Molecular screening of natural compounds targeting KRAS(G12C): a multi-parametric strategy against acute lymphoblastic leukemia

Abstract

Acute lymphoblastic leukaemia (ALL) is a highly aggressive hematological malignancy that necessitates safer, more effective therapies. This study applied a multi-parametric computational approach to identify KRAS (G12C) inhibitors from African natural product databases. Six lead compounds (NA/EA-1 to NA/EA-6) were identified via virtual screening, molecular docking, and induced-fit docking, all showing stronger binding affinities (−14.50 to −10.53 kcal/mol) than the reference inhibitor Sotorasib (−8.34 kcal/mol). These candidates exhibited favorable pharmacokinetic and physicochemical properties, minimal Lipinski’s rule violations, and non-toxic ADMET profiles. Four top hits were subjected to 200 ns molecular dynamics simulations, with NA/EA-3 demonstrating the greatest stability, lowest RMSD, and strongest hydrogen bonding. MM/GBSA analysis confirmed NA/EA-3 as the most promising compound (ΔG_total −54.42 kcal/mol), outperforming Sotorasib (−32.88 kcal/mol). These findings highlight NA/EA-3 as a potential KRAS(G12C) inhibitor for ALL therapy, warranting experimental validation.

Introduction

Acute lymphoblastic leukaemia (ALL) is an extremely heterogeneous and malignant hematological neoplasm with uncontrolled growth of immature lymphoid cells.^1,2 Among all its subtypes, T-cell acute lymphoblastic leukaemia (T-ALL) has a special place due to its exceptionally aggressive nature and accounts for approximately 10–15% of paediatric and 25% of adult ALL cases.^3,4 Despite major advances made with chemotherapy, immunotherapy, and targeted therapies, the problem of treatment failure and relapse persists as a matter of concern, particularly in oncogenic signalling-driven cases, i.e., the RAS pathway.^1,5,6 Therefore, identifying novel treatment strategies based specifically on targeting these mutant pathways is thus the key to optimising clinical performance in T-ALL patients.^7,8

The RAS signalling pathway is a central regulator of cellular proliferation, differentiation, and survival.^9,10 Deregulated activation of this pathway by activating mutations in RAS genes primarily KRAS and NRAS results in leukemic transformation by abnormally activating downstream effectors, including the RAF/MEK/ERK and PI3K/AKT/mTOR pathways.^6,11,12 These mutations result in uncontrolled cell proliferation, apoptosis suppression, and enhanced drug resistance.^13,14 In addition to direct mutations in KRAS and NRAS, mutations in upstream regulators, including PTPN11 (SHP2) and NF1, also contribute to constitutive activation of the RAS pathway, highlighting its central position in T-ALL pathogenesis.^6,15 Among the RAS mutations, the KRAS G12C variant has been recognised as a central oncogenic driver, and it is a promising target for small-molecule inhibitors.^16,17

While the development of small-molecule inhibitors such as KRAS G12C inhibitors, including sotorasib, adagrasib, MRTX-1257, ARS853, ARS1620, JDQ-443, divarasib, garsorasib, and ASP2453, has revolutionised targeted therapy in solid tumours, their application in haematologic malignancies, including acute lymphoblastic leukaemia, remains underexplored.¹⁸ Figure 1 illustrates the selected inhibitors.^19,20 The intrinsic challenges of directly targeting RAS, including its high affinity for GTP and the lack of well-defined allosteric sites, have historically hindered drug development efforts.^21,22 To overcome these limitations, alternative therapeutic strategies are being actively investigated, including allosteric modulation, indirect inhibition via upstream and downstream effectors, and targeted degradation approaches.^6,23,24 The complexity of RAS-driven leukemogenesis underscores the need for innovative and selective inhibitors with high efficacy and minimal off-target toxicity.^6,25

Figure 1.

Selected KRAS G12C inhibitors.

Natural products have historically played a fundamental role in drug discovery, particularly in oncology, due to their structural diversity and bioactivity.^26,27 Many plant-derived and microbial secondary metabolites exhibit potent anticancer properties by modulating key oncogenic pathways, including RAS signalling.^28,29 Certain classes of natural compounds, such as flavonoids, alkaloids, terpenoids, and polyphenols, have demonstrated the ability to interfere with RAS activation, disrupt protein-protein interactions, and modulate downstream effectors. Natural products such as curcumin, epigallocatechin gallate (EGCG), boswellic acids, and betulinic acid have shown promising activity in inhibiting RAS signalling and inducing apoptosis in leukaemia models.^30,31 Given their potential as lead compounds, the identification of novel natural product-based inhibitors targeting KRAS(G12C) may provide a new avenue for T-ALL therapy.

In this study, we perform a systematic survey of African natural product databases containing a diverse array of bioactive compounds traditionally used as medicinal remedies for various diseases,^32,33 with an objective to identify putative KRAS(G12C) inhibitors using advanced computational drug discovery approaches. Our interactive, multi-dimensional in silico system starts with virtual screening(VS) to identify lead compounds with potential, followed by induced-fit docking(IFD) of the most promising candidates to optimise binding interactions. The most promising candidates are then evaluated using molecular dynamics(MD) simulations to investigate binding stability and conformational adaptability.³⁴ An ADMET analysis is also carried out to assess their pharmacokinetic profiles and drug-likeness.³⁵ Through the synergy of computational drug discovery approaches and natural product screening, this study aims to identify novel and selective KRAS(G12C) inhibitors with therapeutic value in T-cell acute lymphoblastic leukaemia(T-ALL). The discoveries of this study have the potential to open up avenues for the discovery of targeted therapies, thus improving clinical outcomes in patients with RAS-driven leukemias.

Material and method

This study employed multi-step, in silico integrative drug discovery strategy against the KRAS(G12C) mutant protein implicated in acute lymphoblastic leukaemia(ALL). As illustrated in Figure 2, included protein retrieval and preparation, natural compound database mining, virtual screening, molecular docking, induced fit docking(IFD), physicochemical and ADMET profiling, dissociation constant(KD) estimation, molecular dynamics(MD) simulations, and MM/GBSA-based binding free energy calculations. Each phase of the research was well-designed for improved acceptability, pharmacokinetic acceptability, and binding stability of candidate compounds.

Figure 2.

Study design and workflow adopted in the current work (Graphical abstract).

Protein retrieval and preparation

Three-dimensional structure of KRAS(G12C) protein(PDB ID: 6OIM, resolution: 1.70 Å) was downloaded from the RCSB Protein Data Bank(PDB) (https://www.rcsb.org/). In-complex model was cleaned in PyMOL v2.5 by removing water molecules, thereby delivering a cleansed protein structure retaining the co-crystallised ligand(MOV).³⁶ Hydrogen atoms were also added, and structural minimisation of the protein was done using the Chimaera v1.16 software to prepare the protein sufficiently for further docking and simulation analysis. The preprocessing step facilitated the removal of steric clashes, hydrogen bond optimisation, and proper preparation of the protein towards molecular docking applications.³⁷

Natural compound database retrieval

To identify potential KRAS inhibitors, natural product databases were explored. Compounds were retrieved from respective African natural products databases (https://african-compounds.org/anpdb/) like, the East African Natural Products Database(EANPDB), the North East African Natural Products Database(NEANPDB), the North African Natural Products Database(NANPDB), and South African Natural Product Database(SANPDB) (https://sancdb.rubi.ru.ac.za/), These databases provided structurally diverse bioactive molecules with potential anticancer activity. The retrieved databases were filtered using Lipinski’s Rule of Five, by filters of FAF-Drugs4 (https://fafdrugs4.rpbs.univ-paris-diderot.fr/) web server to remove non-druggable molecules.^34,37 The final ligand library was prepared in PDBQT format for docking compatibility using Open Babel (https://www.cheminfo.org/Chemistry/Cheminformatics/FormatConverter/index.html).

Gird generation

Before molecular docking, a grid box was purposefully created encircling the co-crystallised ligand binding site (MOV) to cover the most significant active site residues. Grid box was kept at the coordinates x = 1.12, y = −9.28, and z = −0.37, and the dimension was set to 48 × 48 × 40. Special emphasis was given to the key residues that have been earlier reported to be binding with KRAS(G12C), i.e., Lys16, Gln61, Tyr96, and Gln99. These residues were chosen as major targets for interaction while virtually screening.³⁶

Docking validation

To validate the docking protocol’s accuracy and reliability, the positive control AMG-510, a FDA-approved drug was re-docked into the active site of the prepared KRAS(G12C) using EasyDock Vina 2.0 with the AutoDock4 algorithm.³⁴ This step confirmed the accuracy of the docking protocol by comparing the re-docked pose to the crystal ligand conformation.

Virtual screening of selected databases

Virtual screening was performed using EasyDock Vina 2.0, a highly accurate docking software. To ensure compatibility, selected drug candidates were converted into PDBQT format. Docking was conducted with the AutoDock4 algorithm, which offers a user-friendly graphical interface. A two-tiered screening protocol was implemented: a primary screen with an exhaustiveness parameter of 16 to rapidly eliminate low-affinity ligands, followed by a secondary screen at exhaustiveness 64 to refine the top hits and enhance prediction accuracy.³⁷ The top 10% of compounds, based on their docking scores and interaction profiles, were shortlisted for subsequent Induced Fit Docking(IFD) to further assess binding flexibility and stability within the KRAS(G12C) active site.³⁸ Similar approach has been used previously.³⁹

Induced fit docking(IFD)

To improve docking accuracy and account for receptor flexibility, the top 10% of compounds from virtual screening were re-evaluated using IFD.³² This method allowed flexible adjustments of key binding site residues (e.g., Lys16, Gln61, Tyr96, Gln99) to accommodate ligand-induced conformational changes, enhancing the precision of pose prediction. Ligand binding was optimised based on interaction energy and structural complementarity.

Visualisation and analysis

Finally, the top six compounds of the entire database were visually inspected by PyMOL and Schrodinger Maestro. Molecular dynamics(MD) simulations were performed to further confirm the virtual screening outcome.³³ The best-scoring complexes from IFD were further subjected to further MM/GBSA binding free energy calculations and 200 ns molecular dynamics simulations to assess their binding stability and dynamic behaviour in a simulated biological environment.

Physciochemical analysis of lead compounds

The physiochemical properties were retrieved from the ANPDB (https://african-compounds.org/anpdb/compounds_search/)^40,41 and Swiss ADME (http://www.swissadme.ch/)⁴² by searching respective SMILES.^35,43

Pharmacokinetics and ADMET analysis of lead compounds

The pharmacokinetics, and ADMET(Absorption, Distribution, Metabolism, Excretion, and Toxicity) profiles of the best six lead compounds, together with a control drug, were investigated with the pkCSM online platform (https://biosig.lab.uq.edu.au/pkcsm/).⁴⁴ The evaluation was extensive, involving a variety of molecular descriptors and ADMET parameters for the prediction of each compound’s pharmacokinetic performance. The pkCSM platform gave insights into the efficiency of absorption, systemic distribution, metabolic stability, excretion routes, and potential toxicity threats. These forecasts are vital during initial drug development, helping to select compounds with ideal pharmacokinetic characteristics and minimal toxicity issues. Through the inclusion of ADMET analysis in the screening pipeline, this work guarantees the ranking of the best candidates for additional experimental verification and clinical application.^34,45

Dissociation constant (KD) analysis of lead compounds- KRAS(G12C) complexes

Furthermore, the dissociation constant (KD) for the top hits was computationally predicted by using PRODIGY-LIGAND web servers. The KD (dissociation constant) value is crucial in identifying the interactions between target proteins and drugs. The value gives a quantitative measure of the binding affinity, which is important in identifying the efficacy of lead compounds in targeting proteins such as KRAS (G12C).⁴⁶ Therefore, to establish the binding affinity of the compound- KRAS (G12C) complexes, we used the PRODIGY online server (https://wenmr.science.uu.nl/prodigy/) to calculate the KD value.⁴⁷ The lower the KD value, greater will be the binding strength while the higher value of KD indicates lower binding affinity.³⁴ This server was previously used to predict the KD of different molecules used against different diseases.⁴⁸

Although all six identified hit compounds exhibited favourable physicochemical properties, pharmacokinetics, ADMET profiles, and dissociation constants (KD), further validation was carried out for the top six hits, along with the positive control AMG-510. Specifically, 200 ns molecular dynamics (MD) simulations and MM/GBSA binding free energy calculations were performed to evaluate their energetic favorability and structural stability under dynamic conditions.³³

Molecular dynamics (MD) simulations

Atomistic investigation of the binding stability of the four top hit and control (AMG-510) in the KRAS(G12C) active site was performed with the AMBER21 simulation package.⁴⁹ Antechamber module of AMBER was used for the preparation of ligand topologies, and Amber General Force Field (GAFF) and ff19SB force field was used for the solvated complexes.⁴⁹ An Optimal Point Charge (OPC) solvation model was used, and sodium (Na⁺) ions were employed for system neutralisation. Multi-step energy minimisation, heating, and equilibration were performed for the relaxation of the system. Particle Mesh Ewald (PME) algorithm was used for the treatment of long-range electrostatic interactions,⁵⁰ while van der Waals and short-range Coulombic interactions were treated in the force field calculations. Langevin thermostat and Berendsen barostat were used for temperature and pressure control respectively.^39,51 All the ligand-protein complexes were subjected to a 200 ns MD simulation with a 2-fs time step. Structural dynamics and stability of the complexes were analysed with CPPTRAJ and PTRAJ.⁵² The systems were equilibrated after minimisation and heating using a sequence of steps: restrained minimisation, temperature ramping, and unrestrained equilibration to achieve thermal and pressure stability.

Post-simulation analysis of drug-protein complexes

To study the binding affinity of the protein–ligand complexes in dynamic environments, we evaluated crucial interaction-based parameters like root mean square deviation (RMSD) for global structural stability, root mean square fluctuation (RMSF) for residue mobility, radius of gyration (Rg) for protein compactness, and hydrogen bonding studies. Molecular dynamics (MD) simulations were performed for 200 ns using constant pressure (NPT) and constant volume (NVT) ensembles.^35,53–55 Post-simulation trajectory analysis was also carried out using the CPPTRAJ module. The detailed analysis provided us with information regarding the structural flexibility and binding stability upon time. The following equations were used to compute each parameter:

$$\mathit{\text{RMSD}}=\sqrt{\frac{\sum d^{2}i=1}{N_{\mathit{\text{atoms}}}}}$$ (i)

$$\mathit{\text{Thermal factor or}}\mathrm{ }B-\mathit{\text{factor}}=\left[\right(8\pi \mathrm{\ast }\mathrm{\ast }2)/3]\left(\mathit{\text{msf}}\right)$$ (ii)

The radius of gyration serves as a proxy for the degree of structural compactness in proteins.

$$R_{\mathit{\text{gyr}}}^2=\frac{1}{M}\mathrm{ }\sum _{i=1}^N{m}_i\left(r_i-R^2\right)$$ (iii)

where.

$$M={\displaystyle \sum _{i=1}^N{m}_i}$$ (iv)

$$R=N^{-1}{\displaystyle \sum _{i=1}^N{r}_i}$$ (v)

Binding free energy calculation using MM/GBSA

The Molecular Mechanics/Generalised Born Surface Area (MM/GBSA) approach is frequently utilised in drug discovery to estimate binding free energy and predict the affinity of ligand-protein interaction.⁵⁶ MM/GBSA facilitates lead optimisation through critical evaluation of critical interactions, thereby improving binding affinity and specificity. MM/GBSA applies force field calculations based on molecular mechanics in an implicit solvent model to represent the energetic terms in ligand binding to a target protein. MM/GBSA was utilised in the current study to estimate the binding free energy of the most promising ligand-protein complexes.⁵³ The binding energy (ΔG_bind) was estimated using equation (i):

$$\text{¨}\Delta G\left(\mathit{\text{bind}}\right)=\Delta G\left(\mathit{\text{complex}}\right)-\left[\Delta G\left(\mathit{\text{receptor}}\right)+\Delta G\left(\mathit{\text{ligand}}\right)\right]¨$$ (vi)

where: ΔG_complex, is the free energy of the protein–ligand complex, ΔG_receptor, is the free energy of the unbound protein, ΔG_ligand is the free energy of the unbound ligand

The different contributing factors to the total binding energy were calculated by the equation (vii) as follows:

$$\text{¨}G=G\left(\mathit{\text{bond}}\right)+\text{\hspace{0.33em}}G\left(\mathit{\text{ele}}\right)+G\left(\mathit{\text{vdW}}\right)+G\left(\mathit{\text{pol}}\right)+G\left(\mathit{\text{npol}}\right)\text{¨}$$ (vii)

Where the G_(bond) term represents the interactions among covalently bonded atoms, such as bond stretching, angle bending and torsion. The G_(ele) and G_(vdW) terms represent the Coulombic and dispersion forces between non-bonded atoms, respectively.⁴⁸ The G_(pol) and G_(pol) terms represent the solvation effects of the solvent on the solute, which depend on the solvent dielectric constant and the solute accessible surface area.^38,57 The polar term was estimated using the generalised Born (GB) model, which approximates the polar solvation energy by a pairwise interaction potential based on the effective Born radii of the atoms.

Results and discussion

This study highlights the critical role of RAS pathway hyperactivation in T-cell acute lymphoblastic leukaemia (T-ALL), contributing to therapy resistance and poor prognosis. While covalent inhibitors like AMG-510 have demonstrated clinical success in solid tumours by selectively targeting RAS signalling, their application in haematologic malignancies remains limited. By evaluating natural product-derived inhibitors, this work proposes a more biocompatible and potentially less toxic alternative to existing drugs. Overall, the findings support RAS pathway inhibition as a promising therapeutic strategy in ALL, with potential to overcome resistance and enhance current treatment outcomes.

In this context, we explored the chemical space of four African natural product databases to identify novel KRAS(G12C) inhibitors. To validate our docking protocol, the high-resolution crystal structure of KRAS(G12C) (PDB ID: 6OIM) was first docked with AMG-510, a known reference inhibitor. The docking yielded a score of −8.34 kcal/mol, and the ligand formed key interactions with Lys16 and Gln61. A low RMSD of 0.107 Å between the re-docked and crystallised poses confirmed the accuracy and reliability of the docking protocol (Figure 3).

Figure 3.

Superimposed docked and co-crystal structures; 3D alignment of Sotorasib poses A) Superimposed cartoon representation of the docked structure over the co-crystal structure. B) 3D alignment of docked poses of Sotorasib (AMS-510) with the crystal structure (MOV), showing an RMSD of 0.107 Å. C) Chemical structure of the control and Sotorasib (AMS-510).

Subsequent virtual screening of the natural product libraries revealed several compounds with superior docking scores compared to AMG-510. The top 10% of hits were refined using IFD, accounting for receptor flexibility. Six top-ranking candidates were selected for further investigation, exhibiting strong binding scores ranging from −14.50 to −10.53 kcal/mol. Their chemical structures are shown in Figure 4.

Figure 4.

Chemical structures of the top six hit compounds identified through IFD.

The AMG-510 docking score obtained here (−8.34 kcal/mol) is consistent with previously reported ranges of −8.3 to −8.5 kcal/mol in similar docking studies.^58,59 Our top candidates, (-)-shikimic acid-4-O-gallate (−14.50 kcal/mol) and 3-O-acetyl chlorogenic acid (−13.50 kcal/mol), exceed not only AMG-510 but also natural KRAS(G12C) inhibitors such as curcumin analogues and (−)-epigallocatechin gallate (EGCG derivatives) (−9.5 to −10.7 kcal/mol) indicating stronger predicted binding affinities.^45,60

These lead compounds demonstrated favourable interactions with critical residues in the KRAS(G12C) active site, including Gly60, Tyr96, Gln99, Arg68, Gly10, Thr58, Cys12, Asp92, and the key anchoring residues Lys16 and Gln61. A detailed summary of the selected compounds, including their names, sources, biological significance, docking scores, interacting residues, and interaction types, is presented in Table 1. To further evaluate their dynamic behaviour and binding stability, the top four hit selected based on docking scores and interaction profiles from IFD along with the control drug AMG-510, were subjected to 200 ns molecular dynamics (MD) simulations and MM/GBSA binding free energy analysis.

Table 1.

Features of the top-hit compounds, including their names, structures, docking scores, interacting residues, and the nature of their interactions within the 6OIM active site.

Compound name, Code, Database, Source	Docking score	Ligand Interacting functional group	Interacting Residues of 6OIM	Interaction Nature
(-)-shikimic acid-4-O-gallate (NA/EA-1) NANPDB and EANPDB Source:  Euphorbia helioscopia	−14.50	OH(COOH)	Lys16	HB
		CO(COOH)	Gln61	HB
		OH(cyclohexenediol	Gly60	HB
		OH(cyclohexenediol)	Tyr96	HB
		OH(cyclohexenediol	Gln61	HB
		CO(5-Methylpyrogallol)	Gln61	HB
		OH(5-Methylpyrogallol)	Gln61	HB
		OH(5-Methylpyrogallol)	Gln99	HB
		5-Methylpyrogallol ring	Arg68	Pi-Cation
3-O-acetyl chlorogenic acid (NA/EA-2) NANPDB and EANPDB Source:  Carissa edulis	−13.50	CO(ester)	Gly10	HB
		OH(Pyrocatechol)	Gln61	HB
		Pyrocatechol ring	Arg68	Pi-Pi
		OH(pyrocatechol)	Gln99	HB
		OH(pyrocatechol)	Gln99	HB
		OH(acetyl-quinic acid)	Gln60	HB
		CO of COOH(acetyl-quinic acid)	Lys16	HB
		OH of COOH (acetyl-quinic acid)	Lys16	SB
		OH of ester (acetyl-quinic acid)	Cys12	HB
Rosmarinic acid (NA/EA-3) NANPDB and EANPDB Source:  Thalassia hemprichii	−12.40	OH(pyrocatechol)	Asp92	HB
		OH(acetyloxyacetic acid)	Lya16	HB
		CO(acetyloxyacetic acid)	Gly10	HB
		Pyrocatechol ring	Arg68	Pi-Pi
		OH(pyrocatechol)	Gln61	HB
		OH(pyrocatechol)	Gln99	HB
		OH(pyrocatechol)	Gln99	HB
		OH(pyrocatechol)	Gln99	HB
Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4) SANPDB Source: Ecklonia maxima	−11.04	OH(Resorcinol)	Gln99	HB
		OH(Resorcinol)	Gln99	HB
		OH(Resorcinol)	Gln61	HB
		OH(Resorcinol)	Gln61	HB
		OH(Resorcinol)	Thr58	HB
		OH(Resorcinol)	Gly10	HB
		Resorcinol	Arg68	Pi-Cation
3,5,3′-trihydroxy-7,4′-dimethoxyflavone (NA/EA-5) NANPDB and EANPDB Source:  Gardenia ternifolia	−10.60	OH(2-methoxyphenol)	Lys16	HB
		O(2-methoxyphenol)	Lys16	HB
		OH(2-methoxyphenol)	Thr58	HB
		OH (3-hydroxypyran-4-one)	Gln61	HB
		OH(3-methoxyphenol)	Gln99	HB
		3-methoxyphenol	Arg68	Pi-Cation
		2-methoxyphenol ring	Tyr96	Pi-Pi
Quercetin (NA/EA-6) NANPDB and EANPDB Source: Chrozophora species	−10.53	OH(Resorcinol)	Lys16	HB
		OH(Resorcinol)	Thr58	HB
		OH(Resorcinol)	Gln61	HB
		OH(pyrocatechol)	Gln99	HB
		pyrocatechol	Arg68	Pi-Cation
Control(AMG-510)	−8.34	CO(Ketonic)	Lys16	HB
		OH(3-fluorophenol)	Gln99	HB
		OH(3-fluorophenol)	Gln61	HB
		3-fluorophenol ring	Arg68	Pi-Pi
		3-fluoropyridine ring	Tyr96	Pi-Pi

(Metalic bonding: MB, Hydrogen bonding: HB, and pi interaction: Pi-Pi).

The control compound, AMG-510, exhibits a docking score of −8.34, indicating a lower binding affinity to the 6OIM protein compared to the tested natural compounds. AMG-510 is a known KRAS inhibitor, and docking analysis reveals key interactions with Lys16, Gln99, Gln61, Arg68, and Tyr96, contributing to ligand stabilisation through hydrogen bonding and π-π interactions as shown in Figure 5. The ketonic carbonyl group forms a hydrogen bond with Lys16, while the hydroxyl group of the 3-fluorophenol moiety interacts with Gln99 and Gln61. Additionally, π-π stacking interactions occur between the 3-fluorophenol ring and Arg68, as well as the 3-fluoropyridine ring and Tyr96, further stabilising the complex. Despite these interactions, the lower docking score suggests that AMG-510 binds less effectively to 6OIM than the tested natural compounds, highlighting the potential of natural inhibitors as alternative therapeutic candidates.

Figure 5.

A) 2D and B) 3D docked poses of AMG-510 in the KRAS(G12C) active site, highlighting detailed interactions.

The highest-ranking compound, (-)-shikimic acid-4-O-gallate (NA/EA-1), identified from NANPDB and EANPDB and sourced from Euphorbia helioscopia, achieved a docking score of −14.50 kcal/mol, indicating a strong binding affinity to the 6OIM protein. Traditionally, this compound is used for treating skin diseases, gonorrhoea, migraines, parasitic infections, and warts and is also employed as a poison, insecticide, and in Ayurveda for bronchitis and rheumatism.⁶¹ Molecular docking analysis reveals significant interacting residues within the active site of 6OIM, including Lys16, Gln61, Gly60, Tyr96, Gln99, and Arg68, which stabilise the ligand through multiple hydrogen bonds (HB) and π-cation interactions (Figure 6). Specifically, the carboxylate, hydroxyl group of the interacts with Lys16, while the carbonyl forms a hydrogen bond with Gln61. Similarly, the hydroxyl groups of the cyclohexenediol moiety establish hydrogen bonds with Gly60, Tyr96, and Gln61, further stabilising the complex. Besides, 5-methylpyrogallol has carbonyl and hydroxyl functional groups. moiety interacts with Gln61 and Gln99 through hydrogen bonding. There is a π-cation interaction detected between Arg68 and the 5-methylpyrogallol group, which leads to greater stability in the binding site. All these numerous interactions, especially hydrogen bonding and π-cation interactions, demonstrate that (-)-shikimic acid-4-O-gallate has strong affinity with the active site of 6OIM, possibly preventing its action more effectively than the conventional inhibitors. With its high affinity and multiplicity of pharmacological activities, the compound is very promising for further Experimental confirmation in pharmaceutical drug discovery and medical applications.

Figure 6.

A) 2D and B) 3D docked poses of (-)-shikimic acid-4-O-gallate (NA/EA-1) in the KRAS(G12C) active site, highlighting detailed interactions.

The second top-scoring compound, identified from NANPDB and EANPDB is 3-O-acetyl chlorogenic acid (NA/EA-2), sourced from Carissa edulis, exhibits a docking score of −13.50, indicating strong binding affinity to the 6OIM protein. Traditionally, C. edulis is used to treat headaches, chest complaints, rheumatism, oedema, STDs, rabies, fever, sickle cell anaemia, cough, ulcers, toothache, and worm infestations, with pharmacological studies confirming their antiviral, anticonvulsant, antiplasmodial, antimicrobial, analgesic, diuretic, and hypoglycaemic properties.⁶² Docking analysis reveals key interactions with Gly10, Gln61, Arg68, Gln99, Gln60, Lys16, and Cys12, which contribute to ligand stabilisation through hydrogen bonding, salt bridge, and π-π interactions as shown in Figure 7. The carbonyl of the ester forms a hydrogen bond with Gly10, while the hydroxyl of the pyrocatechol moiety interacts with Gln61 and Gln99. A π-π stacking interaction between the pyrocatechol ring and Arg68 enhances ligand stability. The acetyl-quinic acid moiety forms multiple hydrogen bonds with Gln60 and Lys16, with Lys16 also forming a salt bridge by its carboxyl group, further reinforcing binding. Additionally, the ester hydroxyl group interacts with Cys12, adding to the overall stability. These extensive interactions suggest 3-O-acetyl chlorogenic acid as a strong potential inhibitor of 6OIM, making it a promising candidate for further drug development and therapeutic applications.

Figure 7.

A) 2D and B) 3D docked poses of 3-O-acetyl chlorogenic acid (NA/EA-2) in the KRAS(G12C) active site, highlighting detailed interactions.

From NANPDB and EANPDB, the third top-scoring compound listed is Rosmarinic acid (NA/EA-3), derived from Thalassia hemprichii, exhibits a docking score of −12.40, indicating notable binding affinity to the 6OIM protein. This compound is known for its antimicrobial activity.⁶³ Docking analysis reveals key interactions with Asp92, Lys16, Gly10, Arg68, Gln61, and Gln99, contributing to ligand stabilisation through hydrogen bonding and π-π interactions as shown in Figure 8. The hydroxyl group of the pyrocatechol moiety forms hydrogen bonds with Asp92, Gln61, and Gln99, reinforcing its stability within the active site. The acetyloxyacetic acid moiety interacts via hydrogen bonding with Lys16 and Gly10, while the pyrocatechol ring forms a π-π stacking interaction with Arg68, further stabilising the complex. Multiple hydrogen bonds between Gln99 and the hydroxyl groups of pyrocatechol enhance its binding affinity. These interactions suggest that Rosmarinic acid may effectively inhibit 6OIM, making it a strong candidate for further investigation in antimicrobial drug development.

Figure 8.

A) 2D and B) 3D docked poses of Rosmarinic acid (NA/EA-3) in the KRAS(G12C) active site, highlighting detailed interactions.

The fourth top-scoring compound shortlisted from SANCDB is Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4), sourced from Ecklonia maxima, exhibits a docking score of −11.04, indicating a moderate binding affinity to the 6OIM protein. This compound is known for its anti-acetylcholinesterase activity.⁶⁴ Docking analysis reveals key interactions with Gln99, Gln61, Thr58, Gly10, and Arg68, contributing to ligand stabilisation through hydrogen bonding and π-cation interactions as shown in Figure 9. The hydroxyl groups of the resorcinol moiety form multiple hydrogen bonds with Gln99, Gln61, Thr58, and Gly10, reinforcing its stability within the active site. Additionally, a π-cation interaction between the resorcinol ring and Arg68 further enhances binding affinity. These interactions suggest that Dibenzo-1,4-dioxin-1,3,6,8-tetraol may act as a potential 6OIM inhibitor, making it a promising candidate for further investigation in drug development targeting acetylcholinesterase-related disorders.

Figure 9.

A) 2D and B) 3D docked poses of Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4) in the KRAS(G12C) active site, highlighting detailed interactions.

Ranked fifth among the top-scoring compounds, is 3,5,3′-trihydroxy-7,4′-dimethoxyflavone (NA/EA-5), identified from NANPDB and EANPDB and sourced from Gardenia ternifolia, achieved a docking score of −10.60 kcal/mol, indicating a moderate binding affinity to the 6OIM protein. Traditionally, this compound is used in East African medicine to treat infections, pain, fevers, digestive issues, mental disorders, and as an anti-snake venom and purgative.⁶⁵ Docking analysis reveals key interactions with Lys16, Thr58, Gln61, Gln99, Arg68, and Tyr96, contributing to ligand stabilisation through hydrogen bonding, π-cation, and π-π interactions as shown in Figure 10. The hydroxyl and oxygen groups of the 2-methoxyphenol moiety form hydrogen bonds with Lys16 and Thr58, reinforcing ligand stability. The hydroxyl group of 3-hydroxypyran-4-one interacts with Gln61, while the hydroxyl group of 3-methoxyphenol establishes a hydrogen bond with Gln99. Additionally, π-cation interactions occur between the 3-methoxyphenol ring and Arg68, and π-π stacking interactions are observed between the 2-methoxyphenol ring and Tyr96, further stabilising the complex. These interactions suggest that 3,5,3′-trihydroxy-7,4′-dimethoxyflavone may effectively bind to 6OIM, making it a promising candidate for further investigation in drug discovery.

Figure 10.

A) 2D and B) 3D docked poses of 3,5,3′-trihydroxy-7,4’-dimethoxyflavone (NA/EA-5) in the KRAS(G12C) active site, highlighting detailed interactions.

Quercetin (NA/EA-6), obtained from NANPDB and EANPDB and derived from Chrozophora species, ranked sixth among the top-scoring compounds, with a docking score of −10.53 kcal/mol, suggesting a moderate binding affinity to the 6OIM protein. Quercetin is a flavonoid widely found in fruits and vegetables, known for its antioxidant, anti-inflammatory, antiviral, and anticancer properties.⁶⁶ Molecular docking analysis highlights key interactions with Lys16, Thr58, Gln61, Gln99, and Arg68, stabilising the ligand through hydrogen bonding and π-cation interactions as shown in Figure 11. The hydroxyl groups of the resorcinol moiety form hydrogen bonds with Lys16, Thr58, and Gln61, while the hydroxyl group of pyrocatechol interacts with Gln99. Additionally, a π-cation interaction between the pyrocatechol ring and Arg68 further enhances binding affinity. These interactions suggest that Quercetin may serve as a potential inhibitor of 6OIM, making it a valuable candidate for further biological and therapeutic investigations.

Figure 11.

A) 2D and B) 3D docked poses of Quercetin (NA/EA-5) in the KRAS(G12C) active site, highlighting detailed interactions.

Our discovered natural compounds exhibited stronger binding to 6OIM than the control AMG-510 (−8.34), with (-)-shikimic acid-4-O-gallate (−14.50) and 3-O-acetyl chlorogenic acid (−13.50) showing the highest affinities. Key interactions with Lys16, Gln61, Gln99, Arg68, and Tyr96 stabilised these ligands via hydrogen bonding and π-cation interactions. Other compounds also outperformed AMG-510, highlighting their potential as alternative inhibitors. Given their pharmacological relevance, these natural molecules warrant further experimental validation as potential therapeutics. The favourable physicochemical profiles identified here position them as strong candidates for drug development and justify a comprehensive evaluation of their pharmacokinetic and toxicity characteristics to confirm their clinical potential.

Physicochemical analysis of selected natural compounds and control

The selected bioactive compounds exhibit promising physicochemical properties compared to the control (AMG-510) as summarized in Table 2 and illustrated in Figure 12. (-)-Shikimic acid-4-O-gallate (NA/EA-1) has a molecular weight of 340.28 g/mol, moderate flexibility (NRB = 4), strong hydrogen bonding potential (HBD = 6, HBA = 9), and good solubility (LogS = −2.10), but low skin permeability (Log Kp = −8.12 cm/s). 3-O-Acetyl Chlorogenic Acid (NA/EA-2) has a slightly higher MW (396.35 g/mol), good drug-likeness (0 Lipinski violations), and moderate lipophilicity (LogP = 0.15), though its high TPSA (170.82 Ų) might slightly hinder passive absorption. Rosmarinic Acid (NA/EA-3) also follows Lipinski’s rule with good solubility (LogS = −3.44), moderate lipophilicity (LogP = 1.58), and strong hydrogen bonding potential (HBD = 5, HBA = 7), but its high TPSA (144.52 Ų) could impact permeability. Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4) is the smallest (MW = 248.19 g/mol), highly rigid (NRB = 0), with moderate hydrogen bonding (HBD = 4, HBA = 6), good TPSA (99.38 Ų), and slightly better solubility (LogS = −3.36) than some other compounds, indicating potential oral bioavailability. Flavonoid-based compound NA/EA-5 has a molecular weight of 330.29 g/mol, with good hydrogen bonding features (HBD = 2, HBA = 7), moderate flexibility (NRB = 3), and acceptable lipophilicity (LogP = 2.25). While its solubility (LogS = −4.25) and skin permeability (Log Kp = −5.99 cm/s) are within a tolerable range, its TPSA = 109.36 Ų, which supports reasonable passive absorption and oral bioavailability, and it shows no Lipinski violations. Quercetin (NA/EA-6) is another well-balanced compound with a MW of 302.24 g/mol, moderate flexibility (NRB = 1), TPSA = 131.36 Ų, and good solubility (LogS = −3.16), making it a strong drug candidate. In contrast, AMG-510 has a significantly higher MW (560.59 g/mol), high lipophilicity (LogP = 4.02), poor solubility (LogS = −5.82), and violates Lipinski’s rule due to its high molecular weight, which may impact its bioavailability.

Table 2.

Physicochemical properties of selected bioactive compounds and control (AMG-510), including MW, rotatable bonds, H-bond donors/acceptors, TPSA, LogP, solubility, and Lipinski violations.

Property	NA/EA-1	NA/EA-2	NA/EA-3	SA-4	NA/EA-5	NA/EA-6	AMG-510
Molecular Weight (MW) g/mol	340.28	396.35	360.31	248.19	330.29	302.24	560.59
No. of Rotatable Bonds (NRB)	4	7	7	0	3	1	6
No. of H-bond Donors (HBD)	6	5	5	4	2	5	1
No. of H-bond Acceptors (HBA)	9	10	7	6	7	7	8
Topological Polar Surface Area (TPSA)	164.74 Ų	170.82 Ų	144.52Ų	99.38 Ų	109.36 Ų	131.36 Ų	104.45 Ų
Lipophilicity (LogP, Consensus)	−0.10	0.15	1.58	1.47	2.25	1.23	4.02
Water Solubility (LogS, ESOL)	−2.10	−2.09	−3.44	−3.36	−4.25	−3.16	−5.82
Log Kp (Skin Permeation)	−8.12 cm/s	−8.61 cm/s	−6.82 cm/s	−6.14 cm/s	−5.99 cm/s	−7.05 cm/s	−6.90 cm/s
Lipinski Violations	1	0	0	0	0	0	Yes (MW > 500)
Molecule Class	Phenolic	Phenolic	Phenolic	Benzodioxins (Classyfire)	Flavonoid	Flavonoid	//

Figure 12.

Radar plots illustrating the physicochemical property profiles of selected bioactive compounds and the control (AMG-510). Each plot summarises six key drug-likeness parameters: lipophilicity (LIPO), size (SIZE), polarity (POLAR), solubility (INSOLU), saturation (INSATU), and flexibility (FLEX). Compounds analysed include: A) (-)-Shikimic acid-4-O-gallate (NA/EA-1), B) 3-O-Acetyl chlorogenic acid (NA/EA-2), C) Rosmarinic acid (NA/EA-3), D) Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4), E) 3,5,3′-Trihydroxy-7,4′-dimethoxyflavone (NA/EA-5), F) Quercetin (NA/EA-6), G) Control compound (AMG-510). The pink shaded area indicates the optimal range for oral bioavailability, providing a visual comparison of each compound’s suitability as a drug-like candidate.

Overall, the selected natural compounds show more favourable drug-like properties than the control (AMG-510), including lower molecular weights, better solubility, no Lipinski violations, and balanced lipophilicity. Their strong hydrogen bonding capacity and acceptable TPSA values further support their potential for good oral bioavailability. In contrast, AMG-510’s high molecular weight, low solubility, and rule violation may limit its pharmacokinetic profile. These results highlight the selected compounds as strong candidates for further drug development.

ADMET analysis of selected natural compounds and control

Unlike AMG-510, which violates Lipinski’s rule due to high molecular weight (560.59 g/mol) and lipophilicity (LogP = 4.02), all six natural candidates meet drug-likeness criteria. The improved water solubility observed here (LogS −2.09 to −4.25) aligns with previous findings that phenolic- and flavonoid-based inhibitors typically have better aqueous solubility and lower mutagenicity than synthetic KRAS inhibitors.^45,67 Furthermore, none of our compounds displayed AMES toxicity, in contrast to AMG-510 and certain synthetic analogues which have been associated with mutagenic risk.

The absence of hepatotoxicity and skin sensitisation in our candidates reinforces their safety profiles, matching earlier reports on plant-based inhibitors with low off-target toxicity.²⁷ Chronic toxicity (LOAEL) values for our candidates were 1–2 orders of magnitude lower than AMG-510, suggesting a broader therapeutic window.

Table 3 summarises the toxicity and safety profiles of the tested compounds, including (-)-Shikimic acid-4-O-gallate (NA/EA-1), 3-O-acetyl chlorogenic acid (NA/EA-2), Rosmarinic acid (NA/EA-3), Dibenzo-1,4-dioxin-1,3,6,8-tetraol (SA-4), 3,5,3′-trihydroxy-7,4′-dimethoxyflavone (NA/EA-5), and Quercetin (NA/EA-6), in comparison with the control AMG-510. None of the tested compounds showed AMES toxicity, indicating a non-mutagenic nature, whereas AMG-510 was AMES toxic, highlighting potential genotoxic concerns. The maximum tolerated dose (MTD) values of the natural compounds ranged from 0.163 to 0.775 log mg/kg/day, demonstrating their broad safety window compared to AMG-510 (0.438 log mg/kg/day). The hERG I inhibition, associated with cardiac toxicity, was absent in all compounds, including the control, while hERG II inhibition was observed only in 3,5,3′-trihydroxy-7,4′-dimethoxyflavone (NA/EA-5), implying a slightly higher cardiac risk for this particular compound. In terms of oral rat acute toxicity (LD50), all test compounds demonstrated values ranging from 2.173 to 2.607 mol/kg, comparable to AMG-510 (2.482 mol/kg), suggesting a relatively similar acute toxicity profile. However, chronic toxicity (LOAEL) was significantly lower for all natural compounds, ranging from 1.815 to 3.136 log mg/kg_bw/day, whereas AMG-510 exhibited an excessively high LOAEL of 28.35 log mg/kg_bw/day, suggesting greater long-term toxicity for the control. Importantly, none of the natural compounds showed hepatotoxicity or skin sensitisation, reinforcing their safety, similar to AMG-510. T. Pyriformis toxicity was consistent across the compounds, with values ranging from 0.285 to 0.362 log µg/L, whereas Minnow toxicity was much higher for AMG-510 (8.761 log mM), indicating a significantly higher environmental toxicity compared to the natural compounds, which ranged from 0.828 to 4.641 log mM. Collectively, these findings highlight that our natural compounds possess a superior safety profile compared to AMG-510, as they exhibit lower chronic toxicity, lack genotoxic effects, minimal cardiac risks, and reduced environmental impact, making them promising candidates for further therapeutic development.

Table 3.

Summary of Toxicity and Safety Comparison of Natural Compounds against AMG-510.

Property	NA/EA-1	NA/EA-2	NA/EA-3	SA-4	NA/EA-5	NA/EA-6	AMG-510 (Control)
AMES toxicity	No	No	No	No	No	No	Yes
Max. tolerated dose (human) log mg/kg/day	0.2879	0.625	0.163	0.302	0.633	0.775 l	0.438
hERG I inhibitor	No	No	No	No	No	No	No
hERG II inhibitor	No	No	No	No	Yes	No	No
Oral Rat Acute Toxicity (LD50) mol/kg	2.4330	2.211	2.444	2.173	2.356	2.607	2.482
Oral Rat Chronic Toxicity (LOAEL) log mg/kg_bw/day	3.1360	3.087	2.855	2.876	1.815	2.464	28.35
Hepatotoxicity	No	No	No	No	No	No	No
Skin Sensitisation	No	No	No	No	No	No	No
T. Pyriformis toxicity log µg/L	0.285	0.285	0.286	0.362	0.305	0.335	0.285
Minnow toxicity log mM	4.641	3.840	3.363	2.194	0.828	1.957	8.761

These results are consistent with prior reports on plant-derived inhibitors, which frequently show lower mutagenicity, reduced chronic toxicity, and improved environmental safety compared to synthetic small-molecule KRAS inhibitors.^27,58,68 Such favourable ADMET profiles, when combined with strong predicted binding affinities, make these compounds particularly promising for further development as safer KRAS(G12C) therapeutic agents

Binding strength analysis of lead complexes as dissociation constant (KD)

The dissociation constant (KD) is a key parameter for quantifying drug–protein interactions, providing a direct measure of binding affinity between a target protein and its ligand. To assess the interaction strength of our selected lead compounds with KRAS(G12C), we employed the PRODIGY web server, renowned for its high-accuracy binding affinity predictions. Specifically, we used the enhanced PRODIGY-LIG module, optimised for small molecules, which leverages atomic contact information at the binding interface for more precise KD estimation.^28,29 This study presents the first reported KD-based evaluation of KRAS(G12C) inhibitors using this approach. All six lead candidates demonstrated strong predicted binding affinities, with binding free energies ranging from −8.14 to −9.44 kcal/mol. The top-performing compound, NA/EA-3 (−9.28 kcal/mol), exhibited target engagement potential comparable to clinically advanced molecules. Other promising candidates included NA/EA-1 (−8.77 kcal/mol), NA/EA-2 (−9.25 kcal/mol), SA-4 (−8.14 kcal/mol), NA/EA-5 (−9.44 kcal/mol), and NA/EA-6 (−8.58 kcal/mol). These results highlight the strong binding potential of the identified compounds against KRAS(G12C) and support their candidacy as potent inhibitors for the suppression of acute lymphoblastic leukaemia progression.

Although all six compounds showed favourable KD values, the four hits NA/EA-1, NA/EA-2, NA/EA-3, and SA-4 were shortlisted for MD simulations because they combined high docking scores with extensive and well-oriented interactions involving key catalytic residues of KRAS(G12C), ensuring strong predicted binding stability. In addition, these four candidates displayed clean ADMET profiles, with no hERG inhibition, acceptable chronic oral toxicity thresholds (LOAEL > 2.85 log mg/kg-bw/day), and low aquatic toxicity (Minnow toxicity > 2.19 log mM), supporting their drug-likeness and safety. In contrast, NA/EA-5 and NA/EA-6, despite strong KD values, were deprioritized due to lower docking scores, fewer critical residue interactions, and unfavourable ADMET features, including hERG II inhibition risk (NA/EA-5), low chronic oral toxicity thresholds (1.815 and 2.464 log mg/kg-bw/day), and high aquatic toxicity (0.828 and 1.957 log mM), reducing their suitability as drug candidates.

Molecular dynamics simulations analysis of ligands‑KRAS(G12C) complexes

RMSD analysis of selected natural compounds and control

The examination of the dynamic stability of protein–ligand interactions within the binding cavity is important in determining the binding affinity and conformational integrity of small-molecule inhibitors.⁵³ In the current study, molecular dynamics (MD) simulations of 200 ns were conducted to investigate the structural stability of the top four-ranked protein–ligand complexes against control compound AMG-510, targeting KRAS (G12C). Root Mean Square Deviation (RMSD) analysis was a necessary measure for monitoring conformational fluctuations and investigating the dynamic behaviour of each complex over time. The results showed that all compounds investigated had stable RMSD trajectories without visible deviations, indicating stable and persistent binding within the active site. As shown in Figure 13, TopHit1(NA/EA-1) showed early stabilisation during the simulation, with RMSD values within the 1.0–2.5 Å range, indicating strong and consistent binding. TopHit2 (NA/EA-2) also showed a similar profile, stabilising at approximately 2.5 Å with minimal fluctuations reported. Of particular interest, TopHit3 (NA/EA-3) showed the lowest RMSD values, fluctuating between 1.0–1.5 Å during the simulation, indicating exceptional binding stability.

Figure 13.

RMSD-based stability analysis of KRAS (G12C) drug complexes. A) RMSD trajectory of TopHit1-KRAS (G12C) complex. B) RMSD trajectory of TopHit2-KRAS (G12C) complex. C) RMSD trajectory of TopHit3-KRAS (G12C) complex. D) RMSD trajectory of TopHit4-KRAS (G12C) complex. E) RMSD trajectory of the control compound (AMG-510) complex.

RMSD values for NA/EA-3 remained between 1.0–1.5 Å, indicating superior stability to AMG-510 (1.5–2.0 Å) and matching stability benchmarks (<2.0 Å) reported for effective KRAS–ligand complexes [66, FrontPharmacol2022; PMC9821013]. Similarly, the compact Rg profile observed for NA/EA-3 aligns with MD studies of other high-affinity KRAS inhibitors, where stable Rg values indicate persistent complex integrity [66]. RMSF analysis revealed localised flexibility between residues 50–70, consistent with adaptive conformational changes reported in KRAS inhibitor simulations.^69–71

TopHit4 (SA-4) had slightly more flexibility with RMSD values fluctuating between 1.4–1.6 Å, indicating possibly moderate but still favourable binding interactions. In contrast, the control compound AMG-510 had RMSD values fluctuating between 1.5–2.0 Å, used as a control for dynamic stability. Of additional interest, the RMSD values of the compounds investigated, particularly TopHit3 and TopHit4, were either lower than or equal to those of the control compound, indicating superior or equivalent dynamic stability in the KRAS (G12C) binding site. These results indicate the potential of the screened compounds, particularly TopHit3, as potent therapeutic agents. Additional verification using binding free energy calculations and experimental studies is recommended to validate these computational findings.

Rg analysis of selected natural compounds and control

The radius of gyration (Rg) analysis provides useful information regarding the structural compactness and stability of protein–ligand complexes during the 200 ns molecular dynamics (MD) simulation.⁷² As shown in Figure 14, the Rg of the top four hit and the control compound AMG-510 was tracked to analyse their dynamic characteristics. TopHit1 had an Rg of about 17.6 Å until 30 ns, after which there was a slight rise to about 18.0 Å, indicating stable binding with no appreciable structural enlargement. TopHit2 had an average Rg of 17.4 Å until 70 ns, which later decreased slowly to 17.2 Å before finally stabilising at 17.6 Å before 80 ns. There was a minor decrease to 15.5 Å at around 120 ns, and this tight state was maintained until the simulation’s end, which suggests a structurally stable complex with small fluctuations. TopHit3 had a stable Rg profile with very little fluctuation, retaining a tight conformation throughout, which is indicative of a maintained and tightly bound state. TopHit4, while exhibiting comparatively greater fluctuations at the initial stages, displayed an increase in Rg up to 18.0 Å followed by a sharp drop to 15.4 Å at approximately 120 ns, post which the system returned to a stable state, revealing moderate flexibility with final compactness. Conversely, the control ligand AMG-510 retained the lowest constant values of Rg between 15.5 and 15.6 Å, confirming its long-proven binding compactness. Overall, the compounds investigated demonstrated Rg profiles that were either comparable to or better than the control, suggesting their strong ability to stabilise the KRAS (G12C) binding pocket. The low degree of fluctuation across simulations supports the notion that these ligands form compact, stable complexes, warranting further validation through binding free energy calculations and experimental studies.

Figure 14.

Radius of gyration (Rg) analysis of KRAS (G12C)-ligand complexes over a 200 ns MD simulation. A) Hit1 B) Hit2 C) Hit3 D) Hit4 E) control AMG-510. Results indicate strong binding potential for the tested compounds.

RMSF analysis of selected natural compounds and control

The root mean square fluctuation (RMSF) analysis was performed to assess residue-level flexibility and local dynamic behaviour⁷³ of the KRAS (G12C) complexes with the top four hit and the control (AMG-510). RMSF offers valuable insight into atomic mobility, where lower values denote stable regions and higher fluctuations suggest flexible or disordered segments. As illustrated in Figure 15, all complexes exhibited an average RMSF around 1.0 Å, indicative of overall structural rigidity and stable ligand binding. However, a consistent pattern of elevated flexibility was observed between residues 50–70, a region likely involved in adaptive conformational changes upon ligand interaction. Among the tested candidates, TopHit1 showed the lowest RMSF values, signifying the least structural perturbation and strong, stable binding. The other TopHit2, TopHit3, and TopHit4 demonstrated comparable trends, with only minor residue-specific deviations in flexibility. The control compound AMG-510 displayed a fluctuation profile closely aligned with the test compounds, supporting the structural integrity of all complexes. These results, consistent with the RMSD and Rg analyses, further affirm the potential of the selected compounds, particularly TopHit1, as stable and promising KRAS (G12C) inhibitors. Further experimental and thermodynamic validation will be essential to confirm these computational observations.

Figure 15.

The RMSF trends for all complexes, highlighting the relative stability of each ligand-bound structure.

Hydrogen bond analysis of selected natural compounds and control

The hydrogen bond analysis provides insights into the stability of molecular interactions within the KRAS (G12C) complexes.⁷⁴ To examine alterations in the hydrogen bonding pattern throughout the simulation, the total count of hydrogen bonds in each trajectory was determined. As depicted in Figure 16, TopHit1 exhibits the highest stability with an average of approximately 82 hydrogen bonds, followed by TopHit2 with around 81 hydrogen bonds. Similarly, TopHit3 and TopHit4 maintain stable hydrogen bond counts of approximately 86 and 80, respectively. NA/EA-3 maintained ∼86 hydrogen bonds throughout the simulation, exceeding AMG-510 (∼79) and outperforming many natural inhibitor complexes, which typically maintain 60–75 persistent hydrogen bonds. Persistent hydrogen bond networks are associated with higher residence times and stronger inhibitory activity.^71,75

Figure 16.

Hydrogen bond analysis of KRAS (G12C) complexes over the simulation. A) TopHit1 (∼82 H-bonds), B) TopHit2 (∼81), C) TopHit3 (∼86), D) TopHit4 (∼80). E) control (AMG-510) (∼79 H-bonds, indicating strong binding in the top hits.

In comparison, the control (AMG-510) complex demonstrates a slightly lower hydrogen bond stability, averaging around 79 hydrogen bonds. The results presented herein reveal that all the complexes possess a well-structured and robust hydrogen bonding network, enabling stable interactions among the lead molecules identified and KRAS (G12C). The greater number of hydrogen bonds in the top-ranked complexes reveal higher binding affinities in comparison to AMG-510, hence the efficacy of their use as potential inhibitors. These results concur with molecular docking, RMSD, and RMSF analyses, which further establish the stability of the complexes and their promise as potential drug candidates.

Binding free energies calculation of shortlisted compounds by MM/GBSA approach

The MM/GBSA binding free energy calculations provided valuable insights into the stability and interaction strength of the selected bioactive compounds compared to the control (AMG-510)⁷⁶ (Table 4). The van der Waals (ΔEvdw) interactions played a crucial role in binding affinity, with the control showing −37.99 kcal/mol. In comparison, TopHit1 (NA/EA-1), TopHit2 (NA/EA-2), TopHit3 (NA/EA-3), and TopHit4 (SA-4) recorded −44.4924, −40.3345, −73.3268, and −30.2649 kcal/mol, respectively. The electrostatic energies (ΔEele) for the control and the four compounds were 32.8087, 88.3039, 119.5207, 9.7358, and 68.8063 kcal/mol, respectively. The gas-phase free energy (ΔG_Gas), representing the combined van der Waals and electrostatic interactions, was highest for TopHit3 (46.1939 kcal/mol) and TopHit2 (47.9694 kcal/mol), indicating strong interaction potential. Solvation free energy (ΔG_Solv) contributed notably, with TopHit3 having the highest solvation penalty (−100.6178 kcal/mol), reflecting a significant desolvation cost upon binding, while TopHit1 had the lowest solvation energy (−30.9429 kcal/mol). The total binding free energy (ΔG_total) values were −32.8877 kcal/mol for the control, and −42.6265, −29.2503, −54.4239, and −25.9332 kcal/mol for TopHit1, TopHit2, TopHit3, and TopHit4, respectively. The MM/GBSA total binding free energy for NA/EA-3 (−54.42 kcal/mol) is markedly more favourable than AMG-510 (−32.88 kcal/mol) and higher than many natural KRAS inhibitor reports, which often range from −35 to −50 kcal/mol.^70,77

Table 4.

MM-GBSA Binding Free Energy Calculations (kcal/mol).

Complexes	ΔEvdw	ΔEele	EGB	ESURF	Delta G Gas	Delta G Solv	ΔG total
TopHit1 (NA/EA-1)	−44.4924	32.8087	−25.8755	−5.0673	−11.6837	−30.9429	−42.6265
TopHit2 (NA/EA-2)	−40.3345	88.3039	−72.1025	−5.1172	47.9694	−77.2197	−29.2503
TopHit3 (NA/EA-3)	−73.3268	119.5207	−93.8826	−6.7351	46.1939	−100.6178	−54.4239
TopHit4 (SA-4)	−30.2649	9.7358	−1.5049	−3.8992	−20.529	−5.4041	−25.9332
Control (AMG-510)	−37.99	68.8063	−59.4479	−4.2562	30.8163	−63.704	−32.8877

Among the tested compounds, NA/EA-3 exhibited the highest binding affinity, making it a particularly promising candidate for further pharmacological evaluation. Importantly, all identified compounds are natural products, known for their safety, biocompatibility, and reduced side effects compared to synthetic drugs. Their broader range of biological activities and therapeutic potential further enhance their suitability for medical applications. The superior safety profile of these natural compounds underscores their value as excellent candidates for experimental validation and clinical studies.

Ligands retention during MD simulation

Figure 17 illustrates the dynamic behaviour of the ligand within the protein’s active site over the course of a 200-nanosecond molecular dynamics simulation. At each observed time frame 0 ns, 50 ns, 100 ns, 150 ns, and 200 ns all the ligands maintain its position within the binding pocket. This consistent occupancy suggests strong and stable interactions between the ligands and key active-site residues, supporting its potential as a promising inhibitor with high binding affinity and structural stability

Figure 17.

Molecular dynamics snapshots of the protein-ligand complex at (a) 0 ns, (b) 50 ns, (c) 100 ns, (d) 150 ns, and (e) 200 ns. The ligand remains stably bound within the active site throughout the simulation, indicating sustained interaction and conformational stability of the complex over time.

Collectively, these comparative analyses demonstrate that our identified compounds not only surpass AMG-510 in predicted binding and safety but also outperform or match previously reported natural KRAS(G12C) inhibitors in stability and affinity.^71,78–80 This dual advantage potency combined with improved pharmacokinetic and toxicity profiles has not been widely reported in the KRAS(G12C) inhibitor literature, particularly in the context of haematologic malignancies.

Conclusion

The current study employed an integrative computational approach to identify natural product-based inhibitors targeting the KRAS(G12C) mutation, a prominent driver of leukemogenesis in acute lymphoblastic leukaemia (ALL). A combination of molecular docking, induced-fit docking, pharmacokinetic (ADMET) analysis, molecular dynamics simulations, binding free energy calculations (MM/GBSA), and KD estimations led to the identification of six top-ranking compounds from African natural product databases. Although NA/EA-5 showed the most favourable KD, its dynamic stability was unverified due to the absence of MD/MM/GBSA analyses. NA/EA-3, with a comparable KD, outperformed across binding energy, 200 ns interaction stability, and safety/pharmacokinetic profiles, making it the overall lead. Other compounds also showed promising characteristics, in several cases outperforming the FDA-approved AMG-510 in binding affinity and toxicity predictions. By integrating static affinity predictions with dynamic and pharmacokinetic evaluations, this study identifies NA/EA-3 as the most promising overall lead, providing a strong foundation for subsequent in vitro and in vivo validation towards potential clinical use in ALL therapy. This outcome underscores that candidate prioritisation was driven by a balanced assessment of affinity, stability, safety, and drug-likeness rather than relying solely on KD values ensuring that the final lead selection reflects true therapeutic potential.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Consent to participate

All authors agreed to participate in this study.

Consent for publication

All authors agreed to publish in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All the generated data are included in the manuscript.