Quantum Reactivity-Guided Optimization of ATP-Competitive Ligands through Multiscale Simulation

Abstract

ATP-binding pockets impose stringent geometric and electronic constraints on small-molecule ligands, requiring preservation of key anchoring interactions while permitting only limited scaffold modification. Rational optimization of ATP-competitive inhibitors therefore remains challenging, particularly when electronic reactivity and metabolic liability must be balanced against binding stability. We explored a multiscale, quantum-informed computational strategy to address this challenge using the receptor tyrosine kinase-like orphan receptor 1 (ROR1) pseudokinase domain, an underexplored intracellular cancer target that retains a druggable ATP-binding pocket, as a chemically informative model system. Using Ponatinib as an internal reference scaffold, eight Fukui-guided analogs (DM1– DM8) were designed through density functional theory analyses integrating Mulliken, Löwdin and Hirshfeld charge partitioning with Fukui function mapping. These calculations identified electronically reactive and metabolically labile regions amenable to site-specific modification at C10, C11, C32, and C35, while preserving the alkyne and carbonyl motifs responsible for anchoring interactions within the ATP-binding pocket. To establish a robust energetic and dynamical baseline, 1 μs all-atom molecular dynamics simulations were performed for the ROR1 pseudokinase domain in both apo and Ponatinib-bound states. The designed analogs were subsequently evaluated using 250 ns simulations combined with relative MM-PBSA binding free-energy analysis, revealing a narrow binding-energy range (ΔG_bind ≈ −37 to −45 kcal mol^–1) comparable to Ponatinib (−41.78 ± 5.65 kcal mol^–1 at 250 ns; −40.23 ± 5.61 kcal mol^–1 at 1 μs). Interaction fingerprint analysis confirmed conservation of the hinge-anchored Glu71-Ile103-Asp181 network and the Phe100-Tyr102 aromatic clamp across the series. In silico ADME profiling differentiated developability among energetically similar compounds, identifying DM3–DM5 as the most balanced analogs. Collectively, quantum-mechanical reactivity descriptors can delineate electronically feasible scaffold modification sites within ATP-binding pocketspositions where chemical perturbation is tolerated without disrupting anchoring pharmacophore interactionsproviding substantively complementary information distinct from, classical energy-based scoring or empirical SAR.

1. Introduction

The re-emergence of embryonic proteins in adult malignancies presents a compelling opportunity for selective cancer therapy, as such targets often exhibit tumor-restricted expression with minimal presence in normal adult tissues. , Receptor tyrosine kinase–like orphan receptor 1 (ROR1), a transmembrane member of the receptor tyrosine kinase family, is primarily expressed during embryogenesis and early development but is highly restricted in normal adult tissues. , This selective expression profile makes ROR1 an exceptionally attractive and tumor-specific therapeutic target. − Aberrant re-expression of ROR1 has been reported across multiple malignancies, including chronic lymphocytic leukemia (CLL), ,,, triple-negative breast cancer (TNBC), − nonsmall-cell lung cancer (NSCLC), − mantle cell lymphoma (MCL), , pancreatic ductal adenocarcinoma (PDAC), and ovarian serous carcinoma, highlighting its broad relevance in oncogenic signaling. −

Functionally, ROR1 promotes tumor progression through multiple interconnected mechanisms. It mediates Wnt5a-dependent noncanonical signaling that activates oncogenic transcriptional programs such as STAT3 and PI3K/AKT, − while also performing kinase-independent scaffolding functions that enhance PI3K/AKT signaling via interactions with cavin-1 and caveolin-1 − and sustain RAS/MAPK pathway activation through receptor cross-talk particularly in drug-resistant cancer cells. These activities collectively promote tumor cell survival, metastasis, and therapeutic resistance, and elevated ROR1 expression correlates with poor prognosis and reduced responsiveness to targeted therapies. ,

To date, therapeutic strategies targeting ROR1 have primarily focused on its extracellular domain, leading to the development of monoclonal antibodies (e.g., cirmtuzumab, h1B8, h6D4), antibody–drug conjugates (ADCs) such as VLS-101 (zilovertamab vedotin), − bispecific T-cell engagers, , and CAR-T cell therapies. , These biologics have demonstrated encouraging preclinical and early clinical efficacy. For example, VLS-101 employs a cathepsin B-cleavable linker to deliver monomethyl auristatin E (MMAE) selectively to ROR1-expressing tumor cells, producing potent antitumor activity in non-Hodgkin lymphoma models and durable clinical responses in the waveLINE-001 trial (NCT03833180). ,, While these biologic modalities underscore the therapeutic promise of ROR1-directed interventions, antibody-based approaches are inherently limited to surface-accessible epitopes and may be affected by antigen shedding, heterogeneous expression, and restricted intracellular access. Consequently, the intracellular pseudokinase domain of ROR1 remains comparatively underexplored as a druggable target.

Pseudokinases, which comprise nearly 10% of the human kinome, were historically considered undruggable due to their lack of catalytic activity and the presence of noncanonical binding pockets. − However, emerging evidence indicates that many pseudokinases function as conformational switches or signaling scaffolds that can be modulated by small molecules through stabilization of specific structural states. − Accordingly, the pseudokinase domain of ROR1 represents an unconventional yet tractable drug target. High-resolution crystal structures of the kringle domain (PDB: 6BA5, 1.6 Å) and pseudokinase domain (PDB: 6TU9) have revealed ligandable cavities and autoinhibited conformations. The extracellular region mediates ligand recognition and dimerization through immunoglobulin-like, Frizzled/Cysteine-Rich (CRD), and kringle domains, ,, whereas the intracellular pseudokinase domain, although catalytically inert, retains a conserved VAIK motif and performs critical scaffolding roles that integrate oncogenic signaling pathways. ,

Crystallographic evidence (PDB: 6TU9) further demonstrated that Ponatinib, a multitargeted type II kinase inhibitor, can occupy the vestigial ATP-binding cleft of the ROR1 pseudokinase domain, inducing conformational stabilization and reducing hydrogen–deuterium exchange across regulatory loops. Despite Ponatinib’s broad kinase promiscuity and cardiotoxicity, it serves as a valuable chemical probe for mapping conformational vulnerabilities and defining pharmacophoric hotspots within the pseudokinase domain. , Additional small molecules, including compound 19h (KD = 0.10 μM)which induces apoptosis in lung and breast cancer cells and KAN0441571C, currently in early clinical development for ibrutinib-resistant CLL, , further support the feasibility of pharmacological modulation of the ROR1 intracellular domain. Collectively, these studies establish a clear proof-of-concept for small-molecule intervention at the ROR1 pseudokinase domain while highlighting the need for selective and rationally optimized inhibitors that preserve ROR1 affinity without incurring off-target kinase liabilities.

Despite this progress, rational optimization of ATP-competitive ligands remains fundamentally challenging. − ATP-binding pockets impose stringent geometric and electronic constraints on small-molecule scaffolds, requiring preservation of key anchoring interactions while permitting only limited chemical modification elsewhere on the molecule. − These constraints are particularly acute in pseudokinase systems, where ligand binding must stabilize a vestigial pocket without catalytic turnover. − Prior computational investigations of ROR1-binding small molecules have primarily relied on conventional docking and molecular dynamics analyses, − without explicitly incorporating quantum-level reactivity descriptors to guide site-specific chemical modification or to rationalize electronically labile regions within the ligand scaffold. As a result, a critical gap remains in connecting electronic structure–driven design principles with protein-scale dynamics and binding energetics for rational pseudokinase inhibitor optimization.

To address this gap, we developed a multiscale computational framework integrating density functional theory-based reactivity analysis with structure-based modeling and long-time scale molecular dynamics simulations to guide the rational redesign of Ponatinib-derived inhibitors targeting the ROR1 pseudokinase domain. Microsecond-scale simulations established a stable Ponatinib binding mode and revealed ligand-induced stabilization of the ATP-binding pocket relative to the apo state. Density functional theory-based charge and Fukui function analyses were then used to identify electronically feasible (probably reactive and metabolically labile regions) within the scaffold, enabling the rational design of eight analogs with preserved anchoring pharmacophore and overall binding geometry. The resulting compounds were evaluated through an integrated workflow combining molecular docking, extended molecular dynamics simulations, relative binding free-energy estimation, and in silico ADME profiling. Collectively, this quantum-informed and dynamics-driven strategy provides a mechanistically grounded and transferable framework for reactivity-guided ligand optimization in pseudokinase systems and next-generation ROR1 inhibitor discovery. Figure summarizes the overall strategy, designed analogs, and key findings.

1.

Summary of the complete computational workflow designed analogs and results.

2. Materials and Methods

The human ROR1 sequence was obtained from UniProt (Q01973). Experimentally resolved ROR1 structures were retrieved from the Protein Data Bank, and all computational analyses were based exclusively on the ponatinib-bound pseudokinase domain structure (PDB ID: 6TU9). , Missing loops and side chains were reconstructed using SWISS-MODEL. ,

Quantum chemical calculations on ponatinib were performed using ORCA 6.1.0, , with geometry optimization at the B3LYP-D3BJ/def2-SVP level and single-point refinement at the B3LYP-D3BJ/def2-TZVP level under SMD implicit solvation. Fukui reactivity indices were computed using a finite-difference approach based on Mulliken charges to identify electronically labile positions for analog design. , Hirshfeld-derived Fukui indices were additionally calculated as an independent, less basis-set-dependent reference for comparison across the different charge-partitioning schemes.

Ponatinib analogs were designed using quantum-informed reactivity descriptors and prepared using RDKit (https://github.com/rdkit/rdkit) and Open Babel v3.1.1 (https://github.com/openbabel) with the MMFF94 force field. Molecular docking was performed using Vina-GPU v2.0, with binding site definition and protocol validation based on redocking of the cocrystallized ligand. −

All-atom molecular dynamics simulations were prepared using CHARMM-GUI , and conducted using GROMACS 2022.4 , with the CHARMM36m force field. Ligand topologies and force-field parameters were generated using the CHARMM-GUI Ligand Reader and Modeler and were used without additional modifications. Apo and ponatinib-bound ROR1 systems were simulated for 1 μs each, while designed molecules (DM) analog complexes were simulated for 250 ns. Protein–ligand interaction fingerprints were computed using ProLIF, and relative binding free energies were estimated using the MM/PBSA method. , Physicochemical and ADMET properties were evaluated using SwissADME. Full methodological details are provided in the Supporting Information (SI).

3. Results and Discussion

3.1. Quantum Mechanical Characterization of Ponatinib

Ponatinib (AP24534) is a third-generation, orally bioavailable multikinase inhibitor originally developed to overcome resistance mutations in BCR-ABL, most notably the T315I gatekeeper mutation. − Structurally, Ponatinib is built around an imidazo­[1,2-b]­pyridazine core that forms key hydrogen-bonding interactions with the kinase hinge region within the ATP-binding site, thereby mimicking essential ATP contacts while maintaining a chemically distinct scaffold. − In contrast to ATPwhich is small, highly polar, and conformationally flexiblePonatinib adopts a rigid, extended architecture that enables simultaneous occupation of the canonical ATP pocket and an adjacent hydrophobic subpocket characteristic of the inactive DFG-out kinase conformation, consistent with its classification as a type II inhibitor. − A rigid ethynyl linker positions a m-trifluoromethylphenyl group for deep hydrophobic engagement, contributing to potency and metabolic stability, while a solvent-exposed methylpiperazine moiety connected through a flexible tether enhances solubility, basicity, and overall pharmacokinetic properties. − Collectively, this modular architecture underlies Ponatinib’s broad kinase inhibitory profile and makes it a chemically informative reference scaffold for structure-guided optimization against atypical kinase targets such as the ROR1 pseudokinase domain.

Single-point electronic structure calculations were performed at the B3LYP-D3BJ/def2-TZVP//B3LYP-D3BJ/def2-SVP level on the optimized geometry, with implicit solvation described by the SMD model (water). The resulting solvated electronic energy was −1824.38237 Eh, including D3BJ dispersion and solvent reaction-field polarization, and is lower than the corresponding SVP-level value, consistent with improved energetic convergence upon basis-set expansion (Figure B). The molecular dipole moment increased from 6.24 to 6.85 D, reflecting enhanced polarity in aqueous solution.

2.

(A) Optimized geometry of Ponatinib. (B) Key quantum chemical parameters at different theoretical levels calculation and comparison. (C–E) Heatmap representation of atomic charge distributions in Ponatinib computed at the B3LYP-D3BJ/def2-SVP level with SMD solvation (water). Mulliken, Löwdin, and differential charges (Δq = Mulliken – Löwdin) are shown for all 66 atoms. Panels correspond to (C) heteroatoms (O, N, F), (D) carbon atoms, and (E) hydrogen atoms. The polar (O28, N1) and electron-deficient (C27, C33) centers are highlighted in the heatmap to show the method-dependent variations in atomic charge assignment.

Harmonic vibrational frequency analysis at 310 K confirmed a true minimum and yielded a total enthalpy of – 1823.82799 Eh and an entropic contribution of – 0.09737 Eh (−61.10 kcal mol^–1), resulting in a final Gibbs free energy of – 1823.92535 Eh (Figure B). The overall thermal correction relative to the electronic energy amounts to G-E _el= +0.45702 Eh (286.78 kcal mol^–1). Table S1 summarizes the key electronic, solvation, and thermochemical parameters obtained from a hierarchical DFT workflow.

Quantum mechanical calculations reveal a pronounced and spatially organized polarization across the Ponatinib scaffold (Figure C–E). Both Mulliken and Löwdin population analyses identify a hydrogen-bond-competent anchoring region centered on the amide linkage, with strong electron density localized on the carbonyl oxygen O28 and heteroaromatic nitrogen N1, consistent with their roles in hinge recognition. In contrast, the amide carbon C27 and the CF₃-substituted aryl carbon C33 emerge as electron-deficient centers, defining regions susceptible to electrophilic modulation and polarization-driven interactions (Figure C–E and Table S2).

Despite numerical differences between Mulliken and Löwdin schemesincluding sign inversions at selected sites such as O28, N1, and the terminal fluorine atoms (F37–F39)both methods reproduce the same spatial polarity pattern. Specifically, the conjugated aryl framework remains π-electron rich, while the amide linkage maintains a pronounced internal dipole across the −C­(=O)–NH– motif (C27–O28–N29), supporting complementary hydrogen-bond donor/acceptor capacity. At the molecular periphery, the CF₃ group displays strong polarization, with a consistently negative carbon center (C35) and method-dependent fluorine charges, indicative of a highly polarizable terminus (Table S2).

Together, these features define two complementary electrostatic faces in Ponatinib: a hydrogen-bond-active amide/heteroaromatic domain and a π-rich aromatic surface capped by a polarizable CF₃ group. This anisotropic electrostatic architecture provides a mechanistic basis for stable accommodation within kinase ATP-binding pockets and informed the selection of modification sites for QM-guided analog design. Complete atom-resolved Mulliken and Löwdin, charge values, along with the atom-numbering scheme, are also provided in the SI (Table S2).

3.2. Orbital Population and Bond-Order Characteristics of Ponatinib

Reduced-orbital population and Mayer bond-order analyses further confirm the chemically saturated yet electronically anisotropic nature of the Ponatinib scaffold. The carbonyl oxygen O28 and heteroaromatic nitrogens (N1, N19) exhibit dominant p-character, consistent with localized lone-pair density and strong hydrogen-bond acceptor capacity, while the amide linkage (C27–O28–N29) maintains a pronounced internal dipole. Several conjugated carbons, notably C27 and the CF₃-substituted aryl carbon C33, display enhanced d/f orbital contributions and reduced p-character, indicating increased π-polarizability within the conjugated framework. Mayer bond-order analysis identifies a rigid alkyne linkage (C3–C24) with strong triple-bond character, a highly polarized carbonyl bond (C27–O28), and delocalized aromatic bond orders consistent with π-conjugation and planarity. Despite carrying a positive atomic charge, C33 remains electronically stabilized through conjugation and exhibits near-zero free valence, underscoring a chemically stable yet polarizable terminus. Together, these results define a scaffold combining rigid structural elements with localized electronic flexibility, supporting stable accommodation within kinase ATP-binding pockets and providing a quantum-mechanical basis for reactivity-guided analog design. Detailed orbital population data and bond-order values are provided in the SI (Tables S3–S6).

3.3. Fukui Reactivity Analysis and Rational Identification of Modification Sites

To enable reactivity-guided optimization of the Ponatinib scaffold, condensed Fukui functions were computed in water at the B3LYP-D3BJ/def2-TZVP level using both Mulliken and Löwdin population schemes. To verify that the neutral reference charge distribution used for Fukui analysis was stable and reproducible, Mulliken/Löwdin charges were extracted from three independently generated single-point data sets (frequency at 310 K, self-consistent polarization, and Fukui-neutral calculations). All three charge sets were numerically indistinguishable (Mulliken mean absolute deviations (MAD) < 7 × 10^–7 e; RMSD < 2 × 10^–6 e; Löwdin identical within printing precision), confirming that the Fukui-neutral charges accurately represent the optimized neutral electron density and can be used consistently as the N-state reference.

Condensed Fukui indices were calculated as f ⁺ = q(N + 1) – q(N) (electrophilic centers; susceptibility to nucleophilic attack), f ^– = q(N) – q(N – 1) (nucleophilic centers; susceptibility to electrophilic attack), and f ⁰ = (f ⁺ + f ^–)/2 (radical susceptibility). Mulliken and Löwdin derived Fukui profiles showed strong agreement in the location of reactive centers (Pearson r = 0.957 for f ⁺; r = 0.734 for f ^–; r = 0.830 for f ⁰; Spearman ρ = 0.803 for f ⁺ and ρ = 0.499 for f ^–). Overall, these results indicate that the principal reactivity topology was broadly preserved across the two population schemes, although nucleophilic-site ranking showed greater scheme dependence (Tables S7–S9).

The Fukui map delineated a small number of electronically responsive positions suitable for scaffold modification. In the Mulliken model, the largest f ⁺ values were observed at C11 (0.0086) and C9 (0.0081), followed by C26 (0.0062) and C31 (0.0038), indicating electrophilic character concentrated along the conjugated linker/aryl region. In the Löwdin model, electrophilic susceptibility shifted in magnitude toward the CF₃-bearing aryl terminus, with C33 (0.0143) and C35 (0.0062) as dominant f ⁺ centers, alongside C12 (0.0047) and N29 (0.0040). Complementarily, nucleophilic susceptibility (high f ^–) was concentrated on heteroaromatic nitrogens and select conjugated carbons: Mulliken identified C31 (0.0323), N20 (0.0207), N19 (0.0153), and N1 (0.0128) as principal nucleophilic sites, while Löwdin highlighted N19 (0.0155), C21 (0.0102), N1 (0.0065), and C31 (0.0035). Radical susceptibility remained uniformly low (largest f ⁰ at C31 in both schemes), supporting a design framework dominated by polar reactivity rather than radical pathways (Table S7).

These atom-resolved trends directly guided analog design by prioritizing modifications at positions that are (i) electronically responsive by Fukui criteria and (ii) peripheral enough to tolerate substitution without perturbing essential scaffold elements. Specifically, C11 (high Mulliken f ⁺), C33/C35 (high Löwdin f ⁺ at the CF₃-bearing aryl region), and secondary permissive sites C10 and C32 (modest positive f ⁺ with otherwise stable indices) were selected as chemically tractable modification sites to tune polarity and electronic distribution. In contrast, electronically stable structural motifsincluding the CC linkage (C3/C24) and the amide carbonyl (C27/O28)exhibited negligible Fukui response (|f⁺| and |f^–| < 0.01), supporting their preservation to maintain scaffold rigidity and hydrogen-bonding capacity. Collectively, this Fukui-guided map defines a chemically constrained but actionable optimization space on Ponatinib and provides the electronic rationale for atom substitution pattern focused on C10/C11/C32/C33/C35 while retaining the core anchoring framework (Tables S7 and S8).

A methodological limitation of the present Fukui analysis has been also acknowledged here. Mulliken-derived condensed Fukui functions are known to be sensitive to basis-set choice and may yield negative values at individual atomic sites, which do not directly reflect physically meaningful reactivity. , Consistent with this limitation, the Spearman rank correlation between Mulliken- and Löwdin-derived f ^– values was moderate (ρ = 0.499), indicating only partial agreement in nucleophilic-site ranking. To address this concern and assess the robustness of the resulting site assignments, Hirshfeld-derived Fukui indices were additionally calculated as an independent reference. , This analysis provided additional support for the overall f ⁺-guided design rationale, although the degree of agreement varied across sites and charge-partitioning schemes. Collectively, the combined Fukui analysis converged on C10, C11, C32, and C35 as candidate modification sites, but not with uniform support at every position. C10 was supported mainly by Löwdin and Hirshfeld, as Mulliken f ⁺ remained negative. C11 was supported mainly by Mulliken and Hirshfeld, whereas C32 was supported mainly by Löwdin and Hirshfeld. In contrast, C35 showed the strongest overall support, with positive f ⁺ values in all three schemes, most prominently in the Hirshfeld analysis. Taken together, these results indicate that the key design-relevant conclusions were not based on Mulliken descriptors alone, but on the combined cross-scheme reactivity pattern, with the exact ranking of individual sites remaining charge-partition dependent (Table S10).

3.4. Design of Ponatinib Analogs Guided by Fukui Reactivity Mapping

Reactivity-guided design of Ponatinib analogs was performed by integrating condensed Fukui-function analysis with local electronic and structural context across the scaffold. Based on combined Mulliken and Löwdin Fukui indices, four positionsC11, C35, C10, and C32were identified as chemically tractable modification sites. These positions either exhibit intrinsic electrophilic character or lie within electronically permissive regions capable of accommodating substitution without compromising scaffold stability or key anchoring interactions in the ROR1 pseudokinase ATP-binding pocket.

At these sites, a chemically diverse set of substituentsincluding CF3 derivatives, sulfonyl-containing groups, conjugated enamides, fluorinated alkyl fragments, tertiary alcohols, and small cycloalkyl moietieswas introduced to systematically probe the effects of electronic polarity, hydrogen-bonding capacity, steric bulk, and hydrophobicity (Figure ). In total, 8 Fukui-guided analogs were generated. These eight analogs (DM1–DM8) were selected for detailed analysis based on electronic diversity, structural moderation, and subsequent molecular dynamics performance.

3.

Designed molecules (DMs) DM1 – DM8 based on ponatinib Fukui Reactivity Mapping and charge analysis.

3.4.1. C11: Replacement of the CF₃ Substituent with Polar and Conjugated Groups

Among all scaffold positions, C11 exhibits the highest electrophilic Fukui index in the Mulliken scheme (f ⁺ = 0.0086), identifying it as the most favorable site for nucleophilic substitution. This pronounced electrophilic character supports replacement of the CF₃ group at C11 with more polar or hydrogen-bond-capable substituents without destabilizing the electronic framework. Accordingly, polar and π-conjugated groups were introduced, including m-1,1-difluoro-1-(trifluoromethoxy)­methylphenyl (DM1), m-pentafluoroethyl phenyl (DM2), conjugated enamide side chains such as prop-2-enamide/acrylamide (DM3), methyl sulfonyl ethynyl (DM4), and sulfonamide (DM5). These substituents span σ-electron-withdrawing (sulfonyl) and π-conjugating (enamide) functionalities, enabling fine control of electronic distribution, solubility, and hydrogen-bonding capacity. The high f ⁺ value at C11 thus provides a direct electronic rationale for polar tuning at this position.

3.4.2. C35: CH₃ → CF₃ Substitution for Dipole and Hydrophobic Optimization

The C35 position, adjacent to the piperazine-linked region, displays electrophilic character in the Löwdin analysis (f ⁺= 0.0062) and a weaker but consistent response in Mulliken (f ⁺≈ 0.0010). Both schemes indicate that C35 carries a net negative charge (Mulliken −0.32 e; Löwdin −0.33 e), while fluorine atoms in a CF₃ group are positive in the Löwdin scheme (+0.18 to +0.19 e), generating a strong localized dipole. Introduction of a CF3 group at this site (DM6) therefore enhances dipolar character and hydrophobic surface area while remaining electronically stable, improving pocket complementarity without chemical overactivation.

3.4.3. C10: Polar Functionalization near the Nucleophilic C31 Region

Although C10 itself exhibits low Fukui reactivity in both charge schemes, it is electronically coupled to C31, the most nucleophilic center in the molecule (Mulliken f ^– = 0.0323; Löwdin f ^– = 0.0035). Together, this C10–C31 segment forms a reactivity corridor capable of transmitting electronic and interaction effects. Functionalization at C10 with an isopropanol (DM7) introduces a proximal hydrogen-bond donor/acceptor near the nucleophilic C31 region, increasing opportunities for hydrogen bonding with surrounding residues and improving solvent stabilization while preserving overall electronic neutrality.

3.4.4. C32: Steric and Hydrophobic Optimization at an Electronically Inert Site

The C32 position displays negligible Fukui activity in both schemes (f ⁺ ≈ 0–0.0017; f ^– ≈ −0.009 to −0.001), identifying it as electronically inert and unsuitable for polar modification. This makes C32 ideal for steric and hydrophobic tuning. Introduction of a cyclopropyl substituent (DM8) modestly increases hydrophobic surface area and adjusts three-dimensional shape without perturbing electronic distribution, improving pocket fitting and shape complementarity.

Overall, Fukui-function analysis delineates distinct electronic domains across the Ponatinib scaffold. Electrophilic hot spots localize at C11, C33, and C35, while nucleophilic character concentrates at C31, N19, N1, and C21. The strong consistency between Mulliken and Löwdin schemes, supported by high intermethod correlations, validates the robustness of these assignments. Together, these data establish a coherent, QM-driven design framework in which C11 and C35 enable electronic and polar tuning, whereas C10 and C32 serve as electronically stable sites for steric and hydrophobic modulation, underpinning the rational design of the DM1– DM8 analog series.

3.5. Crystal Structure Interactions, Docking Validation, and Docking Analysis of the Designed Analog Series

3.5.1. Crystallographic Interaction Network of Ponatinib in the ROR1 Pseudokinase Domain

The cocrystal structure of Ponatinib bound to the ROR1 pseudokinase domain (PDB ID: 6TU9) reveals a highly coordinated binding mode stabilized by a synergistic network of polar, aromatic, hydrophobic, and directional interactions. Ponatinib adopts a deeply buried conformation within the ATP-binding cleft, anchored by a compact and ordered interaction framework.

Three short, highly directional hydrogen bonds define the primary anchoring triad: the ligand amide nitrogen donates to the side-chain carbonyl oxygen of Glu71 (H–A = 2.36 Å), the backbone NH of Ile103 donates to the ligand heteroaromatic nitrogen (H–A = 1.68 Å), and the backbone NH of Asp181 donates to the ligand carbonyl oxygen (H–A = 1.79 Å). These near-ideal geometries lock the heteroaromatic core across the hinge-proximal and lower cleft regions (H–A denotes hydrogen–acceptor distance).

This polar triad is surrounded by a dense hydrophobic enclosure formed by Leu27, Ala52, Phe100, Tyr102, and Leu182, with contact distances of ∼2.4–4.0 Å; the closest contact occurs with Leu182 (2.39 Å), indicating a sterically optimized fit. Aromatic complementarity is reinforced by a parallel-displaced π–π interaction between the ligand phenyl ring and Phe100 (centroid distance 3.70 Å), together with a π–cation interaction between the ligand tertiary amine and His161 (5.91 Å). Directionality is further enhanced by a halogen bond between a ligand fluorine and the backbone carbonyl oxygen of Ile179 (3.29 Å; angle 165°). Collectively, hydrogen bonding, hydrophobic enclosure, aromatic stacking, cation−π recognition, and halogen bonding cooperate to stabilize the inhibitor in a tightly integrated binding architecture.

3.5.2. Docking Validation against the Crystallographic Reference

Redocking of Ponatinib into the 6TU9 binding site reproduced the crystallographic pose with high fidelity, yielding a docking score of −8.1 kcal·mol^–1 (Table S11) and a docked-crystal RMSD of 0.899 Å (Figure ). The canonical hydrogen-bond triad was preserved: amide N-Glu71 (2.16 Å), Ile103 NH-heteroaromatic N (2.01 Å), and Asp181 NH-carbonyl O (2.10 Å). Although slightly elongated relative to the crystal structure, these contacts remain within optimal ranges.

4.

(A) Two-dimensional protein–ligand interaction diagram derived from the ROR1–Ponatinib cocrystal structure (PDB ID: 6TU9), highlighting key hydrogen bonds, hydrophobic contacts, aromatic stacking, and electrostatic interactions within the ATP-binding pocket. (B) Two-dimensional interaction diagram of the top-ranked docked Ponatinib pose in the same binding pocket, illustrating conservation of the primary interaction network observed in the crystal structure. (C) Docking protocol validation shown by superposition of the docked Ponatinib pose and the crystallographic ligand conformation, demonstrating close structural agreement within the binding site (RMSD 0.899 Å). Interaction color codes corresponding to different noncovalent interaction types are shown in the legend box.

The hydrophobic envelope was largely conserved, with tight contacts involving Leu182 (2.47 Å), Val84, Phe100, Leu74, and Lys54 (∼2.5–3.9 Å). The π–π stacking interaction with Phe100 persisted (3.50 Å). Notably, the docked pose introduced a salt bridge between the ligand tertiary amine and Asp181, absent in the crystal structure, while the halogen bond to Ile179 and the π–cation interaction with His161 were lost, reflecting minor peripheral reorientation. Overall, docking accurately reproduced the core interaction network with realistic local rearrangements.

3.5.3. Docking Analysis of Fukui-Guided Designed Analogs

Following validation, the same docking protocol was applied to all Fukui-guided analogs (Table S11). Up to ten poses were generated per ligand, and pose selection followed a hydrogen-bond–enrichment criterion: Pose 1 was retained if it contained the maximum number of hydrogen bonds; otherwise, a lower-ranked, hydrogen-bond-enriched pose was selected.

Among the analogs, DM1 exhibited the most favorable docking score (−8.6 kcal·mol^–1) with three hydrogen bonds, outperforming Ponatinib. DM3 showed a low-scoring Pose 1 (−3.6 kcal·mol^–1) with one hydrogen bond; however, pose 6 formed three hydrogen bonds despite a weaker score (−1.6 kcal·mol^–1) and was therefore selected. DM6 and DM8 each formed three hydrogen bonds (−7.2 and – 7.0 kcal·mol^–1, respectively), while DM7 formed four hydrogen bonds in Pose 1 (−6.1 kcal·mol^–1). Hydrogen-bond-enriched lower-ranked poses were selected for DM2 (Pose 2, −7.7 kcal·mol^–1), DM4 (Pose 3, −5.4 kcal·mol^–1), and DM5 (Pose 2, −7.1 kcal·mol^–1). These curated poses were used for subsequent analyses. The lack of strict correlation between docking score and hydrogen-bond count highlights the complementary value of interaction-based pose selection.

3.5.4. Conserved Interaction Framework across the Analog Series

Across all complexes, a conserved interaction framework involving Leu27, Ile35, Ala52, Lys54, Glu67, Glu71, Leu74, Val84, Phe100, Tyr102, Ile103, His161, Ile179, Asp181, and Leu182 was observed (Figure ). All analogs preserved the canonical hinge-region hydrogen-bond triad with Glu71, Ile103, and Asp181 (H–A ≈ 1.7–2.5 Å) and maintained the aromatic clamp formed by Phe100 and Tyr102 (centroid distances ≈ 3.5–3.9 Å).

5.

Obtained docking pose for designed analogs DM1– DM8 inside the ROR1 pseudokinase pocket.

Secondary polar interactions varied with substituent chemistry: Glu67 contributed additional hydrogen bonding in DM3, Lys54 in DM5, and His161 participated in π–cation or halogen-bond interactions in DM2, DM6, and Ponatinib. Ile179 formed auxiliary halogen bonds in Ponatinib and DM1. Asp181 acted as a multifunctional nodehydrogen-bond partner, salt-bridge acceptor, or hydrophobic contactin all analogs except OLI, DM5, and DM7. DM7 formed an extended four-hydrogen-bond network spanning Glu71, Ile103, and dual contacts with Asp181, while DM5 exhibited the highest overall hydrogen-bond enrichment (six hydrogen bonds) distributed across Lys54, Glu71, Ile103, and Asp181.

Hydrophobic mapping revealed a conserved nonpolar enclosure defined by Leu182, Leu74, Phe100, Tyr102, and Val84, with additional contributions from Ala52, Asp181, and Glu71. Peripheral residues (Leu27, Ile35, Lys54) showed greater analog-specific variability. Collectively, these results demonstrate that all Fukui-guided analogs preserve the essential Ponatinib pharmacophorehinge anchoring, aromatic stacking, and hydrophobic enclosurewhile selectively modulating peripheral polar and halogen-bond interactions without disrupting core ATP-site recognition.

3.6. Molecular Dynamics Analysis of the Ponatinib–ROR1 Pseudokinase Complex and Designed Analogs

Following the quantum- and Fukui-guided design of the analog series, one-microsecond all-atom MD simulations were performed for both the apo and Ponatinib-bound ROR1 pseudokinase domain to establish a dynamic reference framework under physiological conditions. These long-time scale trajectories captured pocket flexibility, interaction persistence, and ligand-induced stabilization beyond static structural models.

Throughout the simulation, Ponatinib retained a stable binding pose within the ATP-binding cleft, defining a dynamic pharmacophore composed of the hinge-binding carbonyl (O28), heteroaromatic and amide nitrogen donors (N20 and N29), the rigid alkyne linker (C3–C24) that preserves distal aryl orientation, and the m-trifluoromethyl-bearing aryl terminus (C33–F37–F39) anchoring the ligand within the hydrophobic subpocket. These elements remained consistently engaged despite thermal fluctuations, underscoring their central role in binding stability.

System-level stability metricsincluding backbone RMSD, per-residue RMSF, radius of gyration (R _g), and solvent-accessible surface area (SASA)confirmed that the Ponatinib–ROR1 complex remained conformationally stable over the full 1 μs time scale. Protein–ligand interaction fingerprints extracted using ProLIF, together with MM/PBSA-based relative binding free-energy estimates, provided quantitative benchmarks for interaction occupancy and energetic robustness. Collectively, this reference ensemble defines the mechanistic and energetic criteria against which the behavior of the Fukui-guided analogs is directly evaluated in subsequent simulations.

3.6.1. Backbone Stability and Global Dynamics of the ROR1 Pseudokinase Domain

To quantify how Ponatinib modulates the conformational dynamics of the ROR1 pseudokinase domain, we analyzed backbone RMSD, radius of gyration (R _g), and solvent-accessible surface area (SASA) over a 1 μs molecular dynamics trajectory for the ligand-free (Apo) and Ponatinib-bound states. Collectively, these metrics reveal a pronounced ligand-induced stabilization that suppresses both long-time scale drift and short-time scale backbone fluctuations.

The Apo ROR1 pseudokinase domain exhibited substantial intrinsic flexibility, characterized by a broad RMSD distribution and sustained structural drift (Figure A,B). The mean backbone RMSD was 0.384 ± 0.058 nm (Table ), with excursions exceeding 0.53 nm, indicating large-amplitude conformational displacements from the starting structure (Figure ). The right-shifted and broadened RMSD probability distribution reflects heterogeneous sampling of multiple displaced substates rather than a single dominant basin (Figure C). Time-resolved analyses further showed repeated increases in rolling RMSD standard deviation, consistent with transient breathing motions and loop rearrangements that persist throughout much of the trajectory (Figure D,E).

6.

Backbone RMSD-based insights from the MD simulations of the ROR1 pseudo kinase domain in the absence (Apo) and presence of Ponatinib. (A) Time-resolved backbone RMSD profiles reveal that the Ponatinib-bound complex equilibrates more rapidly and maintains consistently lower fluctuations compared to the Apo state. (B) The RMSD difference plot (Apo-Ponatinib) highlights transient peaks reaching up to ∼0.174 nm at specific intervals (e.g., 161–775 ns), indicating larger conformational excursions in the Apo ensemble. (C) RMSD probability density distributions show a narrower and left-shifted profile for the Ponatinib-bound system, reflecting greater structural rigidity relative to the broader Apo distribution. (D) Running mean RMSD (10 ns window) shows that the Ponatinib-bound ensemble reaches 95% of its final mean RMSD by ∼256 ns, while the Apo system requires ∼774 ns to achieve similar convergence, confirming faster stabilization upon ligand binding. (E) Rolling RMSD standard deviation analysis (10 ns window) captures short-time scale backbone fluctuations, revealing higher and more variable local flexibility in the Apo system, consistent with transient structural and partial breathing, whereas the Ponatinib-bound state exhibits dampened local dynamics.

1. Summary of Structural Metrics for DM1–DM8 (Ponatinib Analogs) Bound to ROR1 Systems .

ref/analog	RMSD (nm) (mean ± SD)	R g (nm)(mean ± SD)	SASA (nm 2 )(mean ± SD)
Apo	0.384 ± 0.058	2.021 ± 0.018	159.83 ± 3.52
Ponatinib	0.345 ± 0.059	2.052 ± 0.021	162.20 ± 3.44
DM1	0.326 ± 0.044	2.074 ± 0.025	165.64 ± 3.19
DM2	0.317 ± 0.040	2.053 ± 0.013	164.10 ± 2.90
DM3	0.331 ± 0.045	2.077 ± 0.023	165.29 ± 2.97
DM4	0.320 ± 0.042	2.061 ± 0.022	163.91 ± 3.55
DM5	0.319 ± 0.046	2.082 ± 0.050	164.07 ± 3.95
DM6	0.313 ± 0.041	2.043 ± 0.049	162.73 ± 2.90
DM7	0.341 ± 0.048	2.069 ± 0.049	165.18 ± 4.00
DM8	0.322 ± 0.043	2.052 ± 0.046	161.41 ± 2.80

^aMean ± standard deviation values of backbone RMSD, radius of gyration, and solvent-accessible surface area calculated from MD trajectories of Apo ROR1, Ponatinib-bound, and designed Ponatinib analog complexes. Apo and Ponatinib systems were simulated for 1000 ns, while DM1–DM8 complexes were simulated for 250 ns.

In contrast, Ponatinib binding significantly dampened global backbone mobility (Figure B). The Ponatinib-bound complex maintained a lower mean RMSD (0.345 ± 0.059 nm, Table and Figure A) with fewer high-amplitude excursions and a markedly narrower distribution, indicative of a more restricted conformational ensemble (Figure A). The ΔRMSD­(t) profile (ApoPonatinib) remained positive for nearly the entire simulation, confirming that the ligand-free protein consistently deviates further from its reference structure (Figure B). Importantly, although both systems reached ∼95% of their final mean RMSD within the first ∼250 ns, their long-time scale behavior diverged sharply: the Ponatinib-bound system rapidly settled into a stable, drift-free plateau, whereas the Apo trajectory continued to undergo slow structural relaxation for several hundred nanoseconds before achieving comparable convergence (Figure ).

This stabilization is further supported by local fluctuation analyses. Rolling RMSD standard deviations were consistently lower in the Ponatinib-bound ensemble, demonstrating that ligand binding suppresses not only global drift but also short-time scale backbone motions (Figure D,E). Taken together, these results establish Ponatinib as an effective conformational stabilizer of the ROR1 pseudokinase domain, reshaping its dynamic landscape toward a more compact and homogeneous ensemble.

3.6.2. RMSD Signatures of Designed Ponatinib Analogs (DM1–DM8)

To assess whether the designed analogs preserve or enhance this stabilizing effect, we performed 250 ns MD simulations for eight ligand–ROR1 complexes (DM1–DM8, Table and Figure ). All analog-bound systems reached well-defined RMSD plateaus within 0.25–0.40 nm, indicating overall structural integrity and absence of unfolding or progressive drift.

7.

RMSD-based stability and ensemble organization of ROR1 bound to Ponatinib analogs. (A) RMSD time series. Backbone RMSD traces for the eight analogs (DM1–DM8) over 250 ns reveal four stability classes. DM5 shows the lowest fluctuations and most Ponatinib-like stabilization, DM7 displays Apo-like flexibility, DM2 and DM8 occupy compact low-RMSD regimes, and the remaining analogs fall within an intermediate band. (B) 10 ns running mean RMSD. Sliding-window averaging highlights differences in equilibration behavior. DM5 stabilizes earliest, DM7 converges slowly with higher drift, and DM2 and DM8 rapidly settle into compact basins. Other analogs display moderate convergence patterns. (C) RMSD distributions. Probability density distributions confirm these trends: DM5 shows a narrow, Ponatinib-like peak; DM7 exhibits a broad Apo-like profile; DM2 and DM8 shift toward lower RMSD values, indicating distinct compact substates; and DM1, DM3, DM4, DM6 occupy intermediate regions. (D) Rolling RMSD standard deviation. Local 10 ns fluctuation analysis shows that DM5 maintains uniformly low short-time scale variability, DM7 exhibits elevated and irregular fluctuations, and DM2/DM8 show reduced but structurally distinct fluctuation patterns. Intermediate analogs fall between these extremes. (E) Stability radar plot. A composite metric (mean RMSD, KS divergence vs Apo and Ponatinib, rolling SD, equilibration time) summarizes global behavior. DM5 ranks as the most Ponatinib-like stabilizer, DM7 aligns with the Apo-like flexibility class, DM2 and DM8 form alternative compact regimes, and remaining analogs show moderate stabilization. (F) PCA of RMSD distributions. PCA of ensemble RMSD histograms separates the systems into distinct regions of conformational space. Apo and Ponatinib form well-defined reference clusters. DM7 clusters closest to Apo, DM5 lies nearest to Ponatinib, DM2 and DM8 appear as outliers representing unique compact substates, and DM1/DM3/DM4/DM6 populate intermediate positions between the two reference states.

Despite this shared global stability, clear quantitative distinctions emerged among the analogs. DM2 and DM8 displayed the lowest mean RMSD values (0.2578 and 0.2992 nm) and the narrowest RMSD distributions, reflecting highly compact ensembles with strongly suppressed backbone mobility. DM1, DM3, and DM4 exhibited intermediate behavior, maintaining stability while allowing modest fluctuations. In contrast, DM5, DM6, and particularly DM7 showed elevated RMSD magnitudes and broader distributions, indicating increased conformational heterogeneity relative to Ponatinib.

Equilibration kinetics further differentiated these ligands. DM2, DM3, DM4, DM8, and DM1 equilibrated rapidlywithin a few nanosecondsclosely matching the fast-early relaxation observed for Ponatinib. Conversely, DM5, DM6, and DM7 required substantially longer times to converge, approaching the slower equilibration behavior of the Apo system (Table S12). These slower-equilibrating analogs also exhibited larger and more irregular rolling RMSD standard deviations, consistent with enhanced local flexibility.

3.6.3. Statistical and Multivariate Classification of Conformational Ensembles

Two-sample Kolmogorov–Smirnov (KS) tests quantitatively compared RMSD distributions of each analog against the Apo and Ponatinib references (Table S12 and Figure E). All analogs differed significantly from the Apo ensemble, confirming that ligand binding universally suppresses Apo-like flexibility. DM2 and DM8 showed the largest KS distances from Apo, consistent with their strong stabilization. Comparisons against the Ponatinib ensemble revealed a broader spectrum: DM5 emerged as the closest dynamical mimic of Ponatinib, whereas DM2 and DM8 showed the largest deviations. Notably, these deviations do not indicate destabilization; instead, they reflect stabilization of compact conformational substates distinct from those preferentially sampled by Ponatinib.

Principal component analysis (PCA) of RMSD probability distributions provided a global, integrative view of these relationships (Figure F). The Apo and Ponatinib ensembles occupied well-separated regions of PCA space, underscoring their fundamentally different dynamic signatures. The designed analogs populated this space in a structured manner: DM7 clustered closest to the Apo ensemble, consistent with its higher residual flexibility; DM5 localized near the Ponatinib cluster, corroborating its KS-based similarity; and DM2 and DM8 occupied distant regions, reflecting their ability to stabilize unique, compact conformational basins rather than reproducing the Ponatinib ensemble.

3.6.4. Integrated Stability Hierarchy and Implications for Ligand Design

Integrating mean RMSD values, equilibration kinetics, KS statistics, and PCA clustering yields a clear hierarchy of ligand-induced stabilization (Figure E). The Apo system represents the most flexible reference state. Among the analogs, DM7 retains the greatest degree of Apo-like mobility, followed by DM6 and DM5 with intermediate flexibility. DM5 stands out as the closest dynamical surrogate of Ponatinib, whereas DM1, DM3, and DM4 form a midstability class. At the top of the hierarchy, DM2 and DM8 emerge as the strongest stabilizers, inducing compact and dynamically distinct conformational substates with minimal backbone fluctuations.

Collectively, these results demonstrate that quantum-guided scaffold modifications can tune not only binding stability but also the nature of the stabilized conformational ensemble. Rather than uniformly reproducing the Ponatinib-bound state, certain analogs reshape the dynamic landscape of the ROR1 pseudokinase domain, highlighting conformational stabilization as an orthogonal and chemically addressable design dimension for ATP-pocket-targeted inhibitors.

3.6.5. Global Compactness and Shape Regulation Revealed by Radius of Gyration

To complement backbone RMSD analyses and directly assess ligand-dependent changes in global architecture, we examined the radius of gyration (R _g) of the ROR1 pseudokinase domain in the Apo and Ponatinib-bound states over 1 μs of molecular dynamics simulation (Figure A,B and Table ). While RMSD primarily reports positional deviation from a reference structure, Rg provides an orthogonal measure of overall mass distribution and compactness.

8.

Radius of gyration (R _g) and solvent-accessible surface area (SASA) analyses of the ROR1 pseudokinase domain. (A) Time-series R _g profiles of the Apo and Ponatinib-bound ROR1 pseudokinase domain, illustrating raw fluctuations in global compactness over the 1 μs molecular dynamics simulation. (B) Rolling-mean R _g profiles calculated using a 10 ns window for the Apo and Ponatinib-bound states, highlighting long-time scale trends in structural compactness. (C) Time-series SASA profiles of the Apo and Ponatinib-bound ROR1 pseudokinase domain, showing raw solvent exposure fluctuations throughout the simulation. (D) Rolling-mean SASA profiles with corresponding standard deviations (10 ns window), illustrating differences in average solvent accessibility and dynamic variability between the Apo and ligand-bound states.

The Apo ROR1 pseudokinase domain maintained a mean R _g of 2.021 ± 0.018 nm, consistent with a compact but dynamically permissive fold. Upon Ponatinib binding, the mean R _g increased modestly to 2.052 ± 0.021 nm (Table ). This slight expansion does not indicate destabilization; rather, it reflects a controlled conformational adjustment accommodating the bulky inhibitor within the ATP-binding cleft. Importantly, both systems reached 95% of their final mean R _g values within the first ∼10 ns, demonstrating that global shape equilibration occurs rapidly and precedes slower backbone relaxation observed in RMSD analyses.

Despite its marginally larger average R _g, the Ponatinib-bound ensemble exhibited narrower local R _g fluctuations relative to the Apo system. Rolling standard deviation analysis revealed that Ponatinib constrains transient shape deformations, suppressing excessive breathing motions while stabilizing a slightly expanded but well-ordered conformation (Figure B). This combination of mild expansion with reduced variability suggests ligand-induced rigidification rather than loss of structural integrity.

3.6.6. Radius of Gyration Signatures of the Designed Analog Series

To contextualize the designed analogs within the Apo-Ponatinib compactness landscape, R _g profiles were evaluated for all eight ligand-bound complexes over equilibrated 250 ns trajectories (Figure A and Table ). All analogs retained R _g values within a narrow range (2.03–2.08 nm), confirming preservation of the native pseudokinase fold without evidence of global unfolding or collapse.

9.

Radius of gyration (R _g) and solvent-accessible surface area (SASA) analyses of the ROR1 pseudokinase domain in the presence of DM1–DM8 analogs. (A) R _g profiles reflecting ligand-dependent modulation of global compactness. (B) SASA profiles illustrating changes in solvent exposure upon binding of designed analogs.

However, fluctuation amplitudes clearly differentiated the analogs. DM2 displayed the lowest R _g variability (SD = 0.013 nm), identifying it as the most rigid and globally stabilized complex. DM1, DM3, and DM4 formed an intermediate-stability group with modest R _g fluctuations, comparable to the Apo-Ponatinib baseline. In contrast, DM5, DM6, DM7, and DM8 exhibited broader R _g fluctuation ranges, indicating increased exploration of compactness states despite maintaining stable mean Rg values.

Among these, DM7 consistently sampled the upper end of the R _g distribution, approaching an Apo-like expanded ensemble, whereas DM5 retained relatively compact mean R _g values despite elevated fluctuationssuggesting a stable core accompanied by enhanced peripheral breathing. These trends reinforce the notion that analog-induced effects are not binary (stable vs unstable) but instead modulate the mode of compactness sampling.

3.6.7. Ligand-Dependent Modulation of Solvent Exposure Revealed by SASA

To further probe how ligand binding reshapes the global surface properties of ROR1, solvent-accessible surface area (SASA) was analyzed for the Apo and Ponatinib-bound systems over the full 1 μs trajectory. The Apo protein exhibited a mean SASA of 159.83 ± 3.52 nm², while the Ponatinib-bound complex showed a modestly higher mean value of 162.20 ± 3.44 nm² (Figure C,D and Table ). This increase is consistent with localized rearrangements around the ligand-binding pocket that expose peripheral residues without inducing large-scale expansion.

Both systems reached 95% of their final mean SASA values within ∼9 ns, indicating rapid equilibration of solvent exposure. Fluctuation analysis revealed that, although Ponatinib suppresses backbone mobility, it slightly enhances variability in solvent-exposed surface area. This apparent decouplingrigidified internal backbone dynamics accompanied by modestly increased surface breathingsuggests that ligand binding redistributes flexibility from the protein core to peripheral regions rather than eliminating it entirely.

3.6.8. SASA Profiles of the Designed Analog Series

SASA analysis of the analog-bound complexes revealed mean values tightly clustered between 161 and 166 nm², fully consistent with the R _g-derived compactness landscape (Figure B and Table ). DM8 and DM6 exhibited the lowest SASA values, indicating the most solvent-shielded conformations, whereas DM1, DM3, and DM7 showed slightly elevated surface exposure.

Fluctuation amplitudes further resolved stability differences. DM2, DM6, and DM8 displayed the lowest SASA variability, consistent with rigid and well-shielded ensembles. In contrast, DM7 and DM5 exhibited the largest SASA fluctuations, reflecting enhanced surface breathing motions and Apo-like flexibility. Notably, these SASA trends closely mirror RMSD- and Rg-based rankings, underscoring the internal consistency of the stability hierarchy.

3.7. Integrated Interpretation of RMSD, R _g, and SASA

Together, RMSD, R _g, and SASA analyses provide a coherent, multidimensional view of ligand-dependent conformational regulation in the ROR1 pseudokinase domain (Table ). Ponatinib binding reduces global backbone drift and local fluctuations while inducing a mild, well-controlled expansion that accommodates the inhibitor and redistributes flexibility toward solvent-exposed regions. This rigidified yet adaptable conformational ensemble likely underpins high-affinity and specific inhibitor engagement.

Across the designed analogs, DM2 consistently emerges as the most conformationally stable complex, characterized by minimal RMSD, the lowest Rg variability, and tightly regulated SASA, although ligand-level block analysis (50 ns) indicated residual positional drift. DM8 also showed a stable and compact ensemble, whereas DM7 displayed comparatively higher RMSD and greater residual drift at the protein level. DM6 remained compact and solvent-stable despite some protein-level RMSD drift. The remaining analogs occupy intermediate positions, highlighting a graded and tunable stability landscape. Importantly, these results demonstrate that rational scaffold modification can selectively modulate not only binding strength but also the global dynamic architecture of the target protein.

Additional five-block (50 ns) convergence analysis was also performed for both protein- and ligand-level trajectories and is provided in the Supporting Information (Tables S13 and S15). Overall, the simulations showed satisfactory stability within the 250 ns window. Although a subset of systems displayed residual RMSD drift, both protein and ligand R _g and SASA profiles remained uniformly stable, supporting preservation of global compactness and solvent exposure. These results indicate that the observed variability is more consistent with local conformational adjustment than with global instability.

3.8. Local Flexibility Modulation by Ponatinib and Analogs DM1–DM8 Binding

3.8.1. Ligand-Dependent Remodeling of Local Flexibility in the ROR1 Pseudokinase Domain

Backbone RMSF analysis revealed that ligand binding reshapes the spatial distribution of flexibility across the ROR1 pseudokinase domain in a residue-selective and chemically tunable manner (Figures and ). Rather than producing uniform rigidification, Ponatinib and the designed analogs redistribute mobility across a small set of mechanically dominant segmentsmost prominently residues 10–25, 28–38, 45–65, and 120–150which collectively define the principal flexibility hotspots of the Apo ensemble. These segments correspond to loop-rich and regulatory motifs (N-terminal/hinge-adjacent loops, β-hairpin elements, and activation-segment-proximal regions) that commonly gate conformational breathing and substrate-pocket accessibility in pseudokinase-like folds.

10.

Overview of RMSF-based dynamic analyses of the ROR1 pseudokinase domain. (A) Comparison of global mean RMSF values averaged over all backbone residues for the Apo, Ponatinib-bound, and designed analog (DM1 – DM8) systems. (B) Global ΔRMSF comparison calculated as the difference between ligand-bound and apo mean RMSF values (Ligand-Apo). A negative shift indicates ligand-induced suppression of overall protein dynamics relative to the apo state, whereas a positive shift reflects enhanced global conformational dynamics in the ligand-bound state, consistent with adaptive flexibility rather than structural instability. (C) Residue-wise RMSF profiles of backbone atoms for the ROR1 pseudokinase domain, illustrating spatial distribution of flexibility across the protein sequence. (D) Residue-wise ΔRMSF profiles (LigandApo), where negative values indicate ligand-induced suppression of local fluctuations consistent with direct or allosteric stabilization, and positive values reflect enhanced local dynamics suggestive of conformational relaxation or adaptive motion distal to the binding site. (E) Classification of highly fluctuating residues based on ΔRMSF modulation categories, highlighting ligand-dependent redistribution of dynamic hotspots. (F) Density distributions of RMSF values across all systems, illustrating global shifts in flexibility landscapes induced by ligand binding.

11.

Multivariate and correlation analyses of RMSF-derived dynamics for the ROR1 pseudokinase domain. (A) Heatmap representation of residue-wise RMSF values across all systems, including the Apo, Ponatinib-bound, and DM1– DM8 complexes, highlighting ligand-dependent fluctuation patterns. (B) Principal component analysis (PCA) showing the variance explained by the first ten principal components (PC1–PC10), with PC1 accounting for 54.8% of the total variance. (C) PCA score plot projected onto the first two principal components (PC1 vs PC2), illustrating separation and clustering of systems based on collective dynamic behavior. (D) Comparison of residue loadings for PC1 and PC2, with the top 20 contributing residues highlighted for each component, indicating key regions driving dominant collective motions. (E) RMSF correlation matrix across all systems (Apo, Ponatinib, and DM1–DM8), depicting similarity and divergence in residue-level fluctuation patterns. (F) Hierarchically clustered RMSF correlation heatmap, emphasizing groups of systems with shared dynamic signatures. (G) Dendrogram (cluster tree) derived from RMSF correlation analysis, annotated with bootstrap probability (BP%) values to assess cluster robustness.

3.8.2. Apo versus Ponatinib: Localized Rigidification without Global RMSF Collapse

The Apo ensemble (1 μs) displayed high intrinsic mobility (mean RMSF = 0.1709, SD = 0.1358 nm) with pronounced peaks concentrated in the four hotspot regions noted above (Figure A). Ponatinib binding did not collapse the global RMSF average (mean RMSF = 0.1767, SD = 0.1670 nm), but it produced clear localized suppression of fluctuations within the principal mobile segmentsespecially 10–25, 45–65, and 120–150while permitting modestly increased motion at select solvent-exposed residues. This apparent paradox (similar global mean RMSF yet clear rigidification in key segments) indicates that Ponatinib stabilizes functionally coupled loops while redistributing residual flexibility elsewhere, consistent with an induced-fit or conformational selection mechanism localized to hinge/activation-segment neighborhoods rather than generalized protein stiffening (Figure A–F).

3.8.3. Analog Series: a Graded Spectrum from Strong Stabilizers to Apo-like Flexibility

To quantify analog-dependent modulation relative to intrinsic Apo fluctuations, residue-wise ΔRMSF profiles (ligand-Apo) were interpreted using a statistically grounded thresholding scheme derived from the Ponatinib-Apo distribution (mean = −0.0059 nm; SD = 0.0605 nm) (Figure B–D). Within this framework, the analogs spanned a broad stabilizing spectrum:

• Strong stabilizers (DM2, DM8): DM2 and DM8 exhibited the most negative mean ΔRMSF values (−0.0451 and −0.0500 nm) and produced dense stabilizing clusters within the Ponatinib-responsive regions, with additional suppression extending into 28–35 and 135–145. These effects yielded the lowest overall RMSF values (0.1258 and 0.1208 nm, respectively), indicating that these analogs stabilize compact, low-mobility ensembles through coordinated suppression of the dominant hotspot segments (Figure C–E).

• Moderate stabilizers (DM3, DM5, DM6): these ligands produced intermediate reductions (mean ΔRMSF ≈ −0.022 to −0.027 nm), dampening mobility most consistently in 45–65 and 120–150, but with greater residue-to-residue variability than DM2/DM8suggesting partial stabilization of the hotspot network without uniformly anchoring all flexible elements (Figure C–E).

• Weak stabilizers (DM1, DM4): DM1 and DM4 displayed near-baseline modulation (mean ΔRMSF −0.0035 and −0.0072 nm), with most residues remaining in the low-modulation band and only scattered moderate effects. Their RMSF patterns remained largely Apo-proximal, consistent with limited dynamic reprogramming despite stable global folds (Figure C–E).

• Flexibility-enhancing analog (DM7): DM7 uniquely increased local mobility (mean RMSF ≈ 0.1830 nm, mean ΔRMSF = +0.0121 nm), producing concentrated positive ΔRMSF clusters in 28–38 and 132–150. This destabilization-like signature was absent in the other ligands, marking DM7 as the most Apo-like (or more flexible-than-Apo in specific segments) system in the series (Figure C–E).

Together, these results establish that analog optimization can tune not only binding stability but also where flexibility is suppressed or amplifiedan important design axis because dynamic regulation often controls kinase/pseudokinase pocket accessibility and regulatory-state selection. A 50 ns block-based RMSF convergence analysis supported the overall robustness of the residue-level fluctuation profiles for most systems, although DM4 and DM6 showed comparatively weaker convergence; detailed metrics are provided in the Supporting Information (Table S14 and Figure S1).

3.8.4. Multivariate Mapping of RMSF Landscapes: PCA Defines Stabilization Axes and Residue Drivers

To integrate residue-wise fluctuation patterns across systems, PCA of the RMSF matrix identified two dominant dynamical axes. PC1 (54.8% variance) captured the primary flexibility–stabilization gradient, ordering systems in a chemically interpretable hierarchy (Figure , A-C). The strongest stabilizers DM8 and DM2 occupied the extreme stabilized end of PC1 (large positive scores), followed by an intermediate tier (DM3 > DM6 > DM5). DM1 and DM4 clustered near the Apo state, indicating weak stabilization, while DM7 lay at the opposite extreme, consistent with enhanced flexibility. PC2 (9.6% variance) resolved ligand-specific redistribution signatures, particularly within hinge-adjacent and activation-segment regions, separating systems that stabilize the same hotspots via different residue-level patterns (Figure B,C).

PCA loadings localized the origin of system separation to the same RMSF hotspot network10–25, 28–38, 45–65, 120–150demonstrating that global dynamical differences are not diffuse but instead driven by a reproducible subset of mechanically dominant segments (Figure D). PC2 loadings were more localized (notably near 28–35 and 135–150), consistent with ligand-specific tuning of hinge/activation-segment coupling. Mechanistically, these loading profiles indicate that the analog series primarily differs in its ability to suppress (or amplify) fluctuations within a small, functionally relevant flexibility network that likely governs breathing motions around the ATP-binding cleft.

3.8.5. Dynamical Similarity and Robustness: Correlation Structure Supports Two Regimes but Limits Over-Interpretation of Fine Branching

Pairwise correlation of residue-wise RMSF profiles showed uniformly high similarity across systems (r ≈ 0.806–0.959), indicating that all trajectories share a conserved topological pattern of which residues tend to be flexible versus rigid, even as fluctuation amplitudes differ substantially (Figure E–G). Hierarchical clustering of 1 – r distances separated systems into two broad regimes: a Ponatinib-like rigidifying ensemble (Ponatinib with DM4/DM5/DM6 and DM7 in the same neighborhood) and a more heterogeneous, Apo-proximal ensemble (Apo with DM1/DM2/DM3/DM8). Importantly, this separation reflects similarity in the shape of residue-wise RMSF profiles, not necessarily the magnitude of stabilizationexplaining why strongly stabilizing ligands such as DM2 and DM8 can cluster with Apo in correlation space if they preserve the same peak topology while globally reducing amplitudes.

Bootstrap-based pvclust-style support further indicated that only a subset of fine-grained dendrogram relationships is reproducible (Figure G). The DM1–DM3 pairing showed the strongest support, whereas many higher-order branchesincluding larger Ponatinib-centered groupings and Apo-containing nodesdisplayed modest support, suggesting sensitivity to sampling noise and emphasizing that the dendrogram should be interpreted at the level of major regimes rather than over-read at fine branching resolution.

3.8.6. Integrated Interpretation and Design Implications

Collectively, the RMSF, ΔRMSF thresholding, PCA, and correlation/bootstrapping analyses converge on a coherent mechanistic model: ligand binding modulates ROR1 dynamics primarily through a hotspot network spanning N-terminal/hinge-adjacent loops, β-hairpin elements, and activation-segment–proximal regions. Within this framework, DM2 and DM8 emerge as the most effective stabilizers by strongly suppressing fluctuations across multiple hotspots, DM3/DM5/DM6 provide intermediate stabilization with localized variability, DM1/DM4 remain largely Apo-like, and DM7 uniquely enhances flexibility in key regulatory segments. These results support the idea that analog optimization can be guided not only by static binding interactions but also by dynamic state control, enabling rational selection of chemotypes that enforce desired rigidity patterns within the region’s most responsible for conformational breathing of the pseudokinase domain.

3.8.7. Residue-Resolved Interaction Network Governing Ponatinib Binding to ROR1

To define the atomic determinants underlying Ponatinib recognition by the ROR1 pseudokinase domain, we performed residue-resolved interaction fingerprinting using ProLIF over the full 1 μs molecular dynamics trajectory (Figures and A,D). This analysis reveals that Ponatinib binding is governed by a compact and highly conserved interaction network dominated by hydrophobic enclosure and aromatic stacking, reinforced by a single, persistent acidic anchor.

12.

Comparison of ProLIF-derived residue interaction frequency (%) heatmaps for Ponatinib (ligand 0LI) bound to the 6TU9 pseudokinase domain over the 250 ns and 1 μs trajectories. Heatmaps show the percentage of simulation frames in which each residue participates in a given interaction type: (A) hydrophobic, (B) H-bond donor, (C) H-bond acceptor, and (D) π-stacking interactions. Overall, the interaction pattern is largely preserved across both trajectories.

13.

ProLIF interaction frequency from 250 ns MD trajectory and their analysis. (A) Hydrophobic interactions (B) H-bond donor interactions (C) H-bond acceptor interactions (D) Pi-stacking interactions.

3.8.8. Hydrogen-Bonding Interactions: Selective But Functionally Asymmetric

Hydrogen-bond interactions were limited in number and generally intermittent, indicating that they play a supporting rather than dominant role in binding stabilization. Among donor interactions, Glu71 and Val84 occasionally donated hydrogen bonds to Ponatinib acceptor groups, with modest occupancies (8.19 and 0.90%, respectively), suggesting transient anchoring during specific conformational states. In contrast, hydrogen-bond acceptor interactions were strongly asymmetric: while several residues (Val84, Val160, His161, Lys162) engaged sporadically, Asp181 and Ser180 emerged as the principal polar anchors, with Asp181 exhibiting near-continuous acceptor engagement (∼88% occupancy) and Ser180 showing substantial but more variable participation (Figure B,C).

This asymmetric hydrogen-bonding pattern indicates that Ponatinib binding is not driven by a dense hydrogen-bond network typical of classical kinases but instead relies on a single dominant acidic anchor (Asp181) supplemented by transient polar contacts that assist orientation rather than enforce binding (Figure B,C).

3.8.9. Aromatic sTacking Defines Ligand Pose Fidelity

Aromatic interactions were remarkably persistent and represent one of the most stabilizing features of the Ponatinib–ROR1 complex. Phe100 and Tyr102 formed continuous π–π stacking interactions with Ponatinib’s fused aromatic core, with occupancies of ∼96 and ∼78%, respectively. These residues create a rigid aromatic clamp that locks the ligand into a precise orientation throughout the trajectory, minimizing rotational and translational drift. The exceptional persistence of these interactions highlights aromatic stackingnot hydrogen bondingas the primary determinant of pose fidelity and van der Waals stabilization in the ROR1 binding pocket (Figure D).

3.8.10. Hydrophobic Enclosure Dominates the Interaction Landscape

Hydrophobic interactions constituted the most extensive and persistent interaction class. A total of 31 residues contributed measurable hydrophobic contacts, but only a restricted subset formed a rigid, conserved core. Among these, Met75, Phe100, Tyr102, Leu154, Leu170, Ala52, and Phe159 maintained near-continuous engagement (>90–100% of frames), defining a structural hydrophobic cage surrounding the inhibitor scaffold. This core is complemented by semiflexible residues (e.g., Val84, Ile103, Ser180, His161) that adapt to substituent-dependent differences without disrupting the central binding architecture (Figure A).

Notably, Asp181 exhibits dual functionality, behaving as both a persistent hydrophobic contact and the dominant hydrogen-bond acceptor across the trajectory. This unusual dual role positions Asp181 as the keystone residue of the binding site, simultaneously contributing electrostatic stabilization and nonpolar packing.

Together, the ProLIF analysis establishes that Ponatinib binding to ROR1 is driven by hydrophobic enclosure and aromatic stacking, with hydrogen bonding playing a secondary, orientational role. This interaction hierarchy explains the high binding persistence observed despite the limited hydrogen-bond network and underscores the noncanonical nature of ROR1 pseudokinase recognition.

3.8.11. Interaction Conservation and Variability Across the DM1–DM8 Analog Series

To assess whether these interaction principles generalize across the designed analogs, ProLIF analysis was extended to DM1– DM8 using 250 trajectory frames per system. Across the series, hydrophobic interactions overwhelmingly dominated, followed by π-stacking, with hydrogen bonds remaining sparse and ligand-selective (Figure A–D). Because the designed analogs were simulated for 250 ns, an additional comparison was performed for Ponatinib (0LI) between the 250 ns and 1 μs trajectories to evaluate whether the shorter time scale captures the principal interaction pattern (Figure ). This analysis showed that the overall binding profile was largely preserved, with only modest changes across most residue-interaction pairs. The most notable differences were mainly confined to hydrophobic contacts, including reduced frequencies at Hse161, Leu184, Val160, Ile83, and Ile179, together with increased frequencies at Lys162 and Leu74. In contrast, hydrogen-bond donor, hydrogen-bond acceptor, and π-stacking interactions remained largely stable, with Phe100 essentially unchanged and Tyr102 showing only a moderate decrease. Overall, the longer trajectory indicates refinement of selected hydrophobic contacts rather than major reorganization of the interaction network, supporting the use of the 250 ns analog trajectories for comparative interaction analysis.

3.8.12. Conserved Hydrophobic Core Across All Inhibitors

Of 39 residues screened, only 24 exceeded the >10% hydrophobic occupancy threshold, and within this subset a highly conserved coreincluding Ala52, Met75, Phe100, Tyr102, Leu154, Phe159, Leu170, and frequently Ile35maintained near-continuous engagement across all ligands. Among these, Met75, Phe100, Tyr102, Leu154, and Asp181 consistently exhibited the highest occupancies (often >95–100%), identifying them as the dominant hydrophobic determinants of binding mode conservation (Figure A).

Semiflexible residues (e.g., Leu27, Leu74, Val84, Ile103, Ser180, Hse161) displayed moderately high but ligand-dependent occupancies, enabling subtle adaptation to analog-specific substituents without perturbing the core architecture. Peripheral residues showed wide variability, reflecting sensitivity to pocket breathing and ligand tail orientation.

Ligand-specific deviations were minimal: only DM6 uniquely engaged Lys162, while DM7 extended into minor peripheral subpockets involving Tyr153 and Hse157. Importantly, these deviations did not disrupt the conserved hydrophobic core, explaining the broadly similar binding poses observed across the series (Figure A).

3.8.13. Hydrogen Bonding: Dominated by a Single Conserved Anchor

Across DM1–DM8, hydrogen-bond acceptor interactions were generally weak and sporadic, with Asp181 standing out as the sole universally persistent acceptor (84–96% occupancy across all ligands). Ser180 showed ligand-dependent variability, while all other residues participated only transiently. Donor interactions were even rarer and highly selective; notably, Asp181 acted as a strong donor only for DM5, indicating a unique geometric complementarity absent in other analogs (Figure B,C).

These results confirm that hydrogen bonding does not define binding mode conservation in this system, instead serving as a fine-tuning mechanism layered atop a hydrophobic–aromatic scaffold.

3.8.14. π-Stacking Is Strictly Conserved

π-Stacking interactions were almost exclusively mediated by Phe100 and Tyr102, with Phe100 exhibiting uniformly high occupancies (>90%) across all analogs. Tyr102 contributed moderately but consistently, whereas Phe159 played no meaningful role. This conserved aromatic clamp enforces scaffold alignment and explains the minimal ligand-to-ligand variability in pose orientation (Figure D).

3.8.15. A Unified Three-Component Binding Motif for ROR1 Inhibition

Integrating interaction data across Ponatinib and DM1– DM8 reveals a robust three-component binding motif that governs recognition and persistence within the ROR1 pseudokinase pocket (Figures and ):

• 1.Hydrophobic enclosure: a rigid core of Met75, Phe100, Tyr102, Leu154, Leu170, Ala52, and related residues forms a nonpolar cage that dominates binding energetics and pose stability.
• 2.Aromatic stacking clamp: persistent π-stacking by Phe100 (primary) and Tyr102 (secondary) locks the ligand’s aromatic scaffold into a fixed orientation.
• 3.Acidic anchoring by Asp181: Asp181 uniquely combines near-continuous hydrophobic contact with persistent hydrogen-bond acceptor behavior, acting as the principal electrostatic and structural anchor across all inhibitors.

This interaction architecture explains both the high binding persistence and the limited structural variability observed across the DM series. Importantly, it highlights that effective ROR1 inhibition is achieved not through dense polar networks but through hydrophobic and aromatic dominance reinforced by a single, strategically positioned acidic residue.

3.8.16. Relative Binding Free Energy Landscape of Ponatinib and Designed Analogs

To quantify the thermodynamic basis of Ponatinib recognition by the ROR1 pseudokinase domain, relative binding free energies were estimated using MM-PBSA analysis over a 1 μs molecular dynamics trajectory. Ponatinib exhibits a relative MM-PBSA favorable binding free energy (ΔG_bind = −40.23 ± 5.61 kcal mol^–1), establishing internal energetic benchmark for comparison with the designed analogs (Figure A–D). This value reflects the known systematic overestimation of single-trajectory MM-PBSA relative to experimental affinities (estimated experimental ΔG ≈ −8 to −9 kcal mol^–1 from DSF and biochemical data, and is used here as a relative benchmark rather than an absolute affinity prediction. ,, Decomposition of the binding free energy reveals that ligand association is dispersion-dominated, with a large favorable van der Waals contribution (VDWAALS = −61.29 kcal mol^–1) supplemented by stabilizing electrostatic interactions (EEL = −26.18 kcal mol^–1). These gas-phase interactions (GGAS = −87.47 kcal mol^–1) are partially offset by an unfavorable solvation penalty (GSOLV = +47.24 kcal mol^–1), primarily arising from polar desolvation (EPB = +53.33 kcal mol^–1), with a modest compensatory contribution from nonpolar solvation (ENPOLAR = −6.09 kcal mol^–1).

14.

Data-driven MM-PBSA/RBFE analysis of DM ligands identifies binding governed by dispersion–solvation balance. (A) Relative binding free energy (ΔG_bind; mean ± SD) ranked from most favorable (more negative) to least favorable, identifying DM4 as the most favorable ligand, followed closely by DM7, DM6, and DM2, with DM1 showing the weak RBFE among all. (B) Compensation between gas-phase (GGAS) and solvation (GSOLV) contributions shown as a dumbbell plot, with TOTAL ΔG_bind marked for each ligand; this highlight energetic self-compensation where strong gas-phase stabilization can be offset by solvation penalties. (C) Decomposition of GGAS into van der Waals (VDWAALS) and electrostatic (EEL) contributions, showing dispersion dominates the favorable gas-phase term across the series. (D) Decomposition of GSOLV into polar (EPB) and nonpolar (ENPOLAR) contributions, showing polar solvation is the principal unfavorable penalty.

This energetic signaturestrong dispersion and electrostatic stabilization balanced by a substantial but controlled solvation penaltyis consistent with Ponatinib’s deep embedding within a hydrophobic pocket reinforced by a limited number of persistent polar contacts. The small standard errors across all energetic components (≤0.25 kcal mol^–1) confirm that these estimates are statistically stable over the microsecond time scale.

3.8.17. Energetic Profiles of the Designed DM1–DM8 Analogs

To assess whether the designed analogs reproduce this energetic regime, MM-PBSA calculations were performed for each DM ligand using equilibrated 250 ns trajectories. Across the series, all analogs rank within a narrow relative energetic window compared to Ponatinib (ΔG_bind ≈ −37 to −45 kcal mol^–1), indicating broadly comparable relative binding free energy estimates within the resolution afforded by the MM-PBSA framework. DM4 exhibits the most favorable MM-PBSA ranking (−45.20 ± 5.39 kcal mol^–1), while DM7, DM6, and DM2 form a closely grouped set of high-affinity ligands (≈−44 to −45 kcal mol^–1). Given the similar standard deviations (≈5–6 kcal mol^–1), these compounds should be regarded as a high-affinity cluster rather than as statistically resolved rank-ordered binders. At the opposite end of the spectrum, DM1 displays the least favorable binding free energy (−37.32 ± 5.68 kcal mol^–1) (Figure A–D).

Because raw trajectory means and frame-wise standard deviations alone are insufficient for statistically robust fine ranking, additional block analysis was performed to assess the temporal stability of the MM-PBSA estimates. Specifically, the ΔGbind time series for each ligand was divided into five contiguous blocks, and block-wise mean values were calculated. The corresponding block-based standard errors were 0.74 kcal mol^–1 for DM1, 1.97 for DM2, 0.74 for DM3, 1.08 for DM4, 0.71 for DM5, 1.36 for DM6, 0.81 for DM7, and 0.54 for DM8. Under this more conservative treatment, several pairwise differences among the top-performing ligands were not statistically robust. Accordingly, the most defensible interpretation is tier-based rather than strictly rank-ordered: DM4, DM7, DM6, and DM2 define the upper energetic tier; DM3 and DM8 form an intermediate tier; DM5 is somewhat less favorable; and DM1 is the weakest binder in the series (Table S16).

Block-wise trends further clarified the temporal behavior of the individual analogs. DM4 became progressively more favorable over the course of the trajectory, with mean block values shifting from −42.63 kcal mol^–1 in the first block to −47.66 kcal mol^–1 in the final block, indicating increasing stabilization in the later simulation segments. DM6 showed a similar pattern and ended with the most favorable late-block value in the series (−47.89 kcal mol^–1). In contrast, DM2 displayed marked interblock heterogeneity, with highly favorable early block values (approximately −48 kcal mol^–1) followed by less favorable late-block values (approximately −39.5 to −39.9 kcal mol^–1), explaining its comparatively large block-based SEM and suggesting less uniform convergence across the full trajectory. DM7 remained strongly favorable throughout but showed moderate block-to-block variation, whereas DM3 and DM8 retained intermediate mean energies with comparatively smaller temporal drift. DM1 remained the least favorable analog across all five blocks, although it showed modest improvement in the later trajectory segments (Table S16).

Importantly, binding trends across the DM series cannot be rationalized by maximizing any single energetic component. For example, DM3 exhibits the strongest gas-phase stabilization (GGAS = −111.00 kcal mol^–1) and the most favorable electrostatic contribution (EEL = −44.80 kcal mol^–1), yet does not emerge as a top binder because these gains are largely negated by a substantial solvation penalty (GSOLV = +68.33 kcal mol^–1), dominated by polar desolvation. In contrast, the high-affinity cluster (DM4, DM2, DM6, DM7) achieves binding energies comparable to Ponatinib by maintaining strong van der Waals stabilization (≈−63 to −68 kcal mol^–1) coupled with moderate electrostatics and restrained solvation penalties (≈+47–49 kcal mol^–1). This balanced energetic regime minimizes cancellation between favorable gas-phase interactions and unfavorable solvation, resulting in robust net binding (Figure A–D).

The weakest binder, DM1, illustrates the converse scenario: although it incurs the smallest solvation penalty among the analogs, its comparatively weak gas-phase stabilization results in the least favorable overall binding free energy. These observations highlight that balanced dispersion–electrostatic stabilization with controlled desolvation, rather than extreme electrostatics or minimal solvation cost alone, defines optimal binding within the ROR1 pocket (Figure A–D).

3.8.18. Comparison to Ponatinib and Implications for the Analog Design

Using Ponatinib as a validated energetic reference, the DM ligands collectively recapitulate a Ponatinib-like energetic balance. Ponatinib itself exhibits nearly identical binding free energies when evaluated over 250 ns (−41.78 ± 5.65 kcal mol^–1) and 1 μs (−40.23 ± 5.61 kcal mol^–1) trajectories, with closely matched energy decomposition profiles. This agreement indicates that the dominant energetic features of binding are already captured on the 250 ns time scale, with longer simulations primarily reducing statistical uncertainty rather than altering the energetic landscape (Figure A–D).

Several DM analogs (DM2, DM4, DM6, DM7) achieve mean ΔG_bind values within the same energetic regime as Ponatinib, differing by less than the associated statistical uncertainty. These results therefore support energetic equivalence rather than superiority, indicating that rational scaffold modification can preserve the delicate dispersion–solvation balance required for stable ROR1 engagement. In conjunction with structural and dynamical analyses, the MM-PBSA results suggest that successful analog design hinges on maintaining Ponatinib-like hydrophobic embedding while avoiding excessive polar exposure that would incur prohibitive desolvation costs (Figure A–D).

3.9. Physicochemical and ADME Profiling of Fukui-Guided Analogs

To assess whether the designed analogs retain drug-like physicochemical properties while preserving Ponatinib-like binding energetics, in silico ADME profiling was performed using SwissADME. This analysis provides a rapid, structure-based evaluation of lipophilicity, polarity, solubility, absorption, and metabolic liability, enabling early identification of favorable chemical space prior to experimental testing (Figure A–D and Table S17).

15.

Integrated ADME profiling of Ponatinib and DM1– DM8 analogs. (A) Heatmap summarizing in silico ADME properties of Ponatinib and DM1–DM8. Each descriptor was min–max normalized (0–1) across the compound set. For directionality, lipophilicity (consensus LogP) and Lipinski rule violations were inverted (lower values correspond to more favorable properties), whereas solubility (ESOL LogS) was not inverted (higher, less negative values indicate improved solubility). Cells report raw/normalized values; TPSA is reported in Ų and Lipinski values denote the number of rule violations. The color scale ranges from least favorable (0) to most favorable (1), with grid lines delineating individual cells. (B) Chemical space projection of lipophilicity versus polarity, plotting consensus LogP against TPSA (Ų). Dashed reference lines at TPSA = 90 Ų and Log P = 4.5 indicate commonly used thresholds for permeability and lipophilicity. Compound labels were offset to minimize overlap; DM8 was positioned on the opposite side to avoid collision with DM2. (C) Comparative ADME radar plot displaying normalized descriptors (as in panel A) for the top four compounds ranked by composite ADME score. Axes include inverted Log P, Log S, TPSA, bioavailability score, and inverted Lipinski violations, enabling visualization of multidimensional ADME balance. (D) Composite ADME score ranking (0–1), calculated using weighted contributions of inverted Log P (0.25), Log S (0.25), TPSA (0.20), bioavailability score (0.20), and inverted Lipinski violations (0.10). Higher scores indicate a more favorable overall ADME profile. Across the series, DM5 ranks highest, driven by low lipophilicity (Log P = 2.10), improved solubility (Log S = −4.34), high polarity (TPSA = 134.3 Ų), preserved bioavailability (0.55), and a single Lipinski violation. DM3 and DM4 also display favorable ADME balance, combining increased polarity (TPSA ≈ 109 Ų), moderate lipophilicity, and acceptable drug-likeness. Ponatinib occupies an intermediate position, whereas DM2 and DM8 rank lowest due to reduced predicted bioavailability (0.17), two Lipinski violations, higher lipophilicity, and poorer solubility.

All compoundsincluding Ponatinib and the DM1–DM8 analogsare predicted to exhibit high gastrointestinal absorption, indicating that none of the Fukui-guided scaffold modifications compromise oral permeability. In contrast, all DM analogs are predicted to lose blood–brain barrier permeability relative to Ponatinib, reflecting a systematic increase in molecular polarity. This shift away from CNS exposure is pharmacologically advantageous for oncology-focused indications, where peripheral selectivity is preferred.

A clear polarity–lipophilicity gradient emerges across the series. Analogs DM3–DM5 occupy a more polar physicochemical regime, with elevated topological polar surface areas (TPSA ≈ 108–134 Ų) and correspondingly improved predicted aqueous solubility relative to Ponatinib (ESOL Log S – 5.73). Among these, DM5 shows the most pronounced solubility enhancement (Log S ≈ −4.34) and the lowest lipophilicity (Log P ≈ 2.10), consistent with a favorable balance between polarity and hydrophobicity. By contrast, DM1, DM2, DM6, and DM8 shift toward higher lipophilicity (Log P ≈ 4.75–5.11) with poorer predicted solubility (Log S ≈ −6.37 to −6.74), delineating a less favorable physicochemical space for oral drug development (Figure A–D).

Drug-likeness metrics reinforce this distinction. DM3–DM5 retain only a single Lipinski rule violation and exhibit predicted bioavailability scores (≈0.55) comparable to Ponatinib, whereas DM2 and DM8 incur two Lipinski violations and show markedly reduced predicted bioavailability (≈0.17). Predicted cytochrome P450 inhibition profiles are broadly conserved across the series, indicating no systematic improvement or deterioration in metabolic liability. However, DM5 displays a distinct metabolic trade-off, losing predicted CYP2C19 inhibition while gaining CYP3A4 inhibition, underscoring that targeted physicochemical optimization can redistribute, rather than eliminate, metabolic risk.

3.10. Integration with Binding Energetics and Interaction Mechanisms

Importantly, the physicochemical trends observed in SwissADME predictions are consistent with binding energetics derived from MM-PBSA analysis (Figures and A–D). DM3–DM5 reside within the same binding free energy window as Ponatinib, within the statistical resolution of 250 ns simulations, indicating that increased polarity and solubility do not compromise binding stability. Molecular dynamics-derived interaction fingerprints further support this conclusion: these analogs preserve the hinge-proximal interaction triad (Glu71-Ile103-Asp181) and the Phe100-Tyr102 aromatic clamp characteristic of Ponatinib binding. Subtle differences among the triosuch as enhanced hydrogen-bond engagement for DM5 or reinforced hydrophobic and π-stacking interactions for DM3 and DM4reflect fine-tuning rather than disruption of the core binding mode.

By contrast, highly lipophilic analogs such as DM2 and DM8, although capable of achieving strong MM-PBSA binding energies, exhibit less favorable predicted solubility, bioavailability, and drug-likeness profiles. These compounds therefore define the upper lipophilicity boundary beyond which physicochemical liabilities begin to outweigh gains in binding affinity.

3.11. Medicinal Chemistry Implications

Taken together, SwissADME-based profiling delineates a coherent structure–property–energetics landscape for the Fukui-guided analog series. Scaffold modifications successfully tune polarity and solubility while preserving oral absorption and Ponatinib-like binding energetics, demonstrating that electronic optimization can be coupled to favorable ADME characteristics. From a medicinal chemistry perspective, DM3, DM4, and DM5 emerge as the most compositionally balanced analogs, combining improved physicochemical properties, acceptable drug-likeness, retained binding stability, and conserved molecular interaction motifs (see Supporting Information animation files, 0LI.mpg, DM3.mpg, DM4.mpg and DM5.mpg). In contrast, more lipophilic analogs such as DM2 and DM8 highlight the physicochemical limits of the scaffold and serve as useful boundaries for further optimization. Interestingly, among all DM1– DM8 analogs, only DM6 has been found to be already present in the PubChem database search (having CID no. 118494438).

4. Conclusions

In this work, we present a quantum-informed, multiscale computational strategy for the rational optimization of small-molecule inhibitors targeting the ROR1 pseudokinase domain. By integrating density functional theory-derived charge distributions and Fukui reactivity indices with structure-based design, long-time scale molecular dynamics, interaction fingerprinting, relative binding free-energy analysis, and in silico ADME profiling, we demonstrate how electronic reactivity information can be translated into chemically precise and dynamically validated ligand modifications.

Fukui-guided substitutions at electronically and metabolically labile positions (C10, C11, C32, and C35) enabled systematic tuning of polarity, solubility, and interaction strength while preserving the core binding architecture required for ROR1 engagement. Microsecond-scale simulations of apo and Ponatinib-bound ROR1 established a rigorous dynamic reference framework, revealing that ligand binding stabilizes functionally relevant loop networks without inducing global rigidification. Across the DM1–DM8 series, all analogs maintained binding free energies within a narrow window comparable to Ponatinib, confirming that electronic optimization can be achieved without compromising thermodynamic stability.

Crucially, integration of binding energetics with physicochemical and pharmacokinetic prediction differentiated developability among energetically similar compounds. While several analogs achieved Ponatinib-like binding energies, DM3–DM5 emerged as the most compositionally balanced candidates, combining enhanced polarity and aqueous solubility, preserved oral absorption, acceptable drug-likeness, and retention of the conserved hinge-anchored interaction triad and aromatic clamp that define stable ROR1 binding. These results underscore the importance of multiobjective optimization, in which electronic structure, conformational dynamics, and ADME constraints are treated as coupled design variables rather than sequential filters.

More broadly, this study establishes the ROR1 pseudokinase domain as a tractable and chemically addressable target for small-molecule inhibition and demonstrates that quantum-informed reactivity mapping provides a transferable design principle for ATP-pocket-like systems in which classical docking or SAR approaches alone fail to capture subtle electronic and dynamic constraints. The framework described here is readily extensible to other pseudokinases and conformationally regulated targets, offering a general pathway toward next-generation kinase-family inhibitor design guided by first-principles chemistry.

Glossary

Abbreviations

DM

designed molecule

ATP

adenosine triphosphate

ROR1

receptor tyrosine kinase–like orphan receptor 1

MD

molecular dynamics

SAR

structure activity relationship

DFT

density functional theory

CLL

chronic lymphocytic leukemia

TNBC

Triple-Negative Breast Cancer

NSCLC

Non-Small-Cell Lung Cancer

MCL

Mantle Cell Lymphoma

PDAC

Pancreatic Ductal Adenocarcinoma

Wnt5a

Wnt Family Member 5A

STAT3

Signal Transducer and Activator of Transcription 3

PI3K/AKT

Phosphoinositide 3-kinase/Protein Kinase B

RAS/MAPK

Ras/Mitogen-Activated Protein Kinase pathway

ADCs

Antibody-Drug Conjugates

MMAE

Monomethyl Auristatin E

CRD

Cysteine-Rich Domain

ADME

Absorption, Distribution, Metabolism, and Excretion

RBFE

Relative Binding Free Energy

RMSD

Root Mean Square Deviation

RMSF

Root Mean Square Fluctuation

Rg

Radius of Gyration

PCA

Principal Component Analysis

MMPBSA

Molecular Mechanics Poisson–Boltzmann Surface Area

The study reported in this manuscript used Cartesian coordinates of ROR1 (6TU9) from the PDB database, which is freely available at https://www.rcsb.org/structure/6TU9. PDBs were examined in ChimeraX1.4 and VMD1.9.3. ORCA 6.1.0 has been used for DFT and Fukui charges analysis, the software can be freely available for download at https://orcaforum.kofo.mpg.de/app.php/portal. The PubChem database (https://pubchem.ncbi.nlm.nih.gov/) was used to download the Ponatinib/0LI and designed analogs search. Vina-GPU version 2.0 were used for the virtual screening and docking study. GROMACS 2022.4 was used for the MD simulation study and analysis. The trajectories of the MD simulations were analyzed using the appropriate commands provided in the GROMACS 2022.4 manual/tutorial. The 3D figures were generated using the UCSF chimera. The manuscript was written in the MS office, and some data were analyzed in MS Excel (version 2016). Open-source software Grace (version 5.1.25) and matplotlib with other Python libraries were used for the analysis of data and data graph production. Inkscape, an open-source package, was used for figures processing. ProLIF, a sophisticated Python library, was used to identify and create interaction fingerprints in 2D. The gmx_MMPBSA package was used to calculate relative binding free energies. Animation files (0LI.mpg, DM3.mpg, DM4.mpg and DM5.mpg) of the trajectories were generated using the VMD1.9.3 movie plugin. Supporting Information has been provided, along with the manuscript for the results and discussion. The trajectories are available from the corresponding authors upon request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsphyschemau.6c00012.

• Animation file of 0LI (MPG)

• Animation file of DM3 (MPG)

• Animation file DM4 (MPG)

• Animation file DM5.mpg (MPG)

• Detailed material and method sections, key quantum-chemical parameters of ponatinib, including complete atom-resolved charge values with explicit atom-numbering schemes; Mulliken and Löwdin reduced orbital population analyses (s, p, d, and f contributions); Mayer atomic charges, valence descriptors, and bond orders; and condensed Fukui reactivity indices (f ⁺, f ^–, f ⁰) for selected reactive atoms derived from both Mulliken- and Löwdin-based charge distributions. atoms not identified as Fukui-dominant yet intentionally retained without modification, highlighting their structural, electronic, or pharmacophoric relevance; correlation analyses between Mulliken- and Löwdin-derived condensed Fukui functions for ponatinib; docking scores for selected ponatinib analog modifications are reported alongside molecular dynamics simulation metrics, including mean RMSD values, equilibration times, Kolmogorov–Smirnov (KS) statistics comparing apo and ligand-bound systems, KS comparisons between apo and ponatinib complexes, and stability classifications derived from trajectory analysis; SwissADME-based physicochemical descriptors and pharmacokinetic profiling data at-a-glance assessment of compound developability; Hirshfeld-derived condensed Fukui functions for cross-comparison with Mulliken and Löwdin descriptors; five-block (50 ns) block-averaged backbone RMSD, radius of gyration (R _g), and SASA convergence metrics over the 0–250 ns trajectories; residue-level RMSF convergence metrics and plots across 50 ns simulation blocks for the ligand-bound systems; block-wise convergence summaries for the 0LI–pseudokinase domain complexes; and five-block MM-PBSA binding free energy analysis of the designed analogs against the ROR1 pseudokinase domain (PDF)

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.