Identification of bioactive phytoconstituents as promising ABL2 inhibitors using virtual screening and molecular dynamics simulation Authors: Afnan Elayyan Mousa Elayyan, Abdelbaset Mohamed Elasbali, Marwah Ayed Meshhin Alanazi, Anas Shamsi, Taj Mohammad Published: 2025 ABSTRACT AbstractABL proto-oncogene 2 (ABL2) is an essential nonreceptor tyrosine kinase that regulates cytoskeletal dynamics, cell adhesion, and migration. ABL2 is dysregulated in multiple cancers and represents a promising therapeutic target. However, currently available tyrosine kinase inhibitors are limited by drug resistance and off-target effects. In this study, a structure-based virtual screening approach was performed to identify bioactive phytoconstituents with potential ABL2 inhibitory activity. An initial virtual screening of 11,908 natural compounds from the IMPPAT-2 library was conducted using molecular docking, and the best hits were subsequently filtered based on physicochemical and pharmacokinetic profiles. The top two compounds, Pachyrrhizin and Lupinisoflavone K, showed appreciable binding affinity with appropriate pharmacokinetic properties. Molecular dynamics simulations confirmed stable interactions within the ABL2 binding pocket, supported by hydrogen bonding throughout the simulation, suggesting constant binding of Pachyrrhizin and Lupinisoflavone K with ABL2. The conformational dynamics of these protein-ligand complexes were further confirmed by principal component analysis and free-energy landscape analysis. Overall, the findings suggest that Pachyrrhizin and Lupinisoflavone K may serve as potential ABL2 inhibitors, warranting further in vitro and in vivo validation.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-28195-z.Keywords: Cancer, ABL2 inhibitors, Phytoconstituents, Virtual screening, Molecular dynamics simulationSubject terms: Biochemistry, Cancer, Computational biology and bioinformatics, Drug discovery Introduction Protein kinases are signaling enzymes that regulate key biological processes, such as cell proliferation, differentiation, apoptosis, and migration1. ABL proto-oncogene 2 (ABL2), also known as Abelson-related gene (Arg), is a nonreceptor tyrosine kinase in the Abelson family with structural and functional similarities to ABL12. ABL2 is a central player in both physiological and pathological states, contributing to cytoskeletal rearrangement, cell adhesion, and motility3. While ABL1 has been extensively studied in leukemia, ABL2 was overlooked until recent findings highlighted its role in solid tumors and hematological malignancies, sparking renewed interest in this area4,5. ABL2 dysregulation has been suggested in the oncogenesis of hepatocellular carcinoma, breast cancer, and colorectal cancer6. Aberrant ABL2 signaling promotes tumor progression by increasing cell migration, invasion, and metastasis4. In addition, ABL2 is recognized as more influential in regulating actin cytoskeleton dynamics, which is essential for neurulation, synaptic stability, and the function of immune cells7. Due to its multifaceted role in cancer biology, targeting ABL2 is a promising therapeutic strategy8. This resurgence in interest stems from ABL2’s multidimensional role in cytoskeletal dynamics and oncogenic signaling, making it a compelling therapeutic target. Tyrosine kinase inhibitors (TKIs) have revolutionized the treatment landscape for chronic myeloid leukemia (CML) and many other malignancies, driven by the continuous activity of the kinases. While imatinib and other ATP-competitive inhibitors have shown remarkable clinical success, off-target toxicity, acquired resistance, and limited effectiveness against solid tumors have driven efforts to identify alternative therapeutic approaches9. A common mechanism for resistance to ATP-competitive inhibitors is point mutations in the kinase domain, which can change the binding of the drug and decrease efficacy10. Moreover, long-term consumption of TKIs will result in side effects such as cardiotoxicity, hepatotoxicity, and immune suppression, which will restrict their long-term therapeutic use11. Moreover, secondary mutations, including those in the ABL kinase domain, can render most currently available TKIs ineffective, thereby facilitating disease relapse and progression9. This challenge has prompted ongoing efforts to develop next-generation inhibitors that are more selective and effective while reducing resistance12. To overcome these limitations, new classes of inhibitors that interact with allosteric regulatory sites instead of the ATP-binding pocket have been investigated13. Non-competitive allosteric inhibitors (e.g., ABL-001, GNF-2, and GNF-5) are selective modulators of kinase activity, which have a lower risk of resistance than competing ATP competitors14. These inhibitors present an alternative therapeutic strategy with greater specificity and potentially fewer off-target effects. Yet, while some progress has been made in this field, potent and selective allosteric modulation of ABL2 remains challenging. Allosteric inhibitors, though helpful in overcoming traditional drug resistance mechanisms, remain in early clinical development, with a need for extensive validation in vitro and in vivo as well15. The structural dynamics of ABL2 provide a mechanistic framework for understanding the complex kinome and for orchestrating the rational design of targeted inhibitors with maximal therapeutic benefits, given the inherent complexity of kinase regulation and the balance between ATP-competitive and allosteric sites8. A deeper investigation into new chemical scaffolds available from natural sources is essential to identify next-generation ABL2 inhibitors with better safety and efficacy profiles. Due to their structural diversity, biocompatibility, and multi-targeting potential, bioactive phytochemicals have recently attracted attention as potential drug discovery candidates for treating various diseases16. Historically, natural products have played an important role in anticancer drug discovery, where various phytoconstituents can inhibit kinase activity17. Thus, identifying ABL2 inhibitors from plants may represent an alternative treatment strategy to avoid resistance and limit side effects. Structure-based virtual screening (SBVS) is a well-established computational methodology that enables rapid and accurate prediction of the binding stability and affinity of potential drug leads at the active site of the target protein18. Combined with molecular dynamics (MD) simulations, this method is used on the receptor side to provide molecular-level information on ligand-protein interactions, conformational stability, and binding energetics, aiding rational drug design19. In this study, we used a structure-based virtual screening approach to discover new ABL2 inhibitors from a library of bioactive phytoconstituents. The molecular docking approach was applied to a large library of approximately 11,908 natural compounds, and absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiling was performed to assess drug-likeness. The most promising hits were evaluated in all-atom MD simulations to assess their dynamic stability and binding interactions within the ABL2 kinase domain. Moreover, we performed principal component analysis (PCA) and free energy landscape (FEL) analysis, followed by the molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) analysis, to further understand the conformational changes and binding stability of the elucidated phytochemicals with ABL2. Methods and materials Computing tools and input files An integrated computational workflow based on molecular docking, drug profiling, PASS analysis, and MD simulations was utilized to identify potential ABL2 inhibitors. The molecular docking and interactive analysis used the InstaDock v1.2 tool20. InstaDock is a fast, easy-to-use docking platform designed for high-throughput virtual screening. The ABL2 crystal structure was obtained from the RCSB Protein Data Bank (PDB: 3HMI)21. PDB ID 3HMI was selected for this study owing to its high resolution (1.65 Å) and excellent stereochemical quality, with 98.69% of residues in favored Ramachandran regions (Supplementary Figure S1). Importantly, it contains a co-crystallized inhibitor (K00546) within the ATP-binding pocket, enabling reliable validation of docking accuracy. K00546 and Dasatinib, an FDA-approved dual ABL1/ABL2 inhibitor, were included as a pharmacological reference for comparative docking and ADMET assessment. Preparation of the receptor for virtual screening included adding hydrogen atoms, correcting bond orders, and assigning appropriate atom types. A total of 11,908 bioactive phytochemicals were extracted from the IMPPAT database22 to include structurally diverse natural compounds with possible therapeutic relevance based on a curated library. The IMPPAT 2.0 database was selected as it provides a curated collection of Indian medicinal plant-derived phytochemicals with well-annotated structures, experimental bioactivity data, and structural diversity, ensuring reproducibility and chemical richness during screening. These compounds were initially screened based on Lipinski’s rule of five to exclude compounds that may lack suitable drug-likeness and oral bioavailability. Ligand–receptor complexes were visually assessed using PyMOL23 to confirm that no steric clashes or inappropriate orbital overlaps occurred between the ligands and protein receptor binding sites. GROMACS24 was used for all-atom MD simulations to assess the stability and conformational changes of the ligand-binding protein structures. Molecular docking Molecular docking is a widely used technique in virtual screening to predict interactions between small molecules and target proteins, thereby estimating binding affinities and conformational stability. InstaDock uses an automated, partial flexible docking protocol to search for high-affinity compounds, which was used to perform docking studies. A grid box size of 74 Å × 61 Å × 62 Å was utilized to perform blind docking, centered at the co-crystallized structure (X: −20.703 Å, Y: 11.8 Å, and Z: −10.713 Å) such that the entire binding region was taken into consideration. Although the active site residues of ABL2 are known from the crystal structure, a blind docking approach was adopted to allow ligands to freely explore the entire protein surface and identify potential binding hotspots. This large search grid minimizes the risk of excluding relevant conformations that might otherwise fall outside a predefined pocket. Only the top-ranked poses interacting with the ATP-binding site and catalytic residues were selected for further analysis. Docking calculation was performed using default energy parameters, and top-ranked conformers were picked according to the binding affinity scores. To assess the reliability of the docking protocol, a retrospective redocking was performed using K00546, the co-crystallized ligand of ABL2. The redocked conformation showed strong agreement with the experimentally determined structure, yielding an RMSD of 0.128 Å (Supplementary Figure S2). This close alignment confirms the protocol’s effectiveness in accurately predicting ligand poses within the ABL2 binding site. The co-crystallized ligand K00546 was used solely as a structural benchmark for docking validation. It is not an approved therapeutic, but an experimental inhibitor used in biochemical studies. Dasatinib was thus used as a clinically relevant standard. Using PyMOL and Discovery Studio Visualizer25, ligand-protein interactions were analyzed for the presence of key hydrogen bonds, hydrophobic interactions, and π-π stacking in the ABL2 active site. Compounds were retained for subsequent analysis only if they exhibited key interactions with the ATP-binding pocket. ADMET prediction In silico absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiling was used to assess the pharmacokinetic and toxicological properties of the selected compounds26. Key pharmacokinetic properties, including gastrointestinal (GI) absorption, blood-brain barrier (BBB) permeability, metabolic stability, and renal clearance, were predicted using the pkCSM web server27. Metabolic interactions among the compounds were evaluated by predicting their inhibitory potential against cytochrome P450 (CYP) enzymes because CYP3A4 is one of the main enzymes involved in drug metabolism. Comprehensive CYP450 profiling (CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4) was evaluated to ensure metabolic stability and predict possible drug-drug interactions. Additional ADMET parameters, including all CYP450 isoforms and detailed toxicity indices (AMES, hepatotoxicity, hERG inhibition), are provided in Supplementary Table S1. Moreover, the pan-assay interference compounds (PAINS) filter removed promiscuous molecules that could show non-specific bioactivity or assay interference. The top-ranking compounds with ADMET properties that are good enough for the lead optimization phase, have a low risk of toxicity, and have optimal pharmacokinetic profiles were selected for subsequent studies. PASS analysis The compounds identified via ADMET screening were then predicted for their potential biological activities using the PASS (Prediction of Activity Spectra for Substances) computational tool28. The algorithm, based on machine learning, estimates the probability that a compound will exhibit specific pharmacological activities, using a training set of bioactive molecules. PASS server computes probable activity (Pa) and probable inactivity (Pi) values, which help select compounds with a high probability of causing corresponding therapeutic activities. We evaluated the selected compounds with relevant activity profiles in apoptosis induction, kinase inhibition, and the activation of tumor suppressors for further validation. This predictive method allowed an initial evaluation of the relevance of the compounds for the therapeutic purpose and enabled the quick selection of all the potential drug-like candidates, which were subjected to subsequent MD simulations. MD simulations MD simulations are a valuable tool for gaining a detailed mechanistic understanding of biomolecular interactions, as they mimic the time-dependent behavior of proteins and protein-ligand complexes in a near-physiological environment19,29. MD simulations were used to characterize the overall structural stability, conformational dynamics, and interaction patterns of ABL2 in both the unbound and ligand-bound states. Molecular simulations were carried out with GROMACS 2020 β24 as the molecular simulation package with explicit solvent under 300 K of simulated temperature, while the GROMOS54A7 force field30 was used to simulate ABL2 and its selected ligand complexes, where ligand topologies for the selected ligands were generated using the PRODRG server31 to validate the parametrization for molecular mechanics. Considering that the target system may be subjected to biotreatment events, such as biological absorption, the system was solvated using the SPC216 water model32, which samples a realistic hydration environment. Counterions were included as needed to compensate for the system’s net charge and stabilize its electrostatics. The energy minimization process was performed using the steepest descent algorithm up to 1500 iterations to reach a local energy minimum and remove steric clashes. After the energy minimization, the system relaxed in two equilibration steps: firstly, in a constant number of particles, volume, and temperature (NVT ensemble), and then in a constant number of particles, pressure, and temperature (NPT ensemble). These steps ensured the system reached thermodynamic stability before the production phase began. The final MD simulation was performed for 500 nanoseconds (ns) to follow the evolution and stability of the complexes over time. The time evolution of stability and compactness of the ABL2-ligand complexes was analyzed using simulation trajectories based on root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), solvent-accessibility surface area (SASA), intramolecular and intermolecular hydrogen bonding, and secondary structure composition were computed to assess dynamic stability. Principal component analysis To understand the internal motions and dynamic nature of ABL2, we performed principal component analysis (PCA) on simulated trajectories. The PCA is a classical method for dimensionality reduction that provides dominant motions by analyzing the covariance matrix of atomic fluctuations. This approach enables detection of functionally relevant conformational changes that underlie the ligand’s binding and the protein’s stability33. Eigenvectors and eigenvalues computed from the covariance matrix provided the most principal motion vectors of ABL2 and complexes, showing the essential motions. Free energy landscapes Free energy landscapes (FEL) analysis was subsequently performed to assess protein stability across conformational states34. The FEL was built for each system by projecting the major components in the two-dimensional coordinate and energy space, in which the global minima correspond to the most stable conformations. FELs based on the trajectory data of ABL2 in the unbound form and complexed with the selected compounds were constructed to reveal the metastable states and transition pathways. This study quantitatively elucidated the effects of ligand binding on the conformational dynamics and energy landscape of ABL2, thereby providing necessary information on the drug-like properties of selected phytochemicals. MM-PBSA calculations Binding free energy calculations were carried out using the MM-PBSA approach implemented through the gmx_mmpbsa tool35, which combines functionalities of GROMACS and APBS. A 10-nanosecond segment was selected from the equilibrated region of the MD trajectory, and frames were extracted at 10 picoseconds intervals for analysis. The binding energy was estimated by summing contributions from van der Waals interactions, electrostatic forces, polar solvation, and solvent-accessible surface area (SASA). The overall binding free energy (ΔGbinding) was determined using the following relation: = – ( + ). Here, each term corresponds to the free energy of the respective system component. Additionally, per-residue energy contributions were computed using the Python utilities provided with the gmx_mmpbsa package. Results and discussion Molecular docking A total of ~ 18,000 phytochemicals were extracted from the IMPPAT database22, screened based on their physicochemical properties, and resulted in a final dataset of 11,908 compounds, which were then screened using InstaDock for molecular docking. Virtual screening based on the molecular docking approach was performed to select bioactive phytoconstituents with high binding affinity towards ABL2. The binding affinity profile of the compounds was prioritized in the docking analysis by assigning higher ranks to those with stronger binding potential within the ATP-binding site of ABL2. The top 10 selected compounds showed binding energies ranging from − 10.4 to − 11.3 kcal/mol (Table 1). Ligand efficiency (kcal/mol/non-H atom) was calculated to normalize binding affinity by molecular size, ensuring prioritized hits exhibit optimal binding per heavy atom. The higher binding affinities and ligand efficiency represented by these values, compared to the standard reference inhibitors K00546 (− 8.8 kcal/mol) and Dasatinib (− 9 kcal/mol), indicate the potential of these phytochemicals as promising leads for therapeutic development. These compounds, whose docking scores indicate potential competitive binding of ATP within the kinase domain, could inhibit ABL2 activity. Table 1Top 10 phytochemical hits identified through molecular Docking against ABL2. The table presents the Docking scores of the selected compounds, ranked based on their binding affinity (kcal/mol). Higher negative scores indicate stronger binding interactions within the ATP-binding site of ABL2. Ligand efficiency values are in kcal/mol/non-H atom.S. no.IMPPAT IDPhytoconstituentPlant sourceAffinity (kcal/mol)Ligand efficiency1.IMPHY008069Erythrinin A Erythrina variegata −11.30.472.IMPHY007679Bismurrayaquinone A Murraya koenigii −11.10.353.IMPHY009376Glabridin Glycyrrhiza glabra −11.10.464.IMPHY009292Pachyrrhizin Oxytropis kansuensis −10.70.435.IMPHY009067Picrasidine N Picrasma quassioides −10.70.296.IMPHY003058Rothindin Rothia indica −10.60.337.IMPHY014146Bianthraquinone Senna hirsuta −10.50.338.IMPHY000151Glabroisoflavanone A Glycyrrhiza glabra −10.50.429.IMPHY002792Lupinisoflavone K Lupinus albus −10.50.3310.IMPHY006882Candidine Strobilanthes cusia −10.40.3711.K00546−8.80.3112.Dasatinib−90.27 Top 10 phytochemical hits identified through molecular Docking against ABL2. The table presents the Docking scores of the selected compounds, ranked based on their binding affinity (kcal/mol). Higher negative scores indicate stronger binding interactions within the ATP-binding site of ABL2. Ligand efficiency values are in kcal/mol/non-H atom. ADMET properties The drug-likeness and clinical potential of candidate molecules depend on their pharmacokinetics and toxicity. The top 10 compounds selected from the docking screening were subjected to ADMET evaluation using the pkCSM web server27. To further refine the selection process, a PAINS filter was employed to remove non-specific binders and promiscuous molecules likely to lead to false-positive results. The ADMET profiles of the screened compounds are shown in Table 2. In the ADMET analysis, two of the ten screened compounds showed good pharmacokinetic properties without any toxic patterns. These two compounds, Pachyrrhizin and Lupinisoflavone K, are predicted to possess high GI absorption, BBB permeability, and showed minimal interaction with cytochrome P450 enzymes, particularly CYP3A4, minimizing the likelihood of metabolic instability and low renal clearance36. The screened compounds retained their ADMET properties throughout this screening step without inhibitory potential towards organic cation transporter 2 (OCT2), validating them as candidates suitable for further evaluations. The results demonstrate that the identified phytochemicals offer high binding affinities and desirable pharmacokinetic properties, which can be exploited for subsequent analyses. Although Erythrinin A, Bismurrayaquinone A, and Glabridin displayed slightly better docking affinities, their ADMET analysis revealed unfavorable pharmacokinetic and toxicity profiles (e.g., CYP substrate or predicted hepatotoxicity). In contrast, Pachyrrhizin and Lupinisoflavone K exhibited optimal pharmacokinetics and safety, justifying their prioritization despite marginally lower docking energies. While docking scores reflect binding affinity, overall drug potential depends on a balance between affinity and pharmacokinetic safety. The slightly lower docking energies of Pachyrrhizin and Lupinisoflavone K are compensated by their superior ADMET and toxicity profiles, suggesting better drug-likeness and clinical suitability. Table 2Pharmacokinetic and toxicity parameters (ADMET properties) of the selected ABL2 inhibitors. The table includes key descriptors such as Gastrointestinal (GI) absorption, blood-brain barrier (BBB) permeability, cytochrome P450 (CYP3A4) inhibition, and toxicity predictions, which determine the drug-likeness and safety of the compounds.S. no.PhytoconstituentAbsorptionDistributionMetabolismExcretionToxicityGI absorptionBBB permeationCYP 3A4 substrateOCT2 inhibitorAMES/Herg/Hepato.1.Erythrinin AAbsorbedPenetrableNon-substrateNon-inhibitorToxic2.Bismurrayaquinone AAbsorbedNon-penetrableSubstrateNon-inhibitorToxic3.GlabridinAbsorbedPenetrableSubstrateNon-inhibitorToxic4.PachyrrhizinAbsorbedPenetrableNon-substrateNon-inhibitorSafe5.Picrasidine NAbsorbedPenetrableNon-substrateInhibitorToxic6.RothindinAbsorbedPenetrableSubstrateNon-inhibitorToxic7.BianthraquinoneAbsorbedPenetrableNon-substrateNon-inhibitorToxic8.Glabroisoflavanone AAbsorbedPenetrableSubstrateNon-inhibitorToxic9.Lupinisoflavone KAbsorbedPenetrableNon-substrateNon-inhibitorSafe10.CandidineAbsorbedPenetrableNon-substrateInhibitorSafe11.K00546AbsorbedPenetrableNon-substrateNon-inhibitorSafe12.DasatinibAbsorbedPenetrableSubstrateInhibitorToxic Pharmacokinetic and toxicity parameters (ADMET properties) of the selected ABL2 inhibitors. The table includes key descriptors such as Gastrointestinal (GI) absorption, blood-brain barrier (BBB) permeability, cytochrome P450 (CYP3A4) inhibition, and toxicity predictions, which determine the drug-likeness and safety of the compounds. PASS analysis PASS server is a widely explored resource for predicting molecular properties and calculating probability scores for different versions of pharmacological activity28. PASS server predicts the probability of a compound being able to manifest a variety of specific biological functions. As such, it effectively identifies compounds with therapeutic potential for anticancer activity, apoptosis induction, and tumor suppressor modulation. Evaluation of the screened candidates, Pachyrrhizin and Lupinisoflavone K, showed appropriate drug profiles and relevant anticancer properties. Table 3 shows the predicted biological activities of Pachyrrhizin and Lupinisoflavone K. The expected activity (Pa) values for these compounds are associated with their predicted activity (Pa) and probability of inactivity (Pi) values (0.591 to 0.887, Pa > Pi). This implies that Pachyrrhizin and Lupinisoflavone K have anticancer activities, potentially relevant to ABL2 inhibition. Pachyrrhizin and Lupinisoflavone K alterations in cellular pathways associated with apoptosis and cell cycle regulation were supported by PASS analysis results and may contribute to tumor suppression. These data reinforce the therapeutic value of these agents for malignancies driven by ABL2. Table 3Biological activity predictions of the selected hits (Pachyrrhizin and Lupinisoflavone K) using the PASS server. The table lists the probability of activity (Pa) and probability of inactivity (Pi) for each predicted function, including antineoplastic effects, TP53 expression enhancement, and apoptosis induction. Compounds with Pa > Pi are considered likely to exhibit the predicted activity.CompoundStructurePaPiBiological activityPachyrrhizin 0.7030.006Antimutagenic0.6690.009MMP9 expression inhibitor0.6550.034Antineoplastic0.6170.024Apoptosis agonist0.5910.017Histidine kinase inhibitorLupinisoflavone K 0.8870.002Histidine kinase inhibitor0.7740.015Antineoplastic0.7040.025TP53 expression enhancer0.6940.015Apoptosis agonist0.6090.015MMP9 expression inhibitorK00546 0.6710.003Cyclin-dependent kinase 1 inhibitor0.4540.003AXL receptor tyrosine kinase inhibitor0.4340.005Cyclin-dependent kinase inhibitor0.4080.035Angiogenesis inhibitor0.3110.030Tyrosine kinase inhibitorDasatinib 0.8250.004Tyrosine kinase inhibitor0.8020.002Lck kinase inhibitor0.7940.003Src kinase inhibitor0.4300.002Bcr-Abl kinase inhibitor0.3350.056Angiogenesis inhibitor Biological activity predictions of the selected hits (Pachyrrhizin and Lupinisoflavone K) using the PASS server. The table lists the probability of activity (Pa) and probability of inactivity (Pi) for each predicted function, including antineoplastic effects, TP53 expression enhancement, and apoptosis induction. Compounds with Pa > Pi are considered likely to exhibit the predicted activity. Interaction analysis To evaluate the binding potential of Pachyrrhizin and Lupinisoflavone K to inhibit ABL2, molecular interaction analysis was performed to reveal critical binding interactions in the kinase domain. Interaction analyses were performed to evaluate the binding modes of both Pachyrrhizin and Lupinisoflavone K compared to the reference inhibitor, K00546. This analysis revealed key non-covalent interactions that contribute to complex stability, hydrogen bonding, hydrophobic interactions, and π-π stacking in each ligand-protein complex. Figure 1 represents ABL2 bound with Pachyrrhizin, Lupinisoflavone K, and K00546 in the kinase domain. Analysis showed that Pachyrrhizin and Lupinisoflavone K are bound in the ATP-binding site and superimposed to the reference inhibitor, K00546. Pachyrrhizin bound in the ATP-binding site and showed close interactions such as Lys317, Thr361, Met364, and Arg413 (Fig. 1B). At the same time, Lupinisoflavone K showed direct interactions with the ATP-binding site and a close hydrogen bond interaction with the ABL2 active site, i.e., Asp409 (Fig. 1C). This binding mode is consistent with that observed for K00546, supporting that Pachyrrhizin and Lupinisoflavone K are likely competitive inhibitors that prevent ATP access and disrupt kinase activity (Fig. 1D). The docked Pachyrrhizin and Lupinisoflavone K overlays onto K00546 showed clear overlap within the ATP-binding pocket, indicating them as potential ABL2 inhibitors. Multiple stabilizing interactions suggest a potent and specific ligand-protein association, further reinforcing their candidacy as potential ABL2-targeted therapeutic agents. These findings provide a molecular basis for the inhibitory action of these natural compounds and support their further evaluation. Structural representation of ABL2 in complex with Pachyrrhizin, Lupinisoflavone K, and the reference inhibitor K00546. (A) A magnified cartoon representation of ABL2 with bound ligands, highlighting their binding orientation. (B) Close-up view of the ABL2 binding pocket occupied by Pachyrrhizin. (C) Close-up view of the binding pocket occupied by Lupinisoflavone K. (D) Co-crystallized inhibitor K00546 bound to the ATP-binding pocket of ABL2, serving as a reference for interaction analysis. The two-dimensional plots of ligand-protein interactions of Pachyrrhizin and Lupinisoflavone K with ABL2 confirmed that both molecules established strong hydrogen bonds and hydrophobic contacts in the ATP-binding site (Fig. 2). These interactions are the main determinants affecting their binding and stability in the kinase domain. The interaction profile of Pachyrrhizin reveals important hydrogen bonds and hydrophobic contacts (Fig. 2A). Pachyrrhizin showed three hydrogen bonds with Tyr299, Thr361, and Met364 along with three Pi-sigma interactions with Leu294, Leu416, and Ala426, and various other Alkyl bonds and van der Waals interactions. Similarly, Lupinisoflavone K exhibits a comparable interaction pattern with key hydrogen bonding and hydrophobic interactions (Fig. 2B). It showed one hydrogen bond with Asn414, one Pi-anion with the active site Asp409, and various other Alkyl/Pi-alkyl bonds and van der Waals interactions. As a reference inhibitor to test the efficacy of the screened phytochemicals, Fig. 2C describes the binding mode for K00546 with several similar interactions. This standard complex showed interaction profiles similar to those of the selected phytochemicals, reinforcing their comparable binding stability. Overall, the binding analysis suggests that Pachyrrhizin and Lupinisoflavone K could thus potentially prevent the binding of ATP and inhibition of ABL2, blocking associated downstream signaling pathways. These results justify further engagement in MD simulations and wet laboratory assays to validate these compounds’ stability and inhibitory potency. Two-dimensional interaction analysis of ABL2 with selected ligands. (A) The interaction profile of ABL2 with Pachyrrhizin shows key hydrogen bonding and hydrophobic interactions. (B) Interaction profile of ABL2 with Lupinisoflavone K. (C) Interaction profile of ABL2 with K00546, illustrating its binding mode for comparison with the selected phytochemicals. MD simulations All-atom MD simulations were performed to explore the stability of ABL2 in complex with Pachyrrhizin and Lupinisoflavone K and their dynamic behavior within the ABL2 binding site. All-atom MD simulations (500 ns) were performed for ABL2 complexes with Pachyrrhizin and Lupinisoflavone K to evaluate ligand-induced structural changes and protein-ligand stability. MD analysis showed the significant stability of both phytoconstituents within the ABL2 binding site compared to the system with no ligand bound. For the calculation of the key structural parameters, root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), and solvent-accessible surface area (SASA), key structural parameters were analyzed to illustrate the ABL2 conformational stability following Pachyrrhizin and Lupinisoflavone K binding. Structural deviations RMSD is one of the basic parameters used to assess the overall stability of a protein-ligand complex throughout a simulation37. It tracks the backbone atoms of a protein over time and provides insights into the conformational changes of the ligand within the protein binding pocket. The RMSD values of ABL2 bound to Pachyrrhizin and Lupinisoflavone K were derived and illustrated for time (Fig. 3A). During the 500 ns trajectory, the ABL2-Lupinisoflavone K complex showed greater structural stability, fluctuating very little compared to the ABL2-Pachyrrhizin system. However, with the ABL2-Pachyrrhizin complex, only minor deviations were noted, mainly within the 300–400 ns window frame, and seem to stabilize after that without significant conformational changes. This suggests that the ABL2–Lupinisoflavone K complex was comparatively more stable than those with Pachyrrhizin or K00546. The probability distribution function (PDF) of RMSD supported the fact that the bound states of Lupinisoflavone K maintained similar overall stability as the unbound ABL2 structure (Fig. 3A, lower panel). Structural dynamics of ABL2 upon binding to Pachyrrhizin and Lupinisoflavone K. (A) Root-mean-square deviation (RMSD) plot showing the stability of the ABL2 backbone over the 500 ns simulation for both ligand-bound complexes. The lower panel displays the probability distribution function (PDF), providing a statistical overview of RMSD deviations. (B) Root-mean-square fluctuation (RMSF) plot representing the flexibility of individual residues in ABL2 before and after ligand binding. The lower panel displays the PDF, providing a statistical overview of RMSF variations. To investigate the localized structural perturbations, RMSF analysis was carried out, which describes the flexibility of each amino acid residue during the MD simulation. RMSF reveals the magnitude of conformational changes that different protein regions undergo upon ligand binding, with higher fluctuation and flexibility corresponding to higher RMSF values. From the RMSF profile, the binding of Pachyrrhizin and Lupinisoflavone K only slightly moved the associated residues of key interacting residues, mainly in the ATP-binding pocket of ABL2 (Fig. 3B). The analysis showed some fluctuations in surface-exposed loops and flexible regions that are inherently dynamic within unbound proteins. Residues forming hydrogen bonds and hydrophobic contacts with Pachyrrhizin and Lupinisoflavone K exhibited minor fluctuations, suggesting that these ligands bind firmly and stably. Notably, residues Lys317–Gly325, Cys376–Thr382, Ala426–Pro448, and Tyr481–Asp490 showed moderate fluctuations corresponding to loop and surface-exposed regions, whereas core catalytic residues remained stable, confirming ligand-induced stabilization. The overall RMSF pattern of ABL2-Pachyrrhizin and ABL2-Lupinisoflavone K complexes was comparable to free ABL2, indicating that complexation with respective ligands did not cause significant residual fluctuations. These fluctuations are reasonably acceptable, further strengthening the notion that Pachyrrhizin and Lupinisoflavone K did not destabilize the native fold of ABL2. Moreover, we found that essential regions for kinase function, including the catalytic domain, were structurally preserved in both ligand-bound conformations, providing additional evidence that Pachyrrhizin and Lupinisoflavone K remain stable without any significant conformational perturbations. To examine the overall compactness of the ABL2 structure in both free and ligand-bound forms, the time evolution of the Rg was computed from the simulated trajectory. Rg is an important parameter as it characterizes the distribution of atomic mass about the center of mass of the protein and can provide information on protein folding38. Significantly large deviations from a stable value of the Rg during simulation indicate unfolding, conformational changes, or destabilization over a long period away from the native structure. The Rg analysis suggested that the conformation of ABL2 remained stable while bound to Pachyrrhizin and Lupinisoflavone K, with respective values of 1.90–2.05 nm over the 500 ns simulation (Fig. 4A). These values were similar to those seen with free ABL2, suggesting that the binding of the ligand does not induce a significant expansion or collapse of the overall tertiary structure of the protein. This was also supported by the PDF analysis of the Rg for both ligand-bound systems, which were very similar to that of free ABL2, suggesting that Pachyrrhizin and Lupinisoflavone K binding did not cause extensive structural perturbation (Fig. 4A, lower panel). The minor variations seen in Rg further demonstrate that ABL2 remained in its compact and functional conformation, with Pachyrrhizin and Lupinisoflavone K making them good candidates for inhibitor development. Compactness analysis of ABL2 upon ligand binding. (A) Radius of gyration (Rg) plot depicting changes in structural compactness for free and ligand-bound ABL2. (B) Solvent-accessible surface area (SASA) plot, illustrating alterations in protein-solvent interactions following ligand binding. The lower panels show PDF distributions, providing further insights into the stability of the systems. SASA assesses the extent of protein surface exposure to solvents, providing information about protein stability and conformational flexibility39. SASA measures the extent to which the protein is exposed to surrounding solvent molecules, indicating possible changes in the compactness of structure (and exposure to solvent) when a ligand is bound (or not). The SASA profile shows that ABL2 in complex with Pachyrrhizin and Lupinisoflavone K retained a stable solvent exposure throughout the simulation, as there were no significant deviations from the unbound protein (Fig. 4B). Stable SASA values likely reflect that ligand binding does not result in more solvent exposure, which could indicate unfolding or destabilization of ABL2. Nevertheless, PDF analysis indicates that SASA is slightly increased in the ligand-bound systems (Fig. 4B, lower panel). This slight increase suggests that some ligand accommodation in the ATP-binding site triggered local conformational changes with only a slight alteration in the accessibility of residues to solvent. This is an expected change and is considered an artifact of protein adaptation to ligand binding, rather than destabilization. The stable Rg and SASA values affirm that ligand binding does not disrupt the global folding of ABL2, therefore further indicating the ability of these two phytochemicals to be stable ABL2 binders. Dynamics of hydrogen bonds Hydrogen bonds play an essential role in the binding affinity of ligands, conformational changes of proteins, and overall molecular stability40. Hydrogen bonding is the second most important force that stabilizes the protein-ligand interaction and retains the integrity of the complex. We investigated both intramolecular (within ABL2) and intermolecular (between ABL2 and ligands) hydrogen bond dynamics during the 500 ns MD simulation (Fig. 5). The time evolution of intramolecular hydrogen bonds in ABL2 in its complexes with Pachyrrhizin and Lupinisoflavone K showed that the formation of hydrogen bonds was relatively stable, in that it only fluctuated slightly during the simulation (Fig. 5A). Furthermore, the hydrogen bonding PDF analysis also confirmed that ABL2 retained a stable hydrogen bonding network during the simulation (Fig. 5B). The consistent peaks in the PDF plots reflect well-maintained hydrogen bonding geometries and persistent interactions within the protein and between the protein and ligands. These results affirm that both Pachyrrhizin and Lupinisoflavone K formed energetically favorable interactions with ABL2, contributing to stabilizing the binding complex. Collectively, these findings underline the crucial role of hydrogen bonding in the molecular recognition and stabilization of ABL2-ligand complexes. Intramolecular hydrogen bond analysis during the molecular dynamics simulation. (A) The time evolution of intramolecular hydrogen bonds within ABL2 shows their stability in free and ligand-bound states. (B) Probability distribution function (PDF) of the intramolecular hydrogen bonds, providing a quantitative measure of hydrogen bond fluctuations. The intermolecular hydrogen bonding between ABL2 and the identified compounds, Pachyrrhizin and Lupinisoflavone K, was analyzed to determine the degree of stabilization and time-mediated dynamics (Fig. 6). These interactions were essential for keeping Pachyrrhizin and Lupinisoflavone K in the ATP-binding site. The analysis showed that Pachyrrhizin presented more stable hydrogen bonding interactions with ABL2 than Lupinisoflavone K, according to higher PDF values and more frequent hydrogen bonding occurrences along all the trajectory (Fig. 6, lower panels). These stable hydrogen bonds indicate that Pachyrrhizin and Lupinisoflavone K may be promising binders of ABL2 and can be exploited in therapeutic development. This stability was mirrored by the hydrogen bonding patterns, which also suggest consistent bond formation between Pachyrrhizin, Lupinisoflavone K, and the ABL2 active site with minimal detachment, thus strengthening and further corroborating the conclusion that both compounds act as stable ABL2 binders. The concluded results underscore the hydrogen bonding nature of protein-ligand interaction stability, which could be that the hydrogen bond network found to be stable in the ABL2 binding pocket can successfully obstruct the ATP from binding to the ABL2, and also that Pachyrrhizin and Lupinisoflavone K may function as promising drug candidates. More computational and experimental validations are needed to evaluate the inhibition activity of these compounds. Intermolecular hydrogen bonding between ABL2 and identified phytochemicals. (A) Time evolution of hydrogen bonds between ABL2 and Pachyrrhizin. (B) Time evolution of hydrogen bonds between ABL2 and Lupinisoflavone K. (C) Time evolution of hydrogen bonds between ABL2 and K00546. Secondary structure dynamics of ABL2 The function and stability of a protein are, to some extent, influenced by its secondary structure composition. Ligand binding can modulate the stability of these structural elements by inducing subtle conformational changes. We analyzed the conformational change of ABL2 in the 500 ns MD simulation by studying the evolution of secondary structure elements to find representative conformations of ABL2 under the influences of Pachyrrhizin and Lupinisoflavone K. This analysis allowed us to determine whether ligand binding resulted in a deviation from the native folded conformation of ABL2 and impacted its stability. The study showed that the secondary structure content of ABL2 remained stable in its bound and unbound states (Fig. 7). Binding of Pachyrrhizin and Lupinisoflavone K showed neither drastic loss of structural integrity, as both ligands preserve the overall conformational stability of ABL2. We observed a stable β-sheet content and increased turns, indicating a local stabilization effect due to the binding of the ligand (Table 4). The conversion of flexible features of the structural elements into less flexible secondary structures may also improve the rigidity of ABL2-ligand complexes. It may also contribute to the stability of the complexes (Fig. 7A-D). Accordingly, these results indicate that binding of Pachyrrhizin and Lupinisoflavone K does not drastically perturb the secondary structure of ABL2, supporting the notion that these phytochemicals act as stable binders without perturbing the native folding of the protein. Ligand-induced structural stability of ABL2 reinforces the candidacy of Pachyrrhizin and Lupinisoflavone K as promising candidates. Secondary structure composition of ABL2 before and after ligand binding. (A) Secondary structure distribution in free ABL2. (B) Secondary structure changes upon Pachyrrhizin binding. (C) Secondary structure modifications upon Lupinisoflavone K binding. (D) Secondary structure modifications upon K00546 binding. Structural elements include α-helices, β-sheets, β-bridges, and turns, with minor changes observed upon ligand binding. Table 4Residues involved in the secondary structure formation of ABL2 in its free state and upon binding to pachyrrhizin and Lupinisoflavone K. The table summarizes the composition of different secondary structural elements, including α-helices, β-sheets, β-bridges, turns, and coils, highlighting any ligand-induced structural changes.Secondary structure elementABL2ABL2-PachyrrhizinABL2-Lupinisoflavone KABL2-K00546Coil57545845β-sheet44434444β-bridge3533Bend24232425Turn29333137α-helix91919094π-helix0000310-helix1012910-Helix107910 Residues involved in the secondary structure formation of ABL2 in its free state and upon binding to pachyrrhizin and Lupinisoflavone K. The table summarizes the composition of different secondary structural elements, including α-helices, β-sheets, β-bridges, turns, and coils, highlighting any ligand-induced structural changes. Principal component analysis Protein dynamics and conformational flexibility are among the key aspects in drug discovery, as they significantly impact ligand binding and protein function33,41. Here, we used PCA to determine the dominant motions of ABL2 and to evaluate whether binding the ligands altered its conformational landscape (Fig. 8). PCA was utilized to find collective motions in the protein based on the eigenvectors of the covariance matrix computed from MD trajectories. PCA enables the reduction of complex, high-dimensional atomic motion into a few principal components (PCs) by calculating the eigenvectors and eigenvalues of the covariance matrix derived from the positional fluctuations of the Cα atoms over the trajectory. These principal components correspond to collective motions, which are often functionally relevant. The PCA 2D projections of ABL2, ABL2-Pachyrrhizin, ABL2-Lupinisoflavone K, and K00546 showed that all the systems resided in a similar conformational subspace (Fig. 8A). The time-evolution of the PCA trajectories of both EVs also indicates that the binding of Pachyrrhizin and Lupinisoflavone K did not cause significant conformational changes but stabilized ABL2 in its native dynamics range (Fig. 8B). The similar clustering of the free and ligand-bound states provides further evidence that these compounds bind to ABL2 without disrupting its inherent flexibility. Collectively, these PCA results reinforce the notion that the selected compounds exhibit favorable binding profiles with minimal disturbance to ABL2’s global dynamics, thereby supporting their potential as effective modulators with minimal conformational liabilities. Principal component analysis (PCA) of ABL2 and its ligand-bound complexes. (A) The 2D conformational sampling plots illustrate the dominant motion trajectories of free ABL2, ABL2-Pachyrrhizin, ABL2-Lupinisoflavone K, and ABL2-K00546. (B) The overlapping subspaces indicate that ligand binding does not induce drastic conformational changes but rather stabilizes ABL2 in specific conformational states. Free energy landscapes We carried out FEL analysis on the PCA trajectories to get deeper insights into the ABL2 energetic folding landscape and its ligand-bound states. FEL gives insight into the conformational states of a protein from a thermodynamic perspective through a projection of the free energy minimum that represents the most stable states34,42. FEL plots of ABL2, ABL2-Pachyrrhizin, ABL2-Lupinisoflavone K, and ABL2-K00546 are shown in Fig. 9, with the dark blue regions indicating the lowest free energy conformational states. In the ABL2 free state, two prominent global minima indicated a well-defined conformational landscape (Fig. 9A). Ligand binding changed the energy landscape to contain more than one or two global minima, as was the case for the ABL2-bound complexes of Pachyrrhizin and Lupinisoflavone K (Fig. 9B-C). The ABL2-K00546 also showed two prominent global minima, indicating a well-defined conformational landscape (Fig. 9D). This implies that Pachyrrhizin and Lupinisoflavone K binding can localize stabilization without disrupting the global conformational equilibrium of ABL2. Multiple energy minima in the ligand-bound states could indicate small conformational changes induced by the ligands to fit into the ATP-binding pocket. In summary, PCA and FEL analyses show that Pachyrrhizin and Lupinisoflavone K interact with ABL2, preserving its native structural features while contrasting large-scale conformational shifts. The PCA and FEL analyses collectively suggest that ligand binding stabilizes ABL2 in its native conformational ensemble, minimizing off-target effects, a critical feature for selective kinase inhibitors. These results suggest that these phytoconstituents are stable ABL2 inhibitors, which can be further explored to validate their biological activity in vitro and in vivo. Free energy landscape (FEL) analysis of ABL2 and its ligand-bound states. (A) FEL plot of free ABL2. (B) FEL plot of ABL2-Pachyrrhizin complex. (C) FEL plot of ABL2-Lupinisoflavone K complex. (C) FEL plot of ABL2-K00546 complex, suggesting localized conformational adjustments upon ligand interaction. MM-PBSA analysis To assess the binding affinities of ABL2 in complex with Pachyrrhizin, Lupinisoflavone K, and K00546, we utilized MM-PBSA analysis. This method estimates the thermodynamic favorability of ligand–protein interactions, where more negative binding free energy values indicate stronger and more stable associations43. The total binding energy was calculated by combining contributions from electrostatic and van der Waals interactions, polar solution, and SASA energies. Table 5 summarizes each complex’s average binding free energies and standard deviations. The findings indicate that all energy components played a role in stabilizing the interactions between ABL2 and the respective ligands. Both ligands exhibited strong van der Waals and electrostatic interactions, which contributed significantly to the overall stabilization of the complexes. Notably, unfavorable as expected due to desolvation effects, favorable non-polar contributions, and hydrophobic interactions counterbalanced the polar solvation energies. The relatively low standard deviations suggest that the ligand-protein complexes remained stable throughout the sampled trajectory frames, further affirming the robustness of the predicted binding affinities. These MM-PBSA results are consistent with the earlier dynamic analyses, supporting that Pachyrrhizin and Lupinisoflavone K can form stable and energetically favorable interactions with ABL2, thus underscoring their potential as promising inhibitors. Table 5Estimated binding free energy values for ABL2–ligand complexes calculated using the MM-PBSA method.ComplexΔGVDWAALSΔEELΔEPBΔENPOLARΔGGASΔGSOLV ABL1–Pachyrrhizin–31.92–16.5329.91–3.51–48.4526.40–22.05 ± 2.32ABL1–Lupinisoflavone–25.31–12.9925.83–2.87–38.3022.96–15.35 ± 3.00ABL1–K00546–26.53–22.9933.91–3.22–49.5230.69− 18.83 ± 4.46 Estimated binding free energy values for ABL2–ligand complexes calculated using the MM-PBSA method. Conclusions ABL2 is a critical oncogenic driver and represents a promising therapeutic target in cancer. Nevertheless, selective and potent inhibitors of ABL2 are still lacking, primarily due to resistance mechanisms and the limitations of current small-molecule inhibitors. This study used a structure-based virtual screening approach to screen bioactive phytochemicals that could act as ABL2 inhibitors. Based on molecular docking, ADMET profiling, and MD simulations, two phytochemicals, Pachyrrhizin and Lupinisoflavone K, were identified. Both remained bound to the ATP-binding pocket and the active site of ABL2, maintaining their structure, and exhibited good pharmacokinetics. To further corroborate their ability as stable binders, we performed essential dynamics and MM-PBSA analyses, which suggest appreciable stability of the protein-ligand complexes. Although these findings are computational, future in vitro and in vivo experiments are needed to validate their inhibitory potential. Future work should include experimental validation, structural optimization for improved selectivity, and in vivo evaluations of therapeutic efficacy. This study suggests an opportunity to further explore phytochemicals as ABL2 inhibitors and may also provide basic insights for next-generation drug development. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements This work was funded by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No. (DGSSR-2025-FC-01023). AS acknowledges Ajman University for supporting the publication. Author contributions A.E.M.E.: Conceptualization, Methodology, Data curation, Writing – review & editing.A.M.E.: Conceptualization, Funding acquisition, Formal analysis, Validation, Writing – original draft.M.A.M.A.: Formal analysis, Validation, Writing – original draft.A.S.: Formal analysis, Validation, Funding acquisition, Writing – original draft.T.M.: Methodology, Validation, Writing – original draft. Funding This work was funded by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No. (DGSSR-2025-FC-01023). Data availability All data generated or analyzed during this study are included in this manuscript. All raw data is available on request to the corresponding author. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Abdelbaset Mohamed Elasbali, Email: aeelasbali@ju.edu.sa. Anas Shamsi, Email: anas.shamsi18@gmail.com. Taj Mohammad, Email: taj144796@st.jmi.ac.in. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement All data generated or analyzed during this study are included in this manuscript. All raw data is available on request to the corresponding author.