking consistency.

Table 3.

Molecular docking parameters for inhibitors bound to AKT1 and PIK3CA

	Vina score	Cavity size	Center			Size
			x	y	z	x	y	z
AKT1
Afuresertib_4GV1 (reference)		601		7	12	19	19	19
4GV1_Compound1		601		7	12	19	19	19
4GV1_Compound2		418			14	20	20	20
4GV1_Compound3		418			14	21	21	21
Afuresertib_3QKM (reference)		818	26	7	12	19	19	19
3QKM_Compound1		818	26	7	12	19	19	19
3QKM_Compound2		818	26	7	12	20	20	20
3QKM_Compound3		818	26	7	12	21	21	21
PIK3CA
4JPS_Dactolisib (reference)		528			21.8	23	23	29
4JPS_Apitolisib (reference)		528			21.8	25	25	25
4JPS_Compound1		528			21.8	19	19	29
4JPS_Compound2		528			21.8	20	20	29
4JPS_Compound3		528			21.8	21	21	29
4L23_Copanlisib (reference)		485	30.8	43.7	39.5	25	25	25
4L23_Dactolisib (reference)		485	30.8	43.7	39.5	23	23	23
4L23_Apitolisib (reference)		485	30.8	43.7	39.5	25	25	25
4L23_Compound1		485	30.8	43.7	39.5	19	19	28
4L23_Compound2		485	30.8	43.7	39.5	20	20	28
4L23_Compound3		485	30.8	43.7	39.5	21	21	28

The top-ranked docking poses, based primarily on binding energy, were prioritized for further analysis. The docking score served as a crucial evaluation criterion to filter out potential compounds and their respective targets. Additionally, BIOVIA Discovery Studio Visualizer v24.1.0.23298 and the PLIP web server were utilized to visualize 2D and 3D interactions of docking complexes.

Molecular dynamics simulation

Docking-derived complexes were subjected to MD simulation using GROMACS 2020. The topology and parameter files for the systems were generated using the CHARMM-GUI web server (https://charmm-gui.org/) with the CHARMM36 force field (Vanommeslaeghe et al. 2010). Each system was placed in a cubic simulation box and solvated with TIP3P water molecules, applying periodic boundary conditions in all three dimensions.

To maintain system neutrality, sodium (Na) and chloride (Cl) counterions were added, ensuring an isosmotic condition of 0.15 M NaCl. MD simulations were conducted for 20 ns. The temperature was kept constant at 300.15 K during the NVT equilibration phase, while the NPT equilibration phase maintained a pressure of 1 bar.

Various parameters were utilized to evaluate ligand–protein interactions. Hydrogen bonds were analyzed to assess the interactions between the ligand and the protein. The root mean square deviation (RMSD) of the protein backbone was monitored to gauge its stability throughout the simulation. The radius of gyration (Rg) and the root mean square fluctuation (RMSF) of amino acid residues were also examined to understand the flexibility of the protein and its response to ligand binding.

Conformational clustering of MD trajectories was performed with TTClust v4.10.4 (Tubiana et al. 2018) on backbone-aligned frames (stride = 1). To evaluate sampling robustness, we ran a sensitivity analysis with the number of clusters (NG) ranging from 2 to 6; representative frames and inter-cluster RMSD matrices were recorded and provided in Supplementary File S3.

Results and discussion

Screening of active compounds and target predictions

Figure 2 depicts the three compounds that were selected: Compound 01: Propan-2-yl 2-(1,3-dioxoisoindol-2-yl), CID: 76312568; Compound 02: Butan-2-yl 2-(1,3-dioxoisoindol-2-yl)oxypropanoate, CID: 76319801; Compound 03: Propanoic acid, 2-[(1,3-dihydro-1,3-dioxo-2 H-isoindol-2-yl)oxy]-, 1,1-dimethylethyl ester, (2R)-, CID: 13577112. The drug-likeness of Lepidium sativum compounds was evaluated according to Lipinski’s rules (Lipinski et al. 2012), along with Veber (Veber et al. 2002), Ghose (Ghose et al. 1999), Egan (Egan et al. 2000), and Muegge (Muegge et al. 2001) criteria for drug candidates. These compounds met all specified criteria without violations: fewer than five hydrogen bond donors, fewer than 10 hydrogen bond acceptors, molecular weight below 500 Da, and log P value under five. The drug-likeness assessment results are summarized in Table 1. Since these compounds comply with Lipinski’s rule, they are considered potential orally bioavailable agents.

Fig. 2.

2D structure of selected Lepidium sativum compounds

Table 1.

ADME properties calculated for selected Lepidium sativum-Derived Compounds using SwissADME and ProTox-III webservers

Compounds	Compound 01	Compound 02	Compound 03
MW	277.27	291.3	291.3
TPSA	72.91	72.91	72.91
WLOGP	1.17	1.56	1.56
GI absorption	High	High	High
BBB permeant	No	No	No
Pgp substrate	No	No	No
CYP1A2 inhibitor	No	No	No
CYP2C19 inhibitor	No	No	No
CYP2C9 inhibitor	No	No	No
CYP2D6 inhibitor	No	No	No
CYP3A4 inhibitor	No	No	No
Lipinski violations	0	0	0
Ghose violations	0	0	0
Veber violations	0	0	0
Egan violations	0	0	0
Muegge violations	0	0	0
Bioavailability Score	0.55	0.55	0.55
Hepatotoxicity	Inactive	Inactive	Inactive
Cardiotoxicity	Inactive	Inactive	Inactive
Carcinogenicity	Inactive	Inactive	Inactive
Immunotoxicity	Inactive	Inactive	Inactive
Mutagenicity	Inactive	Inactive	Inactive
Cytotoxicity	Inactive	Inactive	Inactive

As summarized in Table 1, the selected compounds exhibit fewer than ten rotatable bonds, suggesting good oral bioavailability. Their topological polar surface area (TPSA) values (Ertl et al. 2000) fall within the optimal range (20–130 Å), further supporting their suitability as orally available drug candidates.

The BOILED-Egg model (Brain Or IntestinaL EstimateD permeation method) (Fig. 3), which plots TPSA against WLogP (Wildman and Crippen 1999), confirms high gastrointestinal absorption and acceptable lipophilicity. These compounds demonstrate high gastrointestinal tract absorption potential, positioning them as promising drug candidates. Notably, they are not P-glycoprotein (P-gp) substrates, which increases their potential effectiveness against multidrug-resistant cancer cells that overexpress this transporter (Sharom 2008). Their inability to cross the blood–brain barrier suggests limited central nervous system penetration, which may reduce neurological side effects. Predictions regarding potential drug–drug interactions through Cytochromes P450 (CYP) inhibition indicate that these compounds are not inhibitors of CYP1A2, CYP2C19, CYP2C9, CYP2D6, or CYP3A4. This minimizes the risk of metabolic interference with co-administered drugs. These pharmacokinetic attributes support their potential for safe gastrointestinal absorption while limiting blood–brain barrier permeation (Hollenberg 2002; Huang et al. 2008; Kirchmair et al. 2015).

Fig. 3.

Boiled-egg plot (TPSA vs. WLogP) predicting ADME properties of L. sativum compounds. All compounds exhibit high gastrointestinal absorption (white region) without blood–brain barrier penetration or P-glycoprotein substrate activity, indicating favorable oral bioavailability with reduced risk of neurological side effects

Furthermore, safety evaluation (Table 1) demonstrates that Lepidium sativum compounds exhibit a favorable toxicity profile. They show no hepatotoxicity (Siramshetty et al. 2016) and no signs of carcinogenicity (Kroes et al. 2004), immunotoxicity (Schrey et al. 2017), mutagenicity (Ames et al. 1973), or cytotoxicity (Freshney 2015). This favorable safety profile supports their potential for therapeutic applications.

Protein–protein interaction (PPI)

To construct the protein–protein interaction (PPI) network, we used the 34 genes identified from the Venn diagram in Fig. 4, which represent the common intersection between Lepidium sativum targets and prostate cancer-associated genes. These genes were submitted to the STRING database to evaluate their interaction network. The PPI network visually represents the functional relationships among the target genes under disease conditions. In this network, nodes represent proteins, and edges represent predicted or experimentally validated interactions.

Fig. 4.

Venn Diagram of Targets Identified from Lepidium sativum Compounds and Prostate Cancer

Figure 5a shows the resulting PPI network of the 34 genes visualized using Cytoscape v3.10.1. To identify the most influential genes, we applied the CytoHubba plugin and ranked nodes based on four centrality measures: degree, betweenness, closeness, and maximal clique centrality (MCC). Genes that ranked within the top ten in at least two metrics were considered robust hub nodes. This approach identified MDM2, AKT1, GSK3B, STAT3, ESR1, ESR2, MMP9, and PPARA as key hubs. Figure 5b displays the degree-ranked network where these hub genes are highlighted in red. The complete scoring for each centrality measure is available in Supplementary File S5, confirming the convergence of these eight genes as major regulatory nodes.

Fig. 5.

a PPI Network Resulting from the 34 Genes. b Top Ten Genes Based on the Degree Method

The enrichment of the prolactin signaling pathway (hsa04917) is particularly relevant, as prolactin and its receptor are frequently overexpressed in prostate cancer cells. This signaling contributes to the development of androgen-independent (castration-resistant) prostate cancer (CRPC). Mechanistically, prolactin activates the STAT (signal transducer and activator of transcription) family, promoting cell survival and proliferation. If the identified hub genes are modulated by L. sativum components, this could directly disrupt CRPC-associated signaling. Similarly, the enrichment of the proteoglycans in cancer pathway (hsa05205) highlights the role of extracellular matrix (ECM) remodeling in tumor progression. Proteoglycans are overexpressed in the prostate tumor microenvironment and are involved in tumor growth, metastasis, and angiogenesis. The presence of MMP9 among the hub genes further supports this observation, as MMP9 is a matrix metalloproteinase responsible for ECM degradation and metastatic spread. Modulation of this pathway by the selected compounds suggests a potential anti-invasive and anti-metastatic effect through ECM stabilization and inhibition of proteolytic enzymes involved in tumor expansion.

Functional annotation and pathway analysis

The biological functions of the prostate cancer targets were investigated through functional annotation and enrichment analysis of the hub genes selected from the PPI network. This analysis resulted in the enrichment of 65 biological processes (BP), 11 cellular components (CC), 26 molecular functions (MF), and 76 KEGG pathways. A bubble plot of the top enriched terms ranked by p-value and gene count is shown in Fig. 6. GO biological process analysis showed that the targets were involved in negative regulation of gene expression, cellular response to estrogen stimulus, negative regulation of apoptosis, and positive regulation of DNA-templated transcription. The enrichment of “negative regulation of apoptosis” and “positive regulation of transcription” suggests a direct association with two hallmarks of cancer: resistance to cell death and uncontrolled proliferation.

Fig. 6.

Functional annotation and enriched pathways representation in the form of Bubble Plot: GO in terms of biological processes (BP), GO cellular components (CC), KEGG pathway analysis, and GO in terms of molecular function (MF)

The GO cellular component analysis indicated that most hub genes were localized in the cytosol, cytoplasm, and plasma membrane. This localization suggests that several of these proteins are accessible to small-molecule inhibition and are potentially druggable targets.

In terms of GO molecular function analysis, the hub genes were linked to RNA polymerase II transcription factor activity, ligand-activated sequence-specific DNA binding, and enzyme binding. These functions highlight the regulatory role of genes such as STAT3 and ESR1/ESR2 in controlling oncogenic transcriptional programs.

KEGG pathway enrichment revealed that the genes were mainly involved in cancer pathways, prolactin signaling, endocrine resistance (hsa01522), proteoglycans in cancer, chemical carcinogenesis via receptor activation, lipid metabolism, atherosclerosis, and notably, prostate cancer (involving 5 genes: GSK3B, PIK3CA, MDM2, AKT1, MMP9). The enrichment of these clinically relevant pathways confirms that the identified hub genes are central contributors to prostate cancer progression and therapeutic resistance.

Molecular tractability potential of hub genes

As shown in Table 2, the most promising therapeutic targets were AKT1, PIK3CA, ESR1, and ESR2, based on their high global scores and favorable tractability profiles.

Table 2.

Key information on 10 genes and their disease associations, drug actions, clinical trials, and tractability predictions

Gene	Global score	Known drugs	Phase	Tractability predictions
				Small molecule clinical precedence	Antibody clinical precedence	Protac
ESR2	0.485	ESTRAMUSTINE	Phase III	5	3	5
		PHOSPHATE
		SODIUM
STAT3	0.355	–	–	3	1	4
SOD2	0.014	–	–	1	–	3
PPARA	0.011	–	–	5	–	1
MMP9	0.082	–	–	5	4	2
AKT1	0.627	IPATASERTIB	Phase III	5	2	5
		CAPIVASERTIB	Phase III
		MK-2206	Phase II
		AFURESERTIB	Phase I
GSK3B	0.259	LITHIUM CARBONATE	Phase I	5	2	4
PIK3CA	0.62	BUPARLISIB	Phase II	5	1	4
		SAMOTOLISIB	Phase II
		SONOLISIB	Phase II
		APITOLISIB	Phase I
		DACTOLISIB	Phase I
		COPANLISIB	Phase I
ESR1	0.558	TAMOXIFEN CITRATE	Phase III	5	3	3
		FULVESTRAN	Phase II
		ESTETROL	Phase I
MDM2	0.444	–	–	5	1	4

AKT1 (global score 0.627) is targeted by ipatasertib, capivasertib, MK-2206, and afuresertib, which are currently in Phase I–III clinical stages. It shows strong tractability with a small molecule precedence score of 5, antibody precedence score of 2, and PROTAC score of 5.

PIK3CA, with a global score of 0.620, is targeted by inhibitors like buparlisib, samotolisib, sonolisib, apitolisib, dactolisib, and copanlisib, in clinical phases I and II. It also demonstrates high tractability with a small molecule clinical precedence of 5, antibody clinical precedence of 1, and PROTAC at 4. ESR1 has a global score of 0.558 and is targeted by agonists such as tamoxifen citrate, fulvestrant, and estetrol, spanning clinical phases I to III, with excellent tractability predictions: small molecule clinical precedence at 5, antibody clinical precedence at 3, and PROTAC at 5. ESR2, with a global score of 0.485, is modulated by estramustine phosphate sodium in phase III trials, and also shows high tractability with small molecule clinical precedence at 5, antibody clinical precedence at 3, and PROTAC at 5. These genes, highlighted by their higher global scores, indicate a strong association with diseases, and their diverse and advanced clinical tractability predictions make them promising targets for therapeutic research and development.

Mutation analysis

Understanding the mutation status of therapeutic targets is essential to interpret drug response and resistance mechanisms.

Analysis using the cBioPortal revealed that four genes showed no mutations in the 1417 prostate cancer samples, whereas the remaining six genes exhibited multiple alterations, including frameshift, missense, splice, truncating mutations, structural variants, amplifications, and deep deletions (Fig. 7). Mutation frequencies were as follows: PIK3CA (4%), AKT1 (2%), STAT3 (1.4%), GSK3B (0.8%), MDM2 (0.8%), and ESR1 (0.5%). Since these genes are therapeutic targets in prostate cancer, mutations within them may lead to reduced drug binding affinity and treatment failure.

Fig. 7.

Mutational Profile of Key Prostate Cancer Hub Genes

Therefore, their mutational status highlights potential challenges related to drug resistance, further supporting the need for precision-targeted approaches in prostate cancer therapy.

Gene expression

Analysis using GEPIA revealed significant expression changes among the 10 hub genes in prostate adenocarcinomas (PRAD). AKT1, MMP9, and GSK3B were upregulated, while STAT3, ESR1, ESR2, PIK3CA, SOD2, and PPARA were downregulated compared to normal prostate tissues. MDM2 showed no significant change.

All differentially expressed genes showed a with , confirming their relevance in prostate cancer progression (Fig. 8). Kaplan–Meier analysis indicated that none of the 10 hub genes showed a positive association with overall survival, suggesting that their dysregulation may contribute to poor prognosis in prostate cancer patients.

Fig. 8.

Differential gene expression in prostate cancer compared to normal tissues

Alignment examination

A Venn diagram in Fig. 9 was used to integrate the 10 hub genes across multiple systems biology approaches. Among the four functional prediction tools applied, AKT1 and PIK3CA consistently appeared as highly enriched nodes.

Fig. 9.

Overlap of key prostate cancer hub genes identified by different systems biology approaches

They were significant in KEGG pathway enrichment, text-mining relevance, mutation frequency, and differential expression trends in prostate adenocarcinoma. Due to their recurrence across all layers of analysis, AKT1 and PIK3CA were selected as priority targets for molecular docking and downstream evaluation.

Molecular docking analysis

Molecular docking was performed on both wild-type and mutant forms of AKT1 and PIK3CA using the selected compounds. Table 3 summarizes the binding affinities, while Fig. 10 presents the 3D docking interactions of AKT1 with the ligands.

Fig. 10.

3D structural interactions of mutated and non-mutated AKT1 with ligands

For the non-mutant form of AKT1, Phase I clinical inhibitors listed in Table 2 showed binding affinities between and . Afuresertib (4GV1) showed a strong binding affinity of , with key interacting residues (LEU156A, VAL164A, ALA177A, PHE438A, THR291A). These residues were also involved in the docking of Compounds 01, 02, and 03, indicating similar interaction profiles.

For PIK3CA (wild-type), selected drugs exhibited binding affinities between and . Compound 01 interacted with ILE848A, ILE932A, TYR836A, and ASP933A, similar to dactolisib during early-stage trials. The same residue trends were observed for Compounds 02 and 03. In contrast, copanlisib and apitolisib showed fewer shared interactions, suggesting lower similarity to clinical reference ligands at this stage.

For the mutant form AKT1-E17K (3QKM), the binding energies ranged from to . Afuresertib shared four residues (LEU156A, VAL164A, ALA177A, PHE438A) with Compounds 01–03, indicating preserved pharmacophore alignment.

For mutant PIK3CA-H1047R (4JPS), copanlisib, dactolisib, and apitolisib showed the strongest affinities ( to ). These shared residues (ILE848A, ILE932A, TYR836A, or ILE800A, LEU807A) overlapped with those engaged by Compounds 01–03, confirming structural mimicry.

It is important to note that 3QKM (AKT1-E17K) and 4JPS (PIK3CA-H1047R) represent clinically relevant oncogenic variants linked to androgen independence and drug resistance, making them suitable for resistance-aware docking interpretation. To ensure consistency, all reference inhibitors were predocked under identical parameters, and the residue comparison map is available in Supplementary File S3.

Molecular dynamic simulation analysis

TTClust analysis showed that all ligand–protein complexes sampled a restricted conformational space, with average inter-cluster RMSD values between approximately 1.3 and across all NG settings (full clustering outputs are provided in Supplementary File S4). At NG = 3, the inter-cluster RMSD values were (3QKM_Afuresertib), (3QKM_Compound 03), (4GV1_Afuresertib), and (4GV1_Compound 03). These low values confirm stable binding geometries during the simulation window. Extended simulations (e.g., ) or MM/GBSA calculations may yield deeper energetic insights but are more suitable once equilibrium is fully achieved. By applying TTClust to extract dominant centroid frames and comparing these contacts with those of afuresertib (Table 4), we ensured that only persistent interaction profiles were interpreted. The conserved pharmacophore residues relative to clinical inhibitors (Supplementary Table S6) validate that was sufficient for early-stage stability screening.                

Table 4.

Residue-level benchmarking of AKT1 interactions between reference inhibitor afuresertib and candidate compounds based on TTClust-derived dominant conformations

System	Hydrophobic residues	Hydrogen bonds
afuresertib_4GV1 (Reference)	LEU156, VAL164, ALA177, PHE438, THR291	LEU156 (H-bond), THR291 (H-bond)
3QKM_afuresertib (Reference)	VAL164, ALA177, PHE438	GLU234 (H-bond)
3QKM_Compound03 (C3)	ALA177, ASP292, THR291	THR291 (H-bond)
4GV1_afuresertib (Reference)	LEU156, VAL164, PHE438	GLU234 (H-bond)
4GV1_Compound03 (C2)	VAL164, ALA177	None

Bold residues indicate conserved interactions shared between afuresertib and candidate compounds, highlighting maintenance of pharmacophore-relevant contact residues

Subsequently, MD simulations were used to monitor the interaction profiles of Compound 03 and afuresertib with both non-mutated (Fig. 11) and mutated (Fig. 12) forms of AKT1, offering a dynamic perspective on conformational stability and binding persistence.

Fig. 11.

Molecular dynamics simulation of non-mutated AKT1 interacting with compound 03 and afuresertib

Fig. 12.

Molecular dynamics simulation of mutated AKT1 interacting with compound 03 and afuresertib

Root-mean-square deviation (RMSD) analysis

The RMSD values for the 4GV1_Compound 03 complex exhibited higher fluctuations compared with the 4GV1_afuresertib complex. This suggests that the Compound 03 complex is less structurally stable and experiences greater deviations from the initial binding pose over the simulation time. Similarly, the 3QKM_Compound 03 complex also showed larger RMSD values compared with the 3QKM_afuresertib complex, indicating a less consistent molecular structure for the Compound 03 complex when bound to the mutated protein.

Root-mean-square fluctuation (RMSF) analysis

The RMSF analysis was conducted to assess the flexibility and mobility of the protein–ligand complexes. The 4GV1_Compound 03 complex displayed elevated RMSF values for several key amino acid residues compared to the 4GV1_afuresertib complex. This suggests that the Compound 03 complex exhibits greater flexibility and potential instability in certain regions of the protein structure. A similar trend was observed for the 3QKM_Compound 03 complex, where increased RMSF values were observed for specific residues compared to the 3QKM_afuresertib complex.

Hydrogen bond analysis

The formation and persistence of hydrogen bonds between the ligand and the protein is crucial for maintaining stable binding interactions. The MD simulations revealed that the 4GV1_Compound 03 complex formed fewer hydrogen bonds, and these bonds were less persistent over the simulation time, compared to the 4GV1_afuresertib complex. This indicates a weaker network of hydrogen bonding interactions for the Compound 03 complex. The same pattern was observed for the 3QKM_Compound 03 complex, where fewer and less stable hydrogen bonds were formed compared to the 3QKM_afuresertib complex.

Compactness analysis

The compactness of the protein–ligand complexes was analyzed to assess the quality of the binding interactions. The 4GV1_Compound 03 complex exhibited a less compact molecular structure compared to the 4GV1_afuresertib complex, suggesting potentially less favorable interactions. Similarly, the 3QKM_Compound 03 complex showed a less consistent and compact molecular structure compared to the 3QKM_afuresertib complex.

These findings from the MD simulations complement the insights gained from the earlier docking analysis. The docking study provided an initial assessment of the binding poses, and the detailed MD simulation analysis has now illuminated the dynamic profile of the protein–ligand complexes. Instead of merely reinforcing challenges, this analysis specifies the nature of Compound 03’s binding: it is characterized by higher conformational flexibility rather than the rigid binding seen with afuresertib.

Conclusion

Our comprehensive study highlights the significant potential of bioactive compounds derived from Lepidium sativum in targeting prostate cancer. Through rigorous screening, molecular docking, and dynamic simulations, we identified key compounds that exhibit strong binding affinities and stability with critical prostate cancer targets, notably AKT1 and PIK3CA. Among these, Compound 03 emerged as particularly noteworthy due to its unique binding characteristics. While it demonstrates higher conformational flexibility compared to the rigid binding observed with afuresertib, this dynamic profile provides valuable insights for rational optimization efforts. By strategically designing modifications to reduce flexibility and enhance persistent interactions, future iterations of the Compound 03 scaffold could achieve improved therapeutic potential.

The pharmacokinetic predictions and toxicity assessments further support the plausibility of these compounds as drug-like candidates. However, it is important to emphasize that the current work is based exclusively on computational modeling and in silico validation. As such, the findings should be considered hypothesis-generating rather than confirmatory. Experimental validation—such as biochemical binding assays, cellular cytotoxicity profiling, and functional pathway inhibition studies—is required to confirm the translational relevance of the proposed mechanisms.

Despite this limitation, our integrative computational strategy provides a data-driven foundation to guide future in vitro and in vivo investigations and supports the potential of Lepidium sativum-derived compounds as promising leads for prostate cancer drug discovery.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

S.B. and M.H. conceived and designed the study. S.B. performed the in silico analyses, including drug-likeness and pharmacokinetic evaluations. M.H. conducted the protein–protein interaction network analysis and identified hub genes. S.H. supervised the molecular docking and dynamics simulations. S.B. and M.H. wrote the main manuscript text. S.H. reviewed and edited the final draft. All authors analyzed the results, approved the final version, and agreed to submission.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

No datasets were generated or analysed during the current study.

Code availability

Custom scripts used during the current study are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

The authors declare no Conflict of interest.

Ethics approval and consent to participate

Not applicable. This study does not involve human participants or animal experiments.