In silico analysis of PERK interactions with engineered alpha lactalbumin peptides using molecular dynamics simulations

Abstract

This study integrates docking, molecular dynamics (MD) simulation, and free energy estimation to evaluate two engineered peptides, mutated peptide 52 (MuP52) and mutated peptide 250 (MuP250), derived from α-lactalbumin (α-LA), as potential modulators of protein kinase RNA-like endoplasmic reticulum kinase (PERK). Molecular mechanics Poisson–Boltzmann surface area (MMPBSA) calculations revealed that MuP250 exhibited slightly stronger overall binding to mouse PERK (mPERK) (ΔG =  − 25.84 ± 5.27 kJ/mol) compared to MuP52 (ΔG =  − 24.13 ± 6.14 kJ/mol), despite forming fewer hydrogen bonds. Both peptides stabilized the PERK αG-helix region without inducing major structural perturbations, as indicated by analyses of RMSD, RMSF, Rg, SASA, and hydrogen-bonding patterns. In human PERK (hPERK), MuP52 maintained stable binding (ΔG =  − 22.79 ± 3.64 kJ/mol) and reduced local flexibility, suggesting potential cross-species relevance. Peptide binding also modulated the conformational landscape under conditions of endoplasmic reticulum (ER) stress, with principal component analysis (PCA) confirming that MuP250 induced the greatest stabilization in mPERK. These results support the structural plausibility of αG-helix engagement by α-LA–derived peptides in the context of Parkinson’s disease (PD) and highlight the roles of hydrogen-bonding and nonpolar interactions in stabilization. The findings are derived exclusively from MD simulations and future in vitro and in vivo validation will be required to confirm the biological relevance and therapeutic potential of these peptides.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-48319-3.

Introduction

The protein kinase RNA-like endoplasmic reticulum kinase (PERK), a central regulator of the unfolded protein response (UPR), has emerged as a potential therapeutic target in Parkinson’s disease (PD) due to its role in mitigating proteostasis stress in dopaminergic neurons^1,2. Sustained activation of PERK is associated with persistent phosphorylation of eukaryotic initiation factor 2 alpha (eIF2α), which attenuates global protein synthesis, intensifies endoplasmic reticulum (ER) stress, and may contribute to the degeneration of dopaminergic neurons, a hallmark of PD pathology³. Dysregulated PERK-eIF2α signaling has been implicated in key pathogenic processes in PD, including α-synuclein aggregation, mitochondrial dysfunction, and oxidative stress, all of which contribute to motor deficits and neuronal loss^4,5.

Most current strategies targeting PERK, including canonical ATP-competitive inhibitors such as GSK2606414, focus on the highly conserved ATP-binding pocket. These approaches can lead to broader off-target effects due to structural conservation among kinases. In contrast, our approach targets the αG-helix, a structurally defined substrate-docking site for eIF2α⁶, distinct from the ATP-binding pocket. This substrate-competitive strategy is novel and offers the potential for greater selectivity, a concept not yet systematically explored in PD intervention⁷.

Peptide-based strategies provide a promising avenue for selective modulation of PERK signaling. To establish a biological rationale for our peptide source, we selected bovine α-lactalbumin (α-LA) based on its well-documented properties: (i) the ability to cross the blood–brain barrier^8,9, (ii) anti-inflammatory and neuroprotective effects observed in PD models, supported by studies on camel α-LA or α-LA–oleic acid complexes^10,11, and (iii) GRAS (Generally Recognized as Safe) status and widespread use in nutraceuticals^12,13. Although direct experimental evidence for bovine α-LA in PD is limited, these camel α-LA studies provide strong biological plausibility for exploring α-LA-derived peptides in silico. Together, these properties support the use of α-LA as a scaffold for rational peptide design targeting the PERK αG-helix.

Recent work has highlighted peptides designed to influence components of the PERK/eIF2α/activating transcription factor 4 (ATF4) signaling axis, a critical node in ER stress responses implicated in neurodegenerative disorders^14–16. Computational approaches, including molecular docking and dynamics simulations, have also been employed to design small-molecule PERK inhibitors that bind the kinase domain, potentially suppressing overactivation and downstream eIF2α phosphorylation^17,18. Complementary strategies have engineered mutated peptides to target phosphorylated eIF2α directly^19,20, often using residue substitutions to enhance proteolytic stability^21,22.

Given the central role of PERK-eIF2α signaling in PD, there is growing interest in peptide-based approaches to modulate this pathway^4,23. Such strategies are attractive due to their potential for high specificity and have been explored in other biological contexts, including antiviral pathogenesis^24,25. To date, no study has systematically investigated food-derived peptides for direct targeting of the αG-helix in PD⁷.

The methodological innovation of this work lies in a rational mutation strategy guided by integrated computational analysis. A computational framework not previously applied to PERK-targeting peptide design was employed. Specifically, this involved per-residue molecular mechanics Poisson–Boltzmann surface area (MMPBSA) decomposition energy analysis coupled with ΔΔG binding-affinity predictions from the MCSM (Membrane Curvature Server for Molecular Dynamics) server to design affinity-enhancing mutations. Furthermore, the central hypothesis was that a rationally engineered α-LA-based peptide would bind the αG-helix of PERK and modulate its interaction with eIF2α. An integrated computational pipeline, including molecular docking, MD simulations, and binding free energy calculations was used. This approach aimed not only to identify a high-affinity peptide but also to provide mechanistic insights into a substrate-competitive interaction at the αG-helix, distinct from ATP-binding inhibitors. Therefore, the present study does not aim to demonstrate functional inhibition of PERK, but rather to computationally explore the structural feasibility of α-LA–derived peptides interacting with the αG-helix as a non-ATP-competitive site. By integrating docking, MD simulation, and free energy estimation approaches, we sought to generate a hypothesis-driven framework that may guide future experimental validation. The conclusions of this study are limited to in silico predictions and should be interpreted within that scope.

Computational approaches

Peptide library construction and structural prediction

The α-LA (PDB code: 1f6r (https://www.rcsb.org/structure/1f61)) served as the parent sequence for constructing the peptide library. The FASTA sequence of α-LA was used to generate the peptide library with the same number of residues in overlapping sequences. For 3D structural predictions of each library peptide, the I-TASSER (Iterative Threading ASSEmbly Refinement) server (https://zhanggroup.org/I-TASSER/) was used.

Competitive inhibition assessment via molecular docking

To assess the ability of the peptides to inhibit PERK by competing with αG of the α-helix at its substrate-binding site, the HADDOCK (High Ambiguity Driven Bimolecular Docking (https://rascar.science.uu.nl/)) server was employed. The three-dimensional structures of both α-LA and mouse PERK (mPERK) (retrieved from the PDB entries 1f6r and 3qd2 (https://www.rcsb.org/structure/3qd2), respectively) were uploaded into the platform. Based on the scoring, clusters with the top-ranked dockings were selected as the most probable conformational models representing potentially suitable peptide-PERK interaction complexes.

Safety screening of peptide candidates

To exclude peptide candidates with allergenic or toxic properties, comprehensive screening was performed. This step is particularly important for older adults, who are often more prone to hypersensitivity and adverse drug interactions. To maximize therapeutic benefits and minimize health risks for patients, allergenicity and toxicity assessments were conducted using the AllerTOP and ToxinPred. These tools are based on advanced machine learning frameworks, including k-nearest neighbors (k-NN) for allergen prediction and support vector machine (SVM) classifiers for toxicity assessment.

Molecular dynamics (MD) simulation

To investigate the structural behavior of peptide–PERK complexes, MD simulations were performed using GROMACS 2019.1 (10.5281/zenodo.2564761) software with the CHARMM36 force field to generate topology and coordinate files for both the protein and peptides. Each system was placed in a cubic simulation box and solvated with SPC water molecules, maintaining a minimum 1.0 nm buffer distance from the box edge. Charge neutrality was achieved by introducing 0 sodium (Na⁺) and 2 chloride (Cl⁻) counter ions, resulting in an ionic concentration of approximately 0.01 M. Energy minimization was conducted using the steepest descent method for 500,000 steps with an energy step size of 0.01 and a time step of 2 fs, until the maximum force fell below 1000 kJ/mol^-1nm^-1. Equilibration was carried out in two phases containing the NVT ensemble for 1 ns at 300 K using the V-rescale thermostat followed by the NPT ensemble for 2 ns at 1 bar pressure using the Parrinello-Rahman barostat. Long-range electrostatics were treated using the Particle Mesh Ewald (PME) method with a cutoff of 1.2 nm, and bond lengths involving hydrogen atoms were constrained using the LINCS algorithm. Production runs were performed for 150 ns with a time step of 2 fs, and trajectories were recorded every 10 ps for subsequent analysis.

This setup allowed evaluation of structural stability, residue-level interactions, and binding free energy calculations under physiologically relevant conditions while ensuring charge neutrality and appropriate solvent environment. For comparative purposes, the same MD simulations were also conducted on the human PERK (hPERK) structure (PDB ID: 4G31(https://www.rcsb.org/structure/4G31)). It allows us to evaluate the species-specific effects in the αG-helix binding region while maintaining the focus on the mouse template as the primary model. To preliminarily assess the conservation of interactions across species, one of the two final peptide candidates was also simulated in complex with the human PERK structure (PDB: 4G31) under identical conditions. Three independent 150-ns MD simulations were conducted for each peptide–PERK complex (mouse and human).

Molecular dynamics (MD)- based structural analyses

This study employed a set of standard parameters calculated from the MD simulation trajectory to analyze the system’s structural dynamics and stability. These parameters included Root-Mean-Square Deviation (RMSD) to assess overall structural stability^26,27, Root-Mean-Square Fluctuation (RMSF) to investigate local residue flexibility^28,29, Radius of Gyration (Rg) to measure molecular compactness^30,31, Solvent-Accessible Surface Area (SASA) to evaluate the solvent-exposed surface^32,33, and Principal Component Analysis (PCA) to identify and extract the essential collective motions of the system^34,35. The PCA Analysis was performed using a custom Python script (pc2plot.py). The code and detailed instructions for reproducibility are available as supplementary material Table S1.

MMPBSA binding energy calculations

The MMPBSA method widely estimates binding free energies in bimolecular complexes^36,37. It combines molecular mechanics (MM) with implicit solvation models (PBSA/GBSA) to compute contributions from gas-phase interactions (), solvation effects (), and entropy (-T∆S)³⁸. The overall binding free energy is calculated as:

1

A negative indicates favorable binding. MMPBSA balances computational efficiency and accuracy but requires careful interpretation due to approximations (e.g., implicit solvent, rigid binding)^36–38. Total binding energies were computed for both mPERK and hPERK complexes to assess the reproducibility of peptide binding and identify potential species-specific variations in interaction energies.

Mutagenesis

The ΔΔG stability metric was utilized for systematically introducing mutations in the peptide sequence through MCSM. The thermodynamic impact of single-point mutations was investigated based on ΔΔG value, which is the difference in binding free energy between wild-type and mutant peptides, as defined by equation:

2

This parameter serves as a predictor of structural stability changes, where positive ΔΔG values denote mutations that enhance peptide stability, negative values indicate destabilizing mutations, and values approaching zero (ΔΔG ≈ 0) represent neutral mutations with minimal stability impact.

The comprehensive sequence libraries were generated for all favorable mutations identified through MCSM analysis in the case of each residue position in the peptide sequence.

Results and discussion

Structure of αG in PERK

PERK has emerged as a promising therapeutic target for PD due to its central role in regulating proteostatic stress responses in dopaminergic neurons⁴. Chronic PERK activation leads to sustained phosphorylation of eIF2α, which exacerbates ER stress, impairs global protein synthesis, and contributes to α-synuclein aggregation, a hallmark of PD neuropathology³⁹. Building on evidence that milk-derived bioactive peptides can modulate proteostatic dysfunction in neurodegeneration^40,41, we investigated α-LA, a key milk protein involved in lactose synthesis⁴², as a source of potential PERK-modulating peptides. Structural modifications of α-LA may influence its interaction with PERK, a central mediator of the UPR. Within the PERK kinase domain, the αG-helix represents a structurally and functionally important regulatory element⁷. This region is positioned within the C-lobe of the kinase domain, adjacent to the activation loop, and contributes to conformational stability as well as substrate recognition. The αG-helix has been implicated in mediating interactions required for proper substrate docking, including the engagement of eIF2α⁷. Therefore, perturbation or stabilization of this region may influence PERK signaling without directly targeting the conserved catalytic core. Most reported PERK inhibitors primarily act through ATP-competitive mechanisms, binding to the catalytic ATP-binding pocket and suppressing kinase activity (e.g., GSK2606414)⁴³. While effective, ATP-competitive inhibitors often face limitations related to selectivity due to the conserved nature of kinase ATP-binding sites. In contrast, allosteric strategies that target regulatory structural elements offer the potential for greater selectivity and reduced off-target effects^44,45. In this context, the αG-helix represents an attractive alternative site for modulation, as it is spatially distinct from the ATP pocket yet functionally connected to substrate interaction and kinase conformational dynamics. In this study, we engineered α-LA-derived peptides designed to interact with the αG-helix and stabilize its conformation through contacts with the PERK C-lobe. The PERK kinase domain consists of an N-lobe and a catalytic C-lobe⁷, with the C-lobe containing key functional elements including the ATP-binding region, activation loop, and the strategically positioned αG-helix. This helical motif participates in inter-lobe communication and contributes to regulation of kinase structural states. Therefore, our structure-based design focuses on the αG-helix as a non-canonical regulatory site to optimize α-LA-derived peptides as novel modulators of PERK, with potential therapeutic relevance for PD.

Peptide library

A set of 24 primary peptides (PrPs) was derived from the α-LA sequence (Table 1). These peptides were computationally docked into the identified binding site (Fig. 1) using the HADDOCK platform. HADDOCK is an information-driven docking approach that ranks solutions using a weighted scoring function combining van der Waals, electrostatic, and desolvation energy terms⁴⁶. The HADDOCK score reflects the overall interaction energy of the complex, where more negative values indicate more favorable binding. The scoring scheme and weighting strategy are defined within the HADDOCK2.2 framework developed by the Bonvin laboratory. Docked complexes were ranked according to the HADDOCK score, and the template peptide was selected based on the lowest score together with the consistency of intermolecular contacts within the dominant docking clusters. This combined energy- and cluster-based ranking strategy provides a reliable criterion for identifying favorable binding modes prior to MD simulations⁴⁶. The selected complexes were subsequently subjected to MD simulations and MMPBSA calculations to further evaluate binding stability and interaction strength. Through this workflow, PrP11 (YGLFQINNKIW) was identified as the leading template for subsequent optimization, exhibiting the most favorable docking score and stable interaction pattern at the targeted site (Table 1).

Table 1.

Template candidate sequences and their HADDOCK docking scores.

Residues No	Residues	Docking scores
PrP1	EQLTKCEVFR	– 46.8 4.8
PrP2	KCEVFRELKDL	– 47.5 5.5
PrP3	RELKDLKGYGG	– 68.2 3.4
PrP4	LKGYGGVSLPE	– 51.3 3.8
PrP5	GVSLPEWVCTT	– 49.8 0.3
PrP6	EWVCTTFHTSG	– 64.0 6.1
PrP7	TFHTSGYDTQA	– 67.1 3.2
PrP8	GYDTQAIVQNN	– 58.9 1.2
PrP9	AIVQNNDSTEY	– 51.8
PrP10	NDSTEYGLFQI	– 68.9 4.5
PrP11	YGLFQINNKIW	– 77.0 1.9
PrP12	INNKIWCKDDQ	– 56.0
PrP13	WCKDDQNPHSS	– 50.2 3.1
PrP14	QNPHSSNICNI	– 51.5 1.3
PrP15	SNICNISCDKF	– 45.0
PrP16	ISCDKFLDDDL	– 57.5 1.0
PrP17	FLDDDLTDDIM	– 62.1 2.3
PrP18	LTDDIMCVKKI	– 41.2
PrP19	MCVKKILDKVG	– 47.2 2.9
PrP20	ILDKVGINYWL	– 58.0 1.3
PrP21	GINYWLAHKAL	– 48.4 2.0
PrP22	LAHKALCSEKL	– 40.7 7.7
PrP23	LCSEKLDQWLC	– 55.7 4.8
PrP24	LDQWLCEKL	– 52.4 2.1

Significant values are in bold.

Fig. 1.

Three-dimensional (A) and two-dimensional (B) representations of the interaction between PrP11 and the αG-helix in PERK, illustrating effective blockage of the binding site. The representations in (A), and (B) were visualized using Discovery Studio Visualizer 2021 (https://www.3ds.com/products-services/biovia/products/discovery-visualization/discovery-studio-visualizer/), and LigPlus (LigPlot + ; v2.2, EMBL-EBI (https://www.ebi.ac.uk/thornton-srv/software/LigPlus/)), respectively. Color code: A the PrP11 peptide segment (green) B carbon (black), nitrogen (blue), oxygen (red).

The interaction site between PrP11 and the αG in PERK was investigated and indicated in Fig. 1 3D (A) and 2D (B) including H-bonding domains; and effective interactions.

Mutated peptide

Although PrP11 demonstrated strong binding affinity, a 150-ns MD simulation combined with MMPBSA decomposition analysis indicated that the residues No. 1, 2, 4, 5, 7, and 8 exhibited positive binding energies. This observation suggests that these specific residues contributed suboptimally to the overall binding free energy (Supplementary Table S2). Since hydrogen bonds play a vital role in protein–protein interaction⁴⁷, we performed the hydrogen bond occupancy analysis. The results revealed that TRP60, PHE53, and GLN54 highlighted in red are capable of forming a strong connection with the receptor through hydrogen bonds compared to other hotspots (Supplementary Table S3). This suggests that this residue positively contributes to template binding and was kept without replacement. To optimize molecular binding, we strategically targeted residues with energetically unfavorable interactions for mutagenesis. Residues 1, 2, 7, and 8 were selected for mutation, excluding those involved in hydrogen bonding. The MCSM tool was employed to perform these sequence-specific mutations using the ΔΔG framework, which evaluates the energy changes associated with amino acid substitutions. Notably, for residue No.2, no favorable mutant was identified. This analysis identified multiple stabilizing mutations (characterized by positive ΔΔG values) at each targeted position, suggesting improved peptide binding stability. All potential sequence combinations derived from these favorable mutations were then generated to assess their inhibitory efficacy. The 396 mutated peptides (MuPs) were considered based on mutation in PrP11 and filtered based on allergenicity and toxicity predictions (Supplementary Table S4). Finally, subsequent allergenicity and toxicity screening of the mutant sequences yielded 80 non-toxic, non-allergenic candidates listed in Table 2. To further prioritize inhibitory potential, the peptides were docked into the investigated enzyme’s binding cavity. Among the remaining candidates and the initial unmutated template, the MuP250, and MuP52 exhibited the highest binding scores, – 80.2 4.7 and – 77.4 3, respectively. Key predicted biophysical properties of these selected peptides, including solubility, proteolytic stability, aggregation propensity, and cell-penetrating potential, are summarized in Supplementary Table S5.

Table 2.

HADDOCK Docking results of Mutated Peptides and αG in mPERK.

Peptide ID	Docking Score	RMSD from the overall lowest-energy structure (Å)	Van der Waals Energy (kcal/mol)	Electrostatic Energy (kcal/mol)	Desolvation Energy (kcal/mol)	Buried Surface Area (Å)
MuP12	-50.2	4.5 0.5	– 22.9 0.6	– 77.8 14.1	– 14.4 1.0	876.2 26.4
MuP14	– 59.7 5.4	1.4 1.0	– 24.0 3.2	– 137.9 4.4	– 9.2 3.2	914.5 146.5
MuP18	– 50.4 3.6	3.4 0.1	– 22.3 2.2	– 84.4 1.8	– 14.4 1.3	951.9 66.1
MuP19	– 63.5 0.4	3.6 0.0	– 24.4 0.9	– 132.2 8.7	– 13.5 1.6	938.6 11.0
MuP21	– 60.1 4.6	2.3 0.2	– 30.9 3.7	– 48.8 3.7	– 21.4 1.6	1061.0
MuP23	– 60.9 1.2	3.5 0.0	– 27.1 0.7	– 124.9 6.0	– 9.6 0.8	946.1 20.4
MuP32	– 63.6 2.7	1.3 0.9	– 29.0 4.4	– 1121.1 2.5	– 12.4 4.1	998.7 43.8
MuP45	– 62.5 1.5	2.4 1.5	– 27.6 1.1	– 113.7 10.8	– 13.5 0.5	992.6 6.4
MuP46	– 55.2 4.8	1.5 0.9	– 32.0 3.9	– 45.2 14.4	– 15.0 3.4	995.1 56.3
MuP47	– 61.6 4.0	1.5 0.9	– 29.0 3.9	– 107.0 9.7	– 11.8 2.9	960.1 73.4
MuP48	– 72.6 1.9	1.0 0.6	– 31.4 3.0	– 124.5 13.4	– 17.2 1.6	1103.6 69.5
MuP49	– 75.1 1.5	1.5 0.9	– 35.3 4.4	– 113.8 23.2	– 17.3 2.2	1105.8 33.0
MuP50	– 63.7 4.4	1.3 0.8	– 33.1 4.4	– 89 42.3	– 13.8 13.8	1029.9 45.0
MuP51	– 65.8 3.8	1.7 1.3	– 29.7 1.7	– 103.7 13.5	– 15.9 4.4	1032.6 95.8
MuP52	– 77.4 3	1.3 0.8	– 34.2 5.4	– 132.8 32.4	– 17.6 5.3	1075.2 61.8
MuP53	– 73.5 4.6	1.5 1.0	– 34.7 2.6	– 126.4 7.1	– 14.5 4.3	1086.5 41.9
MuP54	– 70.6 3.1	0.8 0.5	– 35.8 2.0	– 104.3 21.0	– 14.8 3.3	1066.3 33.6
MuP58	– 64.3 2.8	1.0 0.6	– 27.3 2.7	– 156.2 13.0	– 8.6 3.5	1006.5 70.1
MuP66	– 69.3 3.8	1.4 1.0	– 32.8 3.1	– 111.1 18.4	– 14.6 3.9	994.6 61.1
MuP68	– 54.7 1.4	3.8 0.3	– 29.2 1.4	– 77.2 15.4	– 11.3 1.9	903.9 54.9
MuP77	– 70.7 3.1	1.6 1.1	– 26.3 5.0	– 153.5 12.1	– 14.5 2.8	1077.8 69.1
MuP89	– 57.3 0.7	0.0 0.0	– 27.8 0.5	– 105.2 5.7	– 10.6 0.8	992.9 15.6
MuP91	– 56.4 5	1.1 0.7	– 26.0 1.6	– 100.8 8.9	– 10.7 4.3	948.9 54.7
MuP94	– 60.3 3	4.3 0.0	– 33.2 2.2	– 131.0 7.6	– 6.0 0.5	868.2 16.2
MuP98	– 67.4 3.8	1.9 1.2	– 31.1 5.9	– 125.7 24.2	– 12.3 1.4	1059.2 44.3
MuP111	– 54.5 1.4	0.1 0.1	– 23.1 0.6	– 95.2 13.3	– 13.9 0.9	939.7 17.9
MuP122	– 60.5 1.9	0.1 0.1	– 33.1 2.9	– 148.1 30.4	– 5.1 2.8	906.4 29.8
MuP128	– 55.2 0.2	25.5 39.8	– 39.8 21.5	– 75.1 25.5	– 20.8 11.7	1144.4 304.9
MuP155	– 56.7 2.3	1.3 0.8	– 32.7 5.0	– 58.9 13.1	– 13.7 3.3	1053.3 42.0
MuP161	– 59.7 1.9	3.7 0.0	– 33.6 2.0	– 125.7 20.9	– 6.0 2.7	882.1 20.0
MuP201	– 63.5 2.5	1.9 0.3	– 28.4 2.1	– 117.7 28.5	– 13.5 5.8	954.0 50.0
MuP208	– 60.6 2.7	1.7 1.3	– 32.2 3.2	– 83.5 26.5	– 13.8 4.3	1045.4 49.6
MuP210	– 53.7 2.9	0.2 0.2	– 23.1 0.7	– 84.3 22.3	– 14.3 2.4	936.4 18.6
MuP211	– 56.4 5.0	2.2 0.4	– 28.1 2.8	– 71.7 22.7	– 14.5 1.2	873.7 51.7
MuP212	– 69.7 3.9	1.0 0.7	– 30.9 3.5	– 137.6 18.3	– 13.3 1.9	1065.4 49.2
MuP215	– 54.9 5.4	2.1 0.4	– 27.3 3.5	– 100.0 22.6	– 9.5 1.7	889.6 26.3
MuP220	– 67.0 1.5	1.2 0.7	– 28.9 1.8	– 134.8 13.4	– 11.8 2.7	1002.7 29.8
MuP221	– 57.4 4.3	1.6 0.5	– 31.3 1.9	– 63.0 12.5	– 15.2 3.6	1026.2 55.1
MuP227	– 62.9 0.9	2.2 0.2	– 27.8 3.3	– 105.9 20.4	– 16.2 1.2	980.9 51.4
MuP231	– 76.0 1.4	1.0 0.6	– 34.1 1.1	– 140.7 12.6	– 14.1 1.7	1073.4 23.2
MuP243	– 61.8 2.3	3.1 0.4	– 27.1 4.4	– 112.0 19.9	– 13.5 2.5	1012.2 44.5
MuP244	– 56.1 0.5	3.7 0.0	– 28.0 1.9	– 82.2 4.0	– 13.4 0.5	976.9 40.6
MuP245	– 61.2 3.3	1.4 0.8	– 28.2 1.7	– 104.7 14.9	– 12.6 4.2	945.6 73.7
MuP246	– 74.4 2.1	1.2 0.7	– 32.5 2.3	– 132.0 21.8	– 16.0 3.7	1090.3 32.2
MuP248	– 70.4 3.3	1.1 0.8	– 36.3 2.6	– 100.9 20.9	– 14.8 1.3	1026.6 35.7
MuP249	– 70.3 4.9	1.4 0.8	– 35.3 2.6	– 93.8 21.5	– 17.0 2.8	1026.3 46.3
MuP250	– 80.2 4.7	2.3 1.3	– 34.3 4.6	– 157.2 14.2	– 15.3 1.4	1085.5 29.5
MuP251	– 72.5 3.7	1.8 1.1	– 34.7 8.1	– 107.5 22.1	– 17.1 1.9	1078.5 56.8
MuP252	– 69.9 2.0	1.3 0.9	– 32.9 32.9	– 111.5 18.2	– 16.1 1.8	1082.5 50.4
MuP254	– 54 4.1	4.3 0.0	– 21.5 0.8	– 118.1 10.4	– 10.1 3.2	901.8 46.7
MuP256	– 72.3 5.1	1.9 1.2	– 30.8 1.9	– 151.6 19.3	– 11.8 5.0	1055.6 26.6
MuP259	– 68.7 1.6	1.4 0.9	– 30.5 3.6	– 164.4 11.8	– 7.7 4.2	1036.8 51.1
MuP260	– 54.1 9.4	1.8 1.4	– 26.6 4.4	– 76.8 29.5	– 14.1 4.8	935.4 49.6
MuP264	– 70.1 2.0	1.3 1.1	– 33.3 4.0	– 125.9 18.8	– 12.1 2.0	1009.9 33.3
MuP269	– 71.4 6.4	1.3 1.0	– 29.7 3.8	– 120.4 24.0	– 18.1 2.6	1073.4 76.8
MuP275	– 73.3 2.4	1.0 0.6	– 30.6 1.1	– 132.2 10.3	– 17.5 2.2	1052.6 47.6
MuP286	– 69.0 3.0	1.9 1.2	– 26.8 4.1	– 141.3 17.2	– 14.2 0.8	1013.2 30.7
MuP287	– 56.2 1.2	3.8 0.0	– 26.9 1.6	– 96.1 14.7	– 11.5 1.8	974.3 7.4
MuP288	– 67.1 0.6	0.0 0.0	– 35.7 0.0	– 43.5 0.0	– 23.3 0.0	1083.2 0.0
MuP289	– 55.7 0.9	4.2 0.5	– 28.8 6.3	– 76.1 30.0	– 12.9 2.2	928.1 43.5
MuP290	– 59.9 1.3	0.0 0.0	– 27.4 0.9	– 115.5 2.0	– 9.9 0.5	940.4 13.2
MuP292	– 56.1 2.1	3.3 0.8	– 28.1 1.5	– 95.4 8.4	– 9.7 1.3	957.3 21.6
MuP293	– 59.5 7.7	1.3 0.8	– 27.4 1.8	– 83.6 22.1	– 17.0 3.6	1055.2 66.6
MuP294	– 64.0 0.8	0.0 0.0	– 24.6 0.4	– 137.8 0.2	– 12.3 0.6	964.4 14.8
MuP295	– 60.3 0.8	0.0 0.0	– 29.6 0.3	– 117.1 1.8	– 7.8 0.6	945.9 18.1
MuP296	– 64.2 4.6	2.1 1.6	– 28.1 4.5	– 126.4 27.9	– 14.2 4.4	1061.0 11.8
MuP300	– 57.6 6.1	2.5 0.4	– 26.3 1.1	– 107.9 29.2	– 11.1 3.4	937.6 48.6
MuP309	– 56.5 3.2	2.8 0.6	– 23.8 0.7	– 104.7 5.0	– 13.1 2.6	933.4 21.1
MuP315	– 56.3 3.2	3.3 0.1	– 34.5 2.3	– 37.9 6.8	– 15.7 1.2	1014.4 30.1
MuP333	– 65.3 5.9	1.7 0.3	– 28.7 5.9	– 130.5 27.0	– 11.3 4.2	949.2 142.7
MuP342	– 69.2 8.3	1.6 1.1	– 33.2 4.2	– 93.5 8.8	– 18.3 4.5	1043.9 39.6
MuP344	– 61.7 3.3	1.1 0.7	– 32.6 2.8	– 69.1 3.4	– 16.0 2.2	1015.5 50.4
MuP345	– 74.7 4.1	1.7 1.1	– 32.0 4.8	– 138.7 25.9	– 15.4 2.3	1079.4 35.2
MuP348	– 64.0 3.9	2.0 1.2	– 32.3 2.0	– 92.7 7.0	– 13.6 3.8	1037.8 30.1
MuP351	– 66.4 2.4	0.8 0.6	– 29.9 2.8	– 106.5 12.9	– 15.6 3.3	1062.0 57.8
MuP353	– 61.6 3.1	0.1 0.0	– 35.5 3.2	– 112.7 2.8	– 7.5 0.5	903.8 4.8
MuP355	– 63.5 1.4	2.3 0.3	– 27.2 3.5	– 136.2 9.6	– 12.0 1.9	1032.0 44.6
MuP386	– 58.6 1.9	0.4 0.4	– 25.9 3.3	– 106.3 3.0	– 13.5 2.1	998.5 38.1
MuP388	– 60.4 5.0	1.5 1.1	– 27.1 2.6	– 85.8 6.1	– 17.0 3.0	994.7 58.2
MuP389	– 58.6 0.6	0.0 0.0	– 27.3 0.6	– 111.2 4.2	– 9.7 0.5	939.2 17.7
MuP395	– 69.3 4.7	1.3 0.9	– 32.7 1.5	– 106.5 28.0	– 16.9 3.5	1070.2 83.8

Significant values are in bold.

Thus, the two peptides were selected, and the structural mutations they underwent are as follows:

MuP52 (YGLFQINNKIW) → (HGLFQIERKIW).

MuP250 (YGLFQINNKIW) → (HGLFQIDRKIW).

The detailed two-dimensional interaction diagrams of these peptides with the αG-helix of mPERK, generated from the top HADDOCK poses, are presented in Supplementary Fig. S1 (for MuP52) and Fig. S2 (for MuP250).

Different diagrams of mPERK and selected mutant peptides (MuPs)

Evaluation of interaction energy profiles

The final peptide candidates, MuP52 (HGLFQIERKIW) and MuP250 (HGLFQIDRKIW), were derived from the wild-type template through structure-guided mutagenesis. Both peptides share Y1H and N8R substitutions and differ at position 7 with N → E (MuP52) and N → D (MuP250), respectively. These targeted mutations introduced charged residues intended to enhance electrostatic complementarity within the mPERK αG-helix binding site (residues 1025–1033: RVRIITDVR). Based on the sequences, positively charged residues in the peptides, such as Arg8, were expected to form stable hydrogen bonds with acidic residues like Asp1031, while polar residues could engage in complementary interactions with nearby polar residues of mPERK.

Initial analysis of the top HADDOCK poses using LigPlot + provided static interaction diagrams (Supplementary Fig. S1 and S2 for mPERK/MuP52 and mPERK/MuP250, respectively), outlining foundational hydrogen-bonding networks and electrostatic contacts. These diagrams represent the initial, static view of peptide–protein interactions prior to MD simulations. Subsequent triplicate 150-ns MD simulations captured the dynamic behavior of these complexes.

For the mPERK–MuP52 complex, the overall stable binding pose is shown in Fig. 2A, while conformational alignment across three independent replicates is presented in Fig. 2B. Representative PyMOL-derived interactions (Fig. 2C a–d) indicate persistent contacts: subpanel a, GLU56:CG—ARG1033:HH1 at 2.37 Å; subpanel b, ARG57:HA—ARG1033:NH1 at 3.68 Å; subpanel c, ASP1031:OD2—ARG57:NH2 at 2.94 Å; and subpanel d, ARG1033:NH1—Ile59:HD2 at 3.6 Å. These interactions align with sequence-based predictions, as positively charged Arg residues in the peptide complement negatively charged Asp1031 and polar residues of the αG-helix. Atomistic validation using Discovery Studio confirmed key stabilizing hydrogen bonds, including Arg8—Asp1031 (1.8 and 2.7 Å) and Arg8—Thr1030 (3.6 Å), which maintain MuP52 within the αG-helix binding site.

Fig. 2.

Structural representations and interaction analyses of the mPERK αG-helix in complex with MuP52. A Three-dimensional view of the overall binding pose. B Superimposition of peptide conformations from three independent MD replicates, illustrating the conformational spread at the binding site. C Representative snapshots of intermolecular interactions observed during the simulations, including hydrogen bonds and van der Waals contacts (panels a–d), not restricted to only the most persistent interactions. All 3D structures in (A), and (B) were rendered using PyMOL v2.5 (https://pymol.org/2/), and representative snapshots in C (a–d) were visualized using Visual Molecular Dynamics (VMD) v1.9.2 (https://www.ks.uiuc.edu/Research/vmd/). In the 3D representations C (a–d), peptide atoms are colored by element: carbon (cyan), nitrogen (blue), oxygen (red), and hydrogen (white).

For the mPERK–MuP250 complex, the binding pose was also stable (Fig. 3A) and conformationally consistent across replicates (Fig. 3B). Representative PyMOL-derived interactions (Fig. 3Ca–d) include ASP1031:OD2—ARG57:HH11 at 2.20 Å, HR1030:HG21—Ile55:O at 3.25 Å, HR1030:HG21—ASP56:O at 2.47 Å, and ASP1031:OD2—ARG57:HH22 at 1.91 Å. These contacts match sequence-based predictions, where polar and charged residues favor hydrogen bonding, and hydrophobic residues (like Ile and Trp) engage in stabilizing hydrophobic interactions. Static HADDOCK/LigPlot + diagrams indicate foundational contacts, while atomistic analysis using Discovery Studio revealed additional stabilizing hydrogen bonds (Trp60—Arg1033 at 1.8 and 1.9 Å) and hydrophobic interactions, such as a pi-sigma contact between Trp60 and Ile994.

Fig. 3.

Structural representations and interaction analyses of the mPERK αG-helix in complex with MuP250. A Three-dimensional view of the overall binding pose. B Superimposition of peptide conformations from three independent MD replicates, illustrating the conformational spread at the binding site. C Representative snapshots of intermolecular interactions observed during the simulations, including hydrogen bonds and van der Waals contacts (panels a–d), not restricted to only the most persistent interactions. All 3D structures in (A), and (B) were rendered using PyMOL v2.5 (https://pymol.org/2/), and representative snapshots in C (a–d) were visualized using Visual Molecular Dynamics (VMD) 1.9.2 (https://www.ks.uiuc.edu/Research/vmd/). In the 3D representations C (a–d), peptide atoms are colored by element: carbon (light blue), nitrogen (blue), oxygen (red), and hydrogen (white).

Together, the multi-scale convergence of static docking, dynamic simulations, and atomistic measurements provides robust validation of the peptide design strategy. The observed interactions are consistent with sequence-based predictions: MuP52 exhibits a balanced polar and ionic interaction profile, whereas MuP250 is primarily stabilized through hydrophobic anchoring. This distinction explains the enhanced binding affinities of both mutant peptides relative to the wild-type template. Overall, PyMOL-derived interactions highlight dynamic contacts, LigPlot + provides static baselines, and Discovery Studio confirms atomistic precision, collectively validating the structural rationale and stability of the designed peptides.

To assess the translatability of the results to the human system, the interaction of MuP52 with hPERK was analyzed, and the findings are presented in Supplementary Fig. S3. Panel A shows the overall stable binding pose of MuP52 at the αG-helix site of hPERK, while panel B presents the superimposition of the peptide from three independent MD replicates, confirming the dynamic stability of the binding mode over the simulation time.

Representative PyMOL-derived interactions from the final simulation frame are summarized in panel C (Fig. S3 a–d). Key contacts include: GLU1025:OE2 forming a strong hydrogen bond with LYS58:HZ3 at 1.68 Å, GLU1025:OE1 interacting with LYS58:HZ3 at 2.28 Å, J1025:HG1 engaging LYS58:HZ1 at 5.58 Å, and J1025:HG2 contacting LYS58:NZ at 5.15 Å. Collectively, these interactions stabilize MuP52 at the αG-helix interface, mirroring the hydrogen-bonding and electrostatic complementarity observed in the mouse system.

The arrangement of hydrogen bonds and longer-range interactions demonstrates that the designed mutations maintain the peptide in a stable binding conformation throughout the simulation, restricting excessive conformational freedom while allowing necessary local flexibility. This supports the consistency of peptide–protein interactions across species and highlights that acidic residues in the αG-helix and positively charged residues in MuP52 are central to interface stabilization.

Overall, this analysis confirms the relevance of the human model and validates the general translatability of the mouse findings. The PyMOL-derived interactions serve as representative examples of the key stabilizing contacts, providing a mechanistic basis for the peptide’s consistent behavior in both mPERK and hPERK systems.

Evaluation of RMSD, RMSF, SASA, Rg

The structural stability and dynamic behavior of mPERK in its apo state and in complex with the designed peptides MuP52 and MuP250 were evaluated through triplicate 150-ns MD simulations. The analysis encompassed global stability, compactness, solvent accessibility, and local flexibility, as summarized in Supplementary Table S6.

RMSD analysis revealed distinct conformational stability profiles across the three systems (Fig. 4A, panels a–c). The apo mPERK system exhibited the lowest and most stable RMSD trajectory with a mean value of 0.40 ± 0.04 nm, indicating a rigid native conformation. The MuP52-bound complex demonstrated slightly higher but consistent RMSD values (0.45 ± 0.06 nm), with good reproducibility across replicates, reflecting the formation of a stable interaction network at the peptide-binding interface. In contrast, the MuP250-bound system showed greater inter-replicate variability (0.43 ± 0.10 nm), suggesting comparatively higher conformational flexibility.

Fig. 4.

Comparative dynamic analysis of the αG region of mPERK in its apo form (green) and in complex with MuP52 (blue) and MuP250 (red) across three independent MD replicates. Time evolution of A RMSD, B Rg, C SASA, and D RMSF via residue number profiles.

Rg analysis indicated that all systems maintained stable global compactness throughout the simulations (Fig. 4B, panels a–c). The MuP52-bound system had a mean Rg of 2.13 ± 0.05 nm, essentially identical to the apo form (2.13 ± 0.03 nm), while the MuP250 complex showed a slightly lower mean Rg (2.11 ± 0.02 nm). The narrow standard deviations confirm that peptide binding does not induce large-scale unfolding or structural expansion.

SASA analysis showed that all systems reached stable plateaus (Fig. 4C, panels a–c). The MuP250 complex presented a marginally higher mean SASA (156.02 ± 1.97 nm²) compared to MuP52 (154.57 ± 4.95 nm²) and the apo state (153.45 ± 4.54 nm²). However, the overlapping standard deviations indicate that these differences are subtle and do not reflect major changes in global solvent exposure.

Local backbone flexibility, assessed via RMSF, showed that peptide binding reduces mobility in specific regions (Fig. 4D, panels a–c). The apo system exhibited the highest mean RMSF (0.18 ± 0.04 nm), while MuP52 and MuP250 complexes showed reduced fluctuations (0.15 ± 0.02 nm and 0.16 ± 0.03 nm, respectively), indicating localized stabilization upon peptide binding. This observation is consistent with the hydrogen-bonding networks and PCA clustering, which further confirm that MuP52 produces the strongest local and global stabilization, whereas MuP250 provides moderate stabilization.

In summary, the integrative dynamic analysis demonstrates that both engineered peptides stabilize local regions of mPERK without perturbing its global fold. The MuP52 complex exhibits slightly greater conformational rigidity and lower fluctuation, consistent with a more stable interaction network at the binding interface. These computational findings provide a mechanistic basis for the differential dynamic behavior of the two peptides, and support the structure-based design strategy.

The conformational stability and dynamic behavior of hPERK in its apo form and in complex with the designed peptide MuP52 were evaluated through triplicate 150‑ns MD simulations, with key statistics summarized in Supplementary Table S7. RMSD analysis (Fig. 5A, panels a–c) showed that the apo hPERK system exhibited a stable trajectory with a mean RMSD of 0.26 ± 0.02 nm, whereas the MuP52-bound complex displayed a slightly higher yet highly reproducible mean RMSD of 0.27 ± 0.02 nm. This indicates that peptide binding induces only subtle global conformational adjustments without compromising overall structural integrity. Rg analysis (Fig. 5B, panels a–c) confirmed retention of compact, folded states in both systems, with a mean Rg of 1.98 ± 0.01 nm for apo and 2.01 ± 0.02 nm for the MuP52 complex, highlighting preservation of global compactness. SASA trajectories (Fig. 5C, panels a–c) remained stable and nearly superimposable, with mean values of 140.61 ± 2.71 nm² and 140.56 ± 1.22 nm² for apo and MuP52-bound states, respectively, indicating minimal perturbation of solvent exposure. Local backbone flexibility, assessed via RMSF (Fig. 5D, panels a–c), showed reduced regional mobility upon MuP52 binding (0.12 ± 0.01 nm) relative to the apo form (0.15 ± 0.02 nm), reflecting a stabilizing effect on specific regions of the kinase domain.

Fig. 5.

Comparative dynamic analysis of hPERK in the apo form (green) and in complex with MuP52 (blue) across three independent MD replicates. Time evolution of A RMSD, B Rg, C SASA, and D RMSF via residue number profiles.

A cross-species comparison with mPERK highlights both conserved and species-specific features. In both organisms, the apo forms of PERK exhibit the highest flexibility, with larger RMSD and RMSF values, indicating that the unbound kinase domain samples a broad conformational space. MuP52 binding stabilizes both mPERK and hPERK, as evidenced by reduced backbone fluctuations (RMSF) and lower variation in global compactness (Rg) and solvent exposure (SASA). Notably, the absolute RMSD, Rg, and SASA values are systematically lower for hPERK, reflecting a more compact native state and slightly reduced structural plasticity compared to the mouse system. The stabilizing effect of MuP52 is consistent across species, with mPERK showing a slightly stronger restraint, likely due to differences in intrinsic flexibility and hydrogen-bonding networks. Interfacial hydrogen bonds contribute significantly to stabilization, with both the number and geometry of these interactions correlating with the observed reductions in structural fluctuations. Overall, MuP52 acts as a well-behaved, structured stabilizer in both systems, constraining local and global dynamics while maintaining overall structural integrity.

Evaluation of hydrogen bonding profiles

The hydrogen-bonding interactions between the engineered peptides and the αG-helix of mPERK and hPERK were analyzed across three independent 150-ns MD simulations (Fig. 6A–C, Supplementary Table S8).

Fig. 6.

Intermolecular hydrogen bond analysis of peptide–PERK complexes. Time evolution of hydrogen bonds formed between the peptides and the αG region during triplicate MD simulations. A mPERK–MuP52 complex, B mPERK–MuP250 complex, and C hPERK–MuP52 complex.

For mPERK, the MuP52-bound system consistently formed a higher number of hydrogen bonds throughout the simulations compared to MuP250. Across the three replicas, the number of H-bonds in MuP52 fluctuated between 2 and 10, with a mean of 2.80 ± 0.73, indicating a stable and persistent hydrogen-bonding interface. In contrast, MuP250 formed fewer hydrogen bonds, ranging from 1 to 7, with greater temporal variability and a mean of 2.34 ± 1.34, suggesting a more transient interaction. This difference highlights that MuP52 establishes a more robust polar interface with mPERK, potentially contributing to enhanced structural stabilization, as reflected in lower RMSD and Rg fluctuations, reduced local flexibility (RMSF), and more consistent solvent exposure (SASA) (Table S6). Interestingly, although MuP250 forms slightly fewer hydrogen bonds than MuP52 in mPERK, it may achieve binding stability through complementary nonpolar interactions, such as π–σ and alkyl contacts, which contribute to overall binding energetics. In contrast, MuP52 combines stable hydrogen bonds with favorable hydrophobic interactions, resulting in an optimized interface geometry and more consistent structural stabilization.

For hPERK, MuP52 formed fewer hydrogen bonds overall, fluctuating between 0 and 7, with a mean of 1.34 ± 0.96, reflecting a more dynamic and transient interface compared to the mouse system (Fig. 6C, Table S8). Despite the lower number of hydrogen bonds, MuP52 binding still contributed to interface stabilization, consistent with observed RMSD, Rg, RMSF, and SASA profiles (Table S7).

The comparative analysis across species demonstrates that mPERK complexes sustain stronger and more persistent hydrogen-bonding interactions than hPERK. Nonetheless, both peptides in mPERK and MuP52 in hPERK maintain a functionally relevant level of hydrogen bonding, which acts as a complementary stabilizing factor rather than the primary determinant of global structural behavior.

Overall, the hydrogen-bonding analysis demonstrates a clear correlation between interfacial stability and global/local protein dynamics. A stable H-bond network in mPERK–MuP52 restrains overall structural fluctuations (RMSD), maintains compactness (Rg), suppresses RMSF, and optimizes solvent accessibility (SASA). A more transient H-bond network in mPERK–MuP250 allows greater structural variability and local flexibility. In hPERK, the MuP52 complex exhibits a more dynamic interfacial profile, yet trends in structural stabilization remain consistent with those observed in the mouse system, highlighting that binding affinity and stabilization are determined not only by the number of hydrogen bonds but also by the balance and geometry of polar and nonpolar interactions at the peptide–αG interface.

Profile of principal component analysis (PCA)

Principal component analysis (PCA) was performed to characterize the global conformational landscape of mPERK in its apo form and in complex with MuP52 and MuP250 peptides. The PCA results for MuP52-mPERK, MuP250-mPERK complexes, and mPERK (apo) are presented in Fig. 7A–C, respectively, with each panel a within the figures showing all three independent replicas (replicas 1–3).

Fig. 7.

Principal Component Analysis (PCA) of mPERK conformational dynamics. PC1–PC2 projections derived from the last 10 ns (1000 frames) of triplicate MD simulations. A Apo mPERK, B mPERK–MuP52 complex, and C mPERK–MuP250 complex. In each panel, the three replicas are shown separately (a–c) to illustrate conformational sampling and clustering behavior. Data analysis and visualization were performed using Python 3.12 (https://docs.python.org/3.12/).

In Fig. 7A (apo mPERK, panel a; replicas 1–3), the three replicas occupy a broad and highly dispersed region of the PCA space, indicating substantial conformational heterogeneity and intrinsic flexibility. Minimal overlap among the replicas reflects significant structural variability in the apo state, highlighting that without peptide binding, mPERK explores a larger conformational space with higher atomic fluctuations, less compactness, and greater solvent exposure.

In contrast, in Fig. 7B (MuP52-bound mPERK, panel a; replicas 1–3), all three replicas cluster tightly within a restricted region of the PCA space, demonstrating that MuP52 binding constrains conformational freedom and promotes sampling of a shared, stable conformational basin. This compact clustering is consistent with the stable hydrogen-bonding network formed by MuP52, which reduces structural fluctuations, stabilizes local regions, and results in a more rigid and compact conformational ensemble, in agreement with the lower RMSF (0.15 ± 0.02 nm), stable Rg (2.13 ± 0.05 nm), and reduced SASA fluctuations (154.57 ± 4.95 nm²).

For the MuP250-bound system (Fig. 7C, panel a; replicas 1–3), each replica shows partial overlap, but the overall distribution is broader than that observed for MuP52, indicating moderate structural restraint while still allowing greater conformational variability. This broader PCA spread corresponds to more transient hydrogen-bonding interactions in the MuP250 complex, leading to higher flexibility and structural variability, as reflected in its intermediate RMSF (0.16 ± 0.03 nm), slightly elevated SASA (156.02 ± 1.97 nm²), and modestly variable Rg (2.11 ± 0.02 nm).

The PCA results are fully consistent with the structural parameters summarized in Supplementary Table S6, including RMSD, RMSF, Rg, and SASA, confirming that systems with lower atomic fluctuations, more compact structures, and reduced solvent exposure occupy more restricted regions of PCA space, whereas more flexible systems like apo mPERK display broader conformational sampling.

The relationship between hydrogen-bonding and PCA further demonstrates that MuP52, which forms a stable and persistent H-bond network, induces a tighter and more stable conformational ensemble, while MuP250, forming more transient H-bonds, allows greater flexibility. Overall, the PCA analysis highlights that peptide association modulates the conformational landscape of mPERK in a peptide-specific manner, with MuP52 producing the strongest stabilization by restricting conformational sampling and reducing structural variability, in complete agreement with the hydrogen-bonding analysis.

Principal component analysis (PCA) was carried out to characterize the global conformational behavior of hPERK in its apo form and in complex with MuP52, as shown in Fig. 8A–B, with each panel a displaying all three independent replicas (replicas 1–3). In Fig. 8A (apo hPERK, panel a; replicas 1–3), the replicas occupy a relatively broad and dispersed region of the PC1–PC2 space with limited overlap, indicating substantial conformational heterogeneity and intrinsic flexibility in the absence of peptide binding. The time-colored single-replica plots reveal wide temporal spread, consistent with continuous exploration of multiple conformational sub-states. This broad dispersion aligns with higher RMSF values, slightly elevated SASA, and a less constrained structural profile, confirming that apo hPERK samples a large conformational space and exhibits significant structural variability.

Fig. 8.

Principal Component Analysis (PCA) of hPERK conformational dynamics. PC1–PC2 projections derived from the last 10 ns (1000 frames) of triplicate MD simulations. A Apo hPERK and B hPERK–MuP52 complex. In each panel, the three independent replicas are shown separately (a–c) to illustrate conformational sampling and clustering behavior. Data analysis and visualization were performed using Python 3.12 (https://docs.python.org/3.12/).

In contrast, Fig. 8B (MuP52-bound hPERK, panel a; replicas 1–3) shows a more compact and convergent cluster of replicas, indicating that MuP52 binding restricts conformational freedom and promotes sampling within a shared, well-defined conformational basin. The temporal evolution within each replica remains dynamic but is confined to a narrower region of PC space compared with apo, reflecting stabilized flexible dynamics. This tighter clustering corresponds with reduced backbone fluctuations (lower RMSF) and modest changes in Rg and SASA, indicating enhanced structural stabilization. The PCA contraction is consistent with a stable hydrogen-bonding network, where MuP52 forms persistent H-bonds that reinforce the global conformational restraint of hPERK, although the stabilizing effect is slightly less pronounced than observed in the mouse system.

Overall, these PCA results are fully consistent with RMSD, RMSF, Rg, and SASA analyses, demonstrating that peptide binding by MuP52 narrows the accessible conformational space and reduces local flexibility. The apo system exhibits the broadest PCA distribution with minimal overlap among replicas, reflecting the inherent flexibility of the unbound kinase domain, whereas the peptide-bound system shows more restricted and overlapping clusters, indicating stabilization. Comparisons with the mouse system reveal that MuP52 exerts a stronger stabilizing effect on mPERK, whereas hPERK responds with a more uniform but slightly less pronounced reduction in conformational freedom.

Taken together, these analyses highlight that peptide-mediated stabilization of PERK is conserved across species, with subtle differences in magnitude and peptide-specific effects. In the human system, the correlation between hydrogen-bonding and PCA demonstrates that formation of a stable network of H-bonds by MuP52 contributes directly to global conformational restraint, supporting a mechanistic link between local interactions and global dynamics.

A cross-species comparison between mPERK and hPERK reveals both conserved and species-specific features in the dynamics of the kinase domains. In both systems, the apo forms exhibit the greatest conformational flexibility, with mPERK showing a slightly broader conformational space than hPERK. This is reflected in higher RMSD and RMSF values for mPERK, indicating a more flexible native state. In contrast, hPERK demonstrates a more compact native structure, with lower RMSD and RMSF values, suggesting reduced conformational plasticity.

When bound to MuP52, both mPERK and hPERK exhibit stabilized conformations, but with notable differences. MuP52 binding induces a stronger stabilization effect in mPERK, leading to tighter PCA clustering, reduced RMSF, and more stable global compactness (Rg), as compared to the human system. MuP52 binding also results in a more persistent and stable hydrogen-bonding network in mPERK, contributing to stronger overall structural restraint. In hPERK, MuP52 binding results in a more dynamic interface with slightly lower hydrogen-bonding persistence, but still leads to notable stabilization of local regions and a more compact structure.

In summary, while both MuP52 and MuP250 stabilize PERK in mouse, MuP52 exerts the most pronounced stabilization effect across species, particularly in mPERK, with MuP250 contributing to intermediate stabilization. The differential impact on local flexibility and overall compactness between the two peptides is consistent with their hydrogen-bonding patterns and PCA clustering, demonstrating that peptide binding modulates PERK dynamics in a species- and peptide-specific manner.

The average MMPBSA binding energies

MMPBSA analyses for the unmutated template (PrP11) are provided in Supplementary Table S9, whereas results for the mutated complexes, including MuP52 and MuP250 in the mouse system and MuP52 in the human system, are summarized in Supplementary Table S10. MMPBSA analysis was conducted to evaluate the binding energetics of mPERK in complex with MuP52 and MuP250 peptides across mPERK and with MuP52 across human systems. For the mouse system, the PrP11 template was used as reference. MuP52 binding exhibited strong electrostatic interactions (− 146.64 ± 79.36 kJ/mol) and favorable van der Waals contributions (− 36.90 ± 13.32 kJ/mol), partially counterbalanced by a substantial polar solvation term (+ 165.14 ± 68.30 kJ/mol), resulting in a total binding energy of ΔG =  − 24.13 ± 6.14 kJ/mol. MuP250 showed similar trends with slightly stronger van der Waals stabilization (− 42.46 ± 5.27 kJ/mol) and moderately reduced electrostatics (− 120.63 ± 10.05 kJ/mol), yielding a total ΔG =  − 25.84 ± 5.27 kJ/mol. In the human system, MuP52 displayed van der Waals energy of − 39.77 ± 1.53 kJ/mol, weaker electrostatics (− 63.91 ± 16.86 kJ/mol), and a polar solvation contribution of + 16.65 ± 88.43 kJ/mol, resulting in a total binding energy of ΔG =  − 22.79 ± 3.64 kJ/mol. The large standard deviation of the polar solvation term reflects transient solvent exposure and orientation fluctuations of the peptide relative to the protein, rather than dissociation or instability of the complex.

Due to statistical correlation among energy components in MMPBSA calculations, the sum of individual component means does not always exactly match the reported total binding free energy. This behavior, particularly evident when the polar solvation component exhibits large fluctuations, is a common feature of MMPBSA calculations and has been reported in the literatures^48,49. Such behavior is not indicative of errors or peptide dissociation, and all calculations have been verified to be stable across the trajectory frames. Overall, the binding energetics are consistent with structural analyses, including RMSD, RMSF, Rg, SASA, and hydrogen-bonding patterns. Apo PERK exhibits higher flexibility and conformational heterogeneity, whereas peptide binding, especially MuP52, reduces atomic fluctuations, stabilizes the αG region, and constrains the accessible conformational space as observed in PCA analysis. In mPERK, binding is dominated by strong electrostatics counterbalanced by polar solvation, whereas in hPERK, van der Waals interactions contribute more prominently to stabilizing the complex.

These results highlight species-specific energetic landscapes of PERK-peptide interactions and justify the observed standard deviations and energy distributions in the MMPBSA calculations. The minor differences in total ΔG between mPERK and hPERK complexes, despite the high sequence similarity of the αG-helix and identical peptide sequence (MuP52), reflect subtle species-specific effects while confirming that the computed binding energetics are consistent and reliable.

Mechanistic distinction from ATP‑competitive PERK inhibitors

The peptide designed in this study modulates PERK through a binding mode fundamentally different from classical small‑molecule PERK inhibitors such as GSK2606414. GSK2606414 is a potent ATP-competitive inhibitor that binds the catalytic ATP-binding pocket of PERK and suppresses its kinase activity, reducing downstream eIF2α phosphorylation⁴³. The crystal structure of PERK in complex with GSK2606414 (PDB 4G31) provides direct structural evidence for this ATP-competitive mechanism. In contrast, the peptide engages the αG-helix of PERK, a distinct region critical for docking the native substrate eIF2α^43,50. Because these ligands interact with non-overlapping sites and inhibit PERK via different molecular principles, direct quantitative comparison of binding energetics or interaction fingerprints is not mechanistically meaningful. This qualitative comparison highlights the complementary nature of substrate-competitive peptides, emphasizing selective disruption of PERK–eIF2α interactions rather than catalytic blockade.

The present computational analyses provide insights into the structural engagement of MuP52 and MuP250 peptides with the PERK αG-helix, highlighting peptide-specific and species-specific variations in conformational dynamics and binding energetics. These observations suggest structural plausibility of αG-targeted modulation, which must be interpreted cautiously in the absence of experimental validation.

Limitations and future perspectives

Several limitations of the present study should be acknowledged. First, molecular docking provides predicted binding poses based on scoring functions and does not confirm physical interaction under physiological conditions. In addition, molecular dynamics simulations assess structural stability within defined force-field parameters but cannot establish functional inhibition or downstream signaling modulation. Moreover, MM/PBSA binding energy calculations rely on approximations such as implicit solvation models and limited entropic sampling, which may affect quantitative accuracy.

Importantly, the predicted interactions were not validated experimentally, as no biochemical kinase assays, phosphorylation measurements, or cellular ER stress reporter experiments were performed. Experimental assessment of peptide bioavailability, proteolytic stability, and blood–brain barrier permeability was also lacking. Therefore, the proposed α-lactalbumin–derived peptides should be regarded as computational candidates that require systematic in vitro and in vivo validation before any biological or therapeutic conclusions can be drawn.

Conclusion

Our computational analyses identified MuP52 and MuP250, derived from bovine α-LA through targeted mutations, as promising candidates for allosteric modulation of the PERK αG-helix. Molecular docking, molecular dynamics (MD) simulations, and MMPBSA calculations consistently show stable binding of these peptides within the αG site, with MuP250 slightly higher than MuP52 in the mPERK system. In mPERK, both MuP52 and MuP250 reduced conformational flexibility, stabilized the αG region, and promoted more compact structural ensembles, as evidenced by RMSD, RMSF, PCA, SASA, and hydrogen-bonding analyses. In hPERK, MuP52 binding also restricted conformational dynamics and stabilized the αG region. However, the energetic contributions differed from those in mPERK, with van der Waals interactions playing a more prominent role in stabilization. These results highlight the peptide-specific and species-specific nature of PERK modulation. Collectively, this work provides a computational foundation for a novel strategy to modulate PERK, a key target in ER stress pathways implicated in PD pathogenesis. These results are based solely on computational analyses. No direct conclusions regarding downstream biological effects, such as mitigation of α-synuclein aggregation or neuronal protection, can be drawn. Experimental validation through in vitro kinase assays, ER stress reporter cell assays, and in vivo models will be required to confirm the biological activity and therapeutic relevance of these peptides. Overall, this computationally based study provides a framework for designing αG-targeted PERK modulators and nominates MuP52 and MuP250 as lead candidates for allosteric regulation. Nevertheless, because these findings lack experimental validation, no conclusions about them in vivo efficacy can be made.

Author contributions

Giti Pishehvarz: Software, Methodology, Formal analysis, Visualization, Writing—original draft, Investigation. Reza Zarghami: Writing—review & editing, Methodology, S. Mohsen Asghari: Writing—review & editing, Methodology. Alireza Bahramian: Conceptualization, Writing—review & editing, Methodology, Investigation, Funding.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This molecular dynamics simulation study utilized exclusively publicly available protein structures from the RCSB PDB database (IDs: 1f6r and 3qd2, and 4G31). No human/animal subjects, biological samples, or laboratory experiments were involved. Therefore, ethical approval was not required under University of Tehran regulations and international standards for computational research.