Comparing the effect of traditional and novel tyrosine kinase inhibitors for epidermal growth factor receptor exon 20 insertions by molecular dynamics simulation

Abstract

Background

Exon 20 insertions of epidermal growth factor receptor (EGFR) exhibit varying sensitivity to traditional and novel tyrosine kinase inhibitors for non-small cell lung cancer.

Methods

Molecular dynamics simulations were used to investigate the structural dynamics of wild-type EGFR and three representative exon 20 insertion mutations: V769_D770insASV (ASV), D770_N771insSVD (SVD), and N773_insNPH (NPH). Furthermore, the binding mechanisms of osimertinib and mobocertinib were explored to understand their efficacy across different EGFR variants.

Results

In this study, we found that binding to osimertinib and mobocertinib changed the free energy landscapes for ASV-, SVD- and NPH-EGFR. Compared with osimertinib, mobocertinib occupies extra space at the back of K745 for the three mutants and forms a hydrogen bond with the gatekeeper residue (T793) using its isopropyl ester group on the pyrimidine ring. When binding to osimertinib, ASV- and SVD-EGFR still revealed two energy minima on their free energy landscapes, but with considerably less conformational probability distribution at collective variable 2 >1.00 Å. In contrast, mobocertinib eliminated the energy minima at collective variable 2 >1.00 Å while decreasing the K745–E762 salt bridge formation rates.

Conclusions

Mobocertinib outperforms osimertinib in targeting specific subtypes of EGFR exon 20 insertions, highlighting its ability to restore the inactive state of this protein.

Background

Exon 20 insertion (ex20ins) mutation accounts for the third most common subtype of epidermal growth factor receptor (EGFR) mutation. ¹ EGFR ex20ins comprise more than 120 subtypes, with only a few located at the C-helix region (A763-M766). The majority subtypes are located at the loop region of EGFR exon 20, which is further classified into near-loop (A767 to P772) and far-loop (H773 to C775). ² V769_D770insX and D770_N771insX represent the most common subtypes, located at the near-loop of EGFR exon 20.^3,4 EGFR ex20ins (except A763_Y764insFQEA) are resistant to traditional EGFR–tyrosine kinase inhibitors (TKIs), with a median progression-free survival (mPFS) of 2.0–2.9 months. ⁵ Novel drugs such as mobocertinib and amivantamab have demonstrated clinical efficacy against EGFR ex20ins.^6,7 However, the objective response rate remains below 43%, with the mPFS not exceeding 8.3 months. Moreover, the MOON trial demonstrated no significant difference in the PFS for patients receiving mobocertinib treatment between near-loop and far-loop insertions. ⁸ Furthermore, compared with chemotherapy, mobocertinib monotherapy failed to demonstrate superior efficacy in the EXCLAIM-2 trial. These results indicate that not all patients benefit from novel TKIs owing to the molecular heterogeneity of EGFR ex20ins. However, other third or fourth generation EGFR–TKIs may demonstrate superior efficacy for specific subtypes due to their unique chemical structures.^9,10

The structural and functional impact of EGFR ex20ins remains poorly understood, particularly regarding their distinct conformational effects relative to wild-type (WT) EGFR and their interaction with traditional and novel EGFR–TKIs such as osimertinib and mobocertinib. The three most frequent subtypes, V769_D770insASV (ASV), D770_N771insSVD (SVD), and N773_insNPH (NPH), occur in the highly flexible loop regions of EGFR exon 20, making it challenging to capture their dynamic behavior using static structural methods. To clarify the structural difference among these mutations, innovative research is warranted to determine how they alter the conformational ensemble of EGFR, including the active and inactive states, as define their influence on drug binding.

Molecular dynamics (MD) simulations offer an effective strategy to explore these structural differences at an atomistic level, enabling the identification of mutation-specific conformational shifts, stability changes, and changes in allosteric communication pathways. This approach is particularly valuable for studying the binding mechanisms of osimertinib and mobocertinib, two inhibitors with varying efficacy across ex20ins variants. By simulating the interaction of these inhibitors with mutant EGFR, MD simulations can provide mechanistic insights into their differential binding modes, energetic profiles, and the structural adaptability of the ATP-binding pocket.

In this study, MD simulations were used to investigate the structural dynamics of WT-EGFR and three representative ex20ins mutations (ASV, SVD, and NPH). Furthermore, the binding mechanisms of osimertinib and mobocertinib were explored to understand their efficacy across different EGFR variants. This approach not only highlights the conformational consequences of these mutations but also elucidates how these structural alterations influence drug binding, providing critical insights for the rational design of next-generation EGFR inhibitors tailored to overcome ex20ins-driven drug resistance.

Methods

Structural preparation of protein and ligand

The WT EFGR apo-protein structure (PDB ID: 2GS2, ¹¹ with residue numbering adjusted by adding 24 to maintain consistency with most EGFR crystal structures) was obtained from the PDB database and subjected to protonation and energy minimization using the MOE software package. ¹² Based on the optimized WT cavity protein, amino acid insertions were introduced at the designated positions using the MOE Protein Mutation module. For each insertion type (ASV, SVD, and NPH), the WT structure was used as the modeling template, and homology modeling was performed via MOE to rebuild the local backbone and side-chain conformations around the insertion sites. The resulting insertion mutant structures were then energy-optimized to obtain reasonable and stable starting models.

For the ligands, the two-dimensional structural formulas of osimertinib and mobocertinib were obtained from the literature. The structures were converted into three-dimensional conformations and energy minimized using the MMFF94 force field in MOE, ¹³ yielding optimized ligand structures for subsequent simulations.

Molecular docking parameter setting

Using the molecular docking module of MOE software, the two abovementioned small molecules were docked with WT and three mutant cavity proteins. The specific parameters of the docking module were set as follows: all conformations of the small molecule obtained through the search were placed in the binding pocket via the Triangle Matcher method, the binding strength of the protein and the small molecule in the process was scored via the London dG algorithm (as explained in the MOE user manual), the 50 conformations with the highest scores were selected for the next step of the induced fit docking optimization operation, and the binding strength of the first 50 conformations and the protein was scored again using the GBVI/WSA dG algorithm. ¹⁴ Based on the order of energy, the last 20 binding conformations were output, and the lowest-energy binding conformation was selected to obtain the initial MD structure according to the binding mode of small molecules and proteins in the existing crystal structure.

MD parameter setting

MD was performed against the apo- and holo-protein systems using the AMBER18 software package, and subsequent analysis was conducted via built-in commands. ¹⁵ The simulation ended when structural root mean square deviations (RMSDs) reached convergence; thus, the simulation time varied for different proteins, as follows: 1 μs for apo-WT-EGFR, 500 ns for apo-ASV- and SVD-EGFR, 800 ns for apo-NPH-EGFR, 200 ns for holo ASV- and NPH-EGFR, and 300 ns for holo SVD-EGFR.

For specific parameter setting of MD simulation, the protein structure parameters were generated using the built-in ff14SB force field of AMBER18, ¹⁶ the small molecule parameters were generated using the GAFF2 force field, ¹⁷ and the AM1-BCC method in the antechamber program was used for small-molecule charge fitting. ¹⁸ The TIP3P water box model was used for all solvent models in this study, with the distance between the edge of the water box and the nearest atom of the solute set to 10.0 Å. ¹⁹ Each system was neutralized by the addition of Na⁺ and Cl⁻ counter ions to maintain charge neutrality. The exact number of water molecules and ions varied depending on the size of each system. First, to eliminate the unreasonable contact between atoms in the simulation system, two-step energy optimization was used to reduce the collision between atoms, and the 2500-step steepest descent method and 2000-step conjugate gradient method were used for optimization. Then, each system was maintained under a constant volume for 2 ns and gradually heated from −273°C to 27°C, and MD equilibrium was performed for 1 ns under atmospheric pressure and 1 standard atmospheric pressure. Finally, unconstrained MD simulations for the cavity system and the complex system were conducted. During the simulation, atomic coordinates were recorded every 2 ps, the system temperature was controlled by a Langevin thermostat, ²⁰ and long-range electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method with a cutoff of 10.0 Å.

Free energy calculation software setting

The binding free energy of osimertinib and mobocertinib with the three EGFR ex20ins subtypes was calculated using the Movable Type (MT) Free Energy Method, as developed by Zheng et al. ²¹ The MT method estimates the molecular partition function by decomposing configurational sampling into a set of local atomic pairwise interactions, which are statistically recombined to approximate the ensemble distribution. This approach eliminates the need for exhaustive conformational sampling while preserving the key thermodynamic contributions.

In this study, the MT_ScoreE protocol of the MT method was applied as a fast absolute binding free energy method capable of predicting protein–ligand binding affinity from accurate complex conformations. For each system, 500 conformations were extracted evenly from the 100-ns MD trajectories in their stable states, and the binding free energies were computed using the GARF scoring function, embedded within the MT method. ²² The MT method provides an efficient and reliable estimate of the overall binding free energy of the studied complexes.

Results

Structural modeling and MD simulations of EGFR ex20ins

We performed molecular modeling to analyze the molecular mechanisms underlying the conformational features of ASV-, SVD-, and NPH-EGFR mutants compared with WT-EGFR. Additionally, we compared the differences in the binding modes between two marketed drugs, osimertinib and mobocertinib, across EGFR variants. Our analysis also considered binding free energy profiles and binding mode differences for both apo- and holo-EGFR targets (Table 1). For each EGFR subtype structure (WT and variants), we collected conformations from the converged region of each simulation trajectory and projected them onto a Cartesian coordinate plane with two collective variables (CV) representing the αC-helix domain conformational change (Figure 1(a)). CV1 (along the x-axis) represents the distance between the Cα atom of K745 and the backbone heavy-atom centroid of the αC-helix front part (residues 753–759). CV1 illustrates how the αC-helix moves in and outward, swinging between the “αC-helix-in” and “-out” conformational states.

Table 1.

Probability distributions of CV1 and CV2 on the FEL for the apo-/holo-protein in the simulation.

CV1
	<12 (Å)	12–13 (Å)	13–14 (Å)	>14 (Å)
WT-apo	0.00%	3.07%	19.22%	77.71%
ASV-apo	3.37%	84.27%	12.35%	0.00%
SVD-apo	39.52%	55.32%	5.15%	0.00%
NPH-apo	0.00%	8.17%	58.99%	32.81%
ASV-osimertinib	0.11%	48.07%	51.35%	0.47%
SVD-osimertinib	3.33%	76.10%	19.93%	0.34%
NPH-osimertinib	1.24%	52.78%	39.43%	6.55%
ASV-mobocertinib	0.00%	44.68%	54.89%	0.43%
SVD-mobocertinib	0.25%	45.01%	54.33%	0.42%
NPH-mobocertinib	0.19%	49.57%	49.40%	0.84%
CV2
	<0.5 (Å)	0.5–1 (Å)	1–1.5 (Å)	>1.5 (Å)
WT-apo	6.86%	24.59%	50.35%	18.20%
ASV-apo	25.05%	43.46%	22.90%	8.60%
SVD-apo	16.48%	21.98%	52.99%	8.55%
NPH-apo	12.37%	45.59%	36.83%	5.20%
ASV-osimertinib	45.41%	49.74%	4.84%	0.01%
SVD-osimertinib	31.96%	57.54%	10.00%	0.51%
NPH-osimertinib	3.71%	77.84%	17.57%	0.00%
ASV-mobocertinib	15.54%	80.27%	4.15%	0.04%
SVD-mobocertinib	61.61%	38.26%	0.13%	0.00%
NPH-mobocertinib	7.26%	66.58%	25.59%	0.57%

CV: collective variables; FEL: free energy landscape; ASV: V769_D770insASV; SVD: D770_N771insSVD; NPH: N773_insNPH; WT: wild-type.

Figure 1.

Free energy surface of the apo- and holo-proteins for the wild-type EGFR and three variants.

The other CV (CV2 along the y-axis) captures the internal RMSD of all the backbone heavy atoms from the entire αC-helix domain (residues 753–768) relative to the initial conformation following system preparation, reflecting internal conformational change in the αC-helix during its departure from or approach to the EGFR activation segment. As shown in Figure 1(a), colors on the Cartesian coordinate plane represent the conformational free energy values, with lower values suggesting higher occurrence probability of a specific conformational feature.

Apo-protein conformational features

The free energy landscape (FEL) of apo-WT-EGFR showed that the αC-helix domain exhibits the largest range of internal conformation changes between the “αC-helix-in” and “αC-helix-out” states. The FEL also highlighted the vital conformational features of the apo-protein in the simulation (Figure 2(a)). The global minimal energy conformation ensemble was marked as state 1 on the FEL. K745 at the β3 strand and E762 at the αC-helix were too far apart to form the KE salt bridge. Instead, two residues from the activation loop formed two separate salt bridges with the K745 and E762 side chains (D855–K745 and K860–E762). The 1.10 Å internal RMSD for the αC-helix was attributed to partial unpacking of the front part of the αC-helix (N756–I759). N756 at the front part of the αC-helix forms two hydrogen bonds, using its side-chain amide group, with the V786 backbone amine and S752 backbone acyl group (Figure 1(b)).

Figure 2.

Representative conformational feature of the global (state 1) and local minimum states (state 2 and state 3) on the free energy landscape for the wild-type and the three subtypes of epidermal growth factor receptor exon 20 insertion mutations. The lower free energy value suggests higher occurrence probability of a certain conformational feature. The pink and the dotted lines represent the hydrogen bond interaction.

In addition to the global minimum, we discovered two local minima (states 2 and 3) on the FEL (Figure 2(a)). State 2 also showed the “αC-helix-out” feature, with the αC-helix domain leaning further outward according to the CV1 value. However, the front part of the αC-helix domain (N756–I759) formed an extended α-helix conformation in state 2, causing N756 to remain away from residues V786 and S752, whereas the β3‐αC loop adopted an extended conformation to accommodate the widely open “αC-helix-out” feature. Furthermore, in state 3, the αC-helix domain moved inward and remained close to the activation segment, with the formation of a KE salt bridge between K745 and E762. The apo-WT-EGFR acquired the “αC-helix-in” feature. Given this conformational feature, T751 in the β3‐αC loop approached F723 in the P-loop through CH–π interaction between the side chains of these two residues.

In contrast, the MD simulation showed that the three EGFR ex20ins apo-proteins exhibited more constrained and patterned αC-helix conformational distributions than apo-WT-EGFR (Figure 1(a)). The two near-loop insertion mutations generated similar conformational distributions for EGFR, with both featuring two “αC-helix-in” conformational minima. One of these insertion mutations exhibited a more extended αC-helix length from residue 752 to residue 768. The other mutation with residues 752–755 unpacked from the αC-helix domain, and the front part of the αC-helix (N756–I759) bent to approach the β-strands (Figure 1(b) and (c)). Moreover, both energy minima achieved a stable K745–E762 KE salt bridge (96.40% probability of formation for apo-ASV-EGFR and 96.74% for apo-SVD-EGFR). Conformations with CV2 >1.00 Å contain additional hydrogen bonds between N756/S752 and V789/A722, forming a “double-locker” interaction network, forcing the αC-helix to remain close to the EGFR activation segment (Figure 2(b) and (c)).

As a far-loop insertion mutant, NPH-EGFR retained greater αC-helix flexibility than the two near-loop insertion mutants but still showed a higher probability (57.9%) for the formation of the K745–E762 salt bridge than WT-EGFR. Different from the conformational features at the two near-loop insertion mutants, in NPH-EGFR, R748 at the β3‐αC loop was usually flipped toward the hydrophobic pocket between the αC-helix and the activation segment and interacted with the residues at the αC-helix and P-loop (Figure 1(b) and (c)). Three energy minima were identified (Figure 2(d)): (a) state 1, in which R748’s side chain forms a hydrogen bonding network at the front part of the αC-helix with E749, F723, and E758; (b) state 2, in which R748 extends deeper into the hydrophobic pocket and forms two salt bridges with E762 and E758; and (c) state 3, in which R748 moves toward the P-loop and enlarges the gap between the αC-helix and β-sheet domain.

Holo-EGFR ex20ins variants

Binding to osimertinib and mobocertinib altered the FELs for ASV-, SVD- and NPH-EGFR by reducing the conformational probability distributions (CPDs) when CV2 was >1.00 Å (Figure 1(d)). When binding to the two inhibitors, intraprotein interactions between the front part of the αC-helix and the β-sheet domain were weakened: residues 752–755 tended to wrap and form extended αC-helices, and these two inhibitors restrained the αC-helix inward and outward moving range. The αC-helix CPDs within 12 Å < CV1 < 14 Å increased, while the probability decreased when the αC-helix moved too close (CV1 < 12 Å) or too far from (CV1 > 14 Å) the activation segment. To demonstrate the internal conformation change in the simulations, we presented various minimal states on the FEL for the two drugs and the three mutants.

Compared with osimertinib, mobocertinib occupied additional space behind K745 for the three mutants and formed a hydrogen bond with the gatekeeper residue (T793) using the isopropyl ester group on the pyrimidine ring (Figure 1(f)), restraining the flexibility of the K745 side chain and pushing it toward the αC-helix. As a result, the αC-helix was stabilized at the abovementioned geometry with CV1 ≈ 13 Å.

When binding to osimertinib, although both ASV- and SVD-EGFR maintained two energy minima on their FELs, they exhibited markedly reduced CPD values at CV2 > 1.00 Å. The conformational ensemble collected from the converged MD trajectory indicated that the K745–E762 KE salt bridge became unstable compared with apo-ASV- and SVD-EGFR (50.72% for osimertinib–ASV-EGFR and 88.57% for osimertinib–SVD-EGFR). In contrast, mobocertinib eliminated the energy minima at CV2 > 1.00 Å while also decreasing the K745–E762 salt bridge formation rates by 64.03% for mobocertinib–ASV-EGFR and 57.07% for mobocertinib–SVD-EGFR relative to the corresponding apo-EGFR mutants. Such changes in the FEL revealed that mobocertinib is generally more effective in hindering the formation of hydrogen bonds between the front part of the αC-helix and the β-sheet domain.

Given that apo-NPH-EGFR exhibits a highly flexible R748 side chain, both inhibitors prevent R748 from extending into the ligand’s allosteric binding site. However, its guanidino side chain still forms hydrogen bonds with S720, G721, and A722 from the P-loop domain and pulls the P-loop away from the ligand’s binding site. Therefore, F723 at the P-loop hairpin turn deviates from the inhibitors and forms CH–π interactions with the hydrophobic residues from the αC-helix, significantly weakening the π–π interaction between F723 and the inhibitors. In contrast, mobocertinib forms a hydrogen bond with the gatekeeper residue T793 using its isopropyl ester group, achieving a better protein–ligand binding affinity than osimertinib (Figure 1(e)).

Free energy calculation

We used the MT free energy method to evaluate the binding free energy of the two drugs against the three EGFR ex20ins variants. Our calculation revealed that mobocertinib exhibits higher binding affinities toward mutants than osimertinib, indicating superior therapeutic effectiveness for all three EGFR ex20ins variants. These results are consistent with the reported half-maximal inhibitory concentration (IC₅₀) values in the cell line. The Pearson correlation coefficient between the Ba/F3 cell experimental data (pIC₅₀ (negative log of the IC50 value)) from the literature ⁶ and our free energy calculation was 0.86, confirming the accuracy of our computational model (Figure 3 and Table 2).

Figure 3.

Pearson correlation coefficient between the Ba/F3 cell experiment data (pIC₅₀) and the free energy calculation (−ΔG). pIC₅₀: negative log of the half-maximal inhibitory concentration value.

Table 2.

Ba/F3 cell experiment data (pIC₅₀) from existing literature and the free energy calculation.

	Osimertinib		Mobocertinib
	Experimentation(pIC50:M/L)	Calculation(kcal/mol)	Experimentation(pIC50:M/L)	Calculation(kcal/mol)
ASV	6.76	−10.93	7.96	−11.69
SVD	6.24	−10.39	7.65	−11.77
NPH	6.33	−11.20	7.74	−11.50

ASV: V769_D770insASV; SVD: D770_N771insSVD; NPH: N773_insNPH; pIC₅₀: negative log of the half-maximal inhibitory concentration value.

Discussion

EGFR is a transmembrane glycoprotein composed of 1186 amino acids, which consists of 3 parts: an extracellular domain, a transmembrane domain, and an intracellular domain. The intracellular domain represents the carboxyl terminal region of a protein kinase domain, which includes three subregions: tyrosine kinase region, near membrane region, and C-terminal region. The tyrosine kinase region has adenosine triphosphate binding sites that can activate downstream signaling pathways. Currently, we can only explain the underlying mechanism by which certain common mutations activate EGFR from a structural biology perspective. ²³ EGFR exists in both activated and inactive states. The major structural difference between these states lies in the conformational characteristics of the αC-helix domain and the extension or folding of the activated loop. In addition, differences in the orientation of key residues, such as KE salt bridge and DFG motif, contribute to state transitions. The activated state exhibits the conformational feature of “αC-helix in,” while the inactive state exhibits the conformational feature of “αC-helix out.” Under physiological circumstances, EGFR maintains a dynamic equilibrium between these two conformational states of EGFR. However, when oncogenic mutations such as EGFR L858R, 19del, and ex20ins occur, the balance shifts toward the activated state, leading to excessive activation of the EGFR gene and further activation of downstream signaling pathways, making it impossible to inhibit cell growth and ultimately promoting the occurrence of tumor lesions.

The common types of EGFR ex20ins are ASV, SVD, and NPH, which are insensitive to targeted drugs such as osimertinib and afatinib, with a response rate less than 10%. In 2021, mobocertinib was approved by the US Food and Drug Administration (FDA) for subsequent-line therapy for patients with non-small cell lung cancer (NSCLC) harboring EGFR ex20ins mutations who had progressed on platinum based chemotherapy regimens. Although osimertinib and mobocertinib are both EGFR–TKIs, their efficacy in EGFR ex20ins NSCLC varies considerably. In this study, we aimed to explore the binding ability of these two targeted drugs to different insertion sites of EGFR ex20ins through MD simulations to elucidate the structural basis underlying their therapeutic differences.

When comparing the MD simulation results of WT-EGFR and the three EGFR ex20ins subtypes (ASV-, SVD-, and NPH-EGFRs), we found that despite having similar structures, MD simulation revealed different conformational distributions, particularly in their αC-helix domains. The FELs of WT-EGFR indicated the largest conformational space for the αC-helix domain, suggesting the capability of apo-WT-EGFR to switch between inactive and active states. For the three apo insertion mutations, the αC-helix domain is locked in the active state, causing sustained activation of various downstream pathways, leading to the development and progression of various diseases, including cancer. This is especially true for the two near-loop insertion mutation variants (SVD- and ASV-EGFRs).

The comparative analysis of the WT protein with the three representative mutants revealed that the mutations disrupted the intrinsic equilibrium between the active and inactive conformations of the kinase, biasing the protein toward the active state. This conformational shift was accompanied by alterations in the binding pocket size and the probability distribution of active conformations. Such changes are expected to influence not only the binding affinity of inhibitors but also their ability to revert the mutant protein from the active to the inactive state. These mutation-induced effects may therefore contribute to differences in drug efficacy across distinct classes of inhibitors.

Regarding osimertinib and mobocertinib binding to the three insertion mutations, the CPD range of the αC-helix domain is also restricted; however, this restriction is less pronounced with mobocertinib than with osmertinib. Specifically, when binding to mobocertinib, ASV- and SVD-EGFRs exhibit a more extended and intact αC-helix. Compared with the two near-loop insertions, the far-loop insertion mutant NPH-EGFR retains greater αC-helix flexibility. The FELs for the holo-NPH-EGFRs revealed that osimertinib has limited capacity to stabilize the αC-helix, while mobocertinib more effectively restricts CPD along CV1 within the 12 Å–14 Å range. Therefore, compared with osimertinib, mobocertinib induces a shift in the conformational distributions of the αC-helix across all three insertion mutations and stabilizes it away from the activation segment. Consequently, the energy required for transition to the fully activated state increases, reducing the likelihood of sustained activation after drug dissociation and supporting the superior therapeutic performance of mobocertinib compared with osimertinib for diseases driven by the three EGFR ex20ins.

Our findings highlight that the isopropyl ester group of mobocertinib not only enhances the binding affinity through additional interactions with surrounding residues (e.g. T793) but also regulates the local conformation of Lys745. Consistent with this, FEL analysis revealed that mobocertinib more effectively hindered the formation of hydrogen bonds between the N-terminal region of the αC-helix and the β-sheet domain, thereby stabilizing the αC-helix at a greater distance from the activation segment. This combination of enhanced local interactions and conformational stabilization underpins its superior performance relative to osimertinib. These insights suggest that rational design of next-generation EGFR inhibitors could benefit from incorporating analogous substituents that occupy the pocket targeted by the isopropyl ester group, thereby leveraging a similar mechanism to improve efficacy against exon 20 insertion mutations.

This study has several limitations. First, our study focused on comparing the differences in the binding modes of mobocertinib and osimertinib to ASV, SVD, and NPH, without including other new drugs or a broader range of exon 20 insertion mutations. Second, at the time of MD simulation experiments, mobocertinib had not yet been delisted globally, which may affect therapeutic relevance. Third, this study is based on a simulation analysis and only provides preliminary mechanistic observations without experimental validation. Therefore, further studies are warranted to validate our findings at the protein level and through organoid models.

Conclusions

Mobocertinib outperforms osimertinib in targeting specific subtypes of EGFR ex20ins and demonstrates a superior ability to restore the inactive state of this protein. These findings advance the understanding of EGFR dynamics in the context of targeted therapies. Further research is warranted to explore the mechanistic diversity of next-generation EGFR–TKIs.