Review Reports

Comments and suggestions

Reply

Comment 1: “These receptors have dozens or hundreds of structures resolved in the PDB, in complex with structurally diverse ligands. Why were these particular structures chosen?”

We thank the reviewer for this valuable observation and for highlighting the importance of explicitly justifying the receptor-template selection. This comment helped us clarify the selection criteria and improve the transparency of the structural screening process. We have clarified the structural criteria used to select the receptor templates: a wild-type human ligand-binding domain, high crystallographic resolution, a resolved ligand-binding pocket and AF-2/H12 region, and an experimentally determined reference-ligand pose suitable for redocking. These criteria guided the selection of PPARα (2P54; 1.79 Å), PPARγ (6MS7; 1.43 Å), and RORα (1N83; 1.63 Å). To make this selection transparent, we also added a record of the 17 screened structures, their selection status and rationale, and the corresponding wwPDB validation metrics.

Changes and specific locations:

·       Main manuscript, Section 2.1, first paragraph, page 4, lines 143–161: explicit selection criteria, identification of the selected templates, and reference to the supplementary screening record.

·       Supplementary Materials, Table S1, pages 2–4: screened structures, selection decisions, and supporting rationale.

·       Supplementary Materials, Table S2, pages 4–5: wwPDB model- and ligand-validation metrics.

Comment 2: “Are there no structures with bound ligands more similar to MaR1 or DHA?”

We appreciate the reviewer’s suggestion to consider chemically related ligand-bound structures. Such structures are available, and we have explicitly acknowledged endogenous fatty-acid-bound PPARα structures (Kamata et al., 2020) and PPARγ structures containing DHA or related oxidized fatty acids (Itoh et al., 2008). Table S1 now includes these chemically related structures as comparative references, including the DHA-bound PPARγ structure 2VV0, and explains their selection status. We clarified that primary template selection prioritized structural quality and independent reference-ligand poses suitable for redocking rather than maximal chemical similarity. This strategy was intended to limit preferential selection of a starting pocket conformation already adapted to either test ligand.

Changes and specific locations:

·       Main manuscript, Section 2.1, first paragraph, page 4, lines 151–156: acknowledgment of chemically related lipid-bound structures, supported by references [36,37], and explanation of the independent-template strategy.

·       Supplementary Materials, Table S1, page 2: PPARα fatty-acid-bound comparators 6LX7, 6LX8, and 7BQ1, with their selection status and rationale.

·       Supplementary Materials, Table S1, pages 3–4: DHA-bound PPARγ structure 2VV0 and related oxidized-fatty-acid complexes; the table note on page 4 explains their use as comparative structural references.

Comment 3: “For instance, the carboxylate group is a well-known pharmacophore in PPARs; what other anchoring points for this group have been observed across these structures, and are they contacted by the ligands studied here?”

We are grateful to the reviewer for highlighting the importance of relating the predicted contacts to established carboxylate-recognition sites. We expanded Section 2.1 to describe the canonical polar networks comprising Ser280, Tyr314, His440, and Tyr464 in PPARα and Ser289, His323, His449, and Tyr473 in PPARγ, supported by references [37,39].

Both ligands contacted these canonical regions in the selected docking poses. During MD, MaR1 maintained carboxylate-associated contacts with Tyr314 and Ser280 in PPARα, whereas DHA formed a network involving His440, Tyr314, Ser280, and Tyr464. In PPARγ, MaR1 sampled His323- and Tyr473-associated carboxylate arrangements with replica-dependent geometry, whereas DHA retained an alternative Arg288-centered anchor. The revised presentation distinguishes initial docking contacts from those sampled during MD and reports contact occupancy separately from geometric proximity.

Changes and specific locations:

·       Main manuscript, Section 2.1, page 4, lines 165–167: literature-supported description of the canonical PPAR carboxylate-recognition regions.

·       Main manuscript, Section 3.1, page 12, lines 514–520, and Table 2, page 11: predicted contacts in the selected initial docking poses.

·       Main manuscript, Section 3.3, pages 13–14, lines 584–607: replica-resolved PPARα and PPARγ contacts and carboxylate-associated distances.

·       Figures 2 and 3, pages 14 and 15, respectively, and Table 4, page 16: contact occupancies, oxygen-distance profiles, and reproducibility of the corresponding interactions.

Comment 4: “In this regard, the work would benefit from a more complete initial analysis of the different binding pockets: their shape (e.g., the three main arms or branches described for PPAR), amino acid composition, the presence of polar or charged residues, etc.”

Thank you for recommending this additional structural characterization. We expanded Section 2.1 to describe the branched, Y-shaped PPAR pockets, their principal arms, and the elongated sterol-binding cavity of RORα, together with their established polar regions, supported by references [35,37–40].

We also added Figure S2 to visualize pocket architecture and the distribution of polar and apolar residues. Table S4 provides SiteMap descriptors of the operationally defined sites centered on the co-crystallized ligands, including volume, enclosure, exposure, and hydrophobic/hydrophilic character. Table S5 reports residue identities, regional assignments, polar/apolar classification, distances to the reference ligand, and literature-supported polar anchors. These additions provide a structural framework for interpreting the receptor-specific contacts.

Changes and specific locations:

·       Main manuscript, Section 2.1, page 4, lines 162–171: expanded description of pocket architecture and polar recognition regions.

·       Main manuscript, Section 2.1, page 4, lines 185–187, and Section 3.1, page 9, lines 413–415: citations to the supplementary pocket characterization.

·       Supplementary Materials, Figure S2, page 34: pocket surfaces showing polar and apolar residue distributions.

·       Supplementary Materials, Table S4, pages 8–9: SiteMap descriptors and the operational definition of the analyzed sites.

·       Supplementary Materials, Table S5, pages 9–14: regional residue inventory, physicochemical classification, reference-ligand distances, and polar anchoring sites.

Comment 5: “MaR1 and DHA are highly flexible ligands, with many degrees of freedom, capable of exploring multiple conformations of similar energy. How many poses, then, can these ligands adopt within the receptor, and were the principal ones explored during docking?”

We acknowledge this sampling concern and thank the reviewer for encouraging a more transparent description of the conformational search. Table S6 now reports 16 rotatable bonds for MaR1 and 14 for DHA, together with 64 retained ConfGen conformers per ligand. These conformer counts describe the pre-docking search and do not quantify Glide’s internal conformational sampling.

The SP–XP workflow advanced the top 10% of SP-ranked poses to XP and requested up to 10 output poses per receptor–ligand pair. The actual outputs for MaR1 and DHA, respectively, were 10 and 1 XP poses in RORα, 4 and 2 in PPARα, and 4 and 1 in PPARγ. All returned poses were retained for pose-level assessment.

These counts describe the docking output, not the total number of accessible binding modes. We therefore do not claim exhaustive sampling or identification of every principal binding arrangement; this limitation is explicitly acknowledged in Section 4.5.

Changes and specific locations:

·       Main manuscript, Sections 2.2–2.3, page 5, lines 195–219: conformer generation, SP–XP workflow, requested versus returned poses, and citation to Table S6.

·       Supplementary Materials, Table S6 and its note, page 14: rotatable-bond counts, retained ConfGen conformers, system-specific XP outputs, and distinction between pre-docking conformers and Glide sampling.

·       Main manuscript, Section 4.5, pages 26–27, lines 1129–1144: explicit limitations concerning alternative orientations and non-exhaustive binding-mode sampling.

Comment 6: “This is not meant to question the docking method used, but rather to point out that selecting a single initial position based solely on the obtained scores could severely restrict the subsequent possibility of identifying the most favorable binding mode through MD simulation, which is the ultimate goal of the work.”

We appreciate this methodological concern and agree that a single starting pose can constrain the binding arrangements explored by conventional MD. We clarified that pose selection combined structural admissibility—occupation of the pocket defined by the co-crystallized ligand—with XP GScore ranking. Emodel was examined only as a secondary within-ligand descriptor, whereas pose-level Prime MM-GBSA was evaluated separately.

The four independent 250 ns replicas for each complex originated from the same selected pose, with distinct initial velocity seeds. They therefore assess the reproducibility of interactions sampled from a common starting configuration, rather than independently validating alternative binding modes. We explicitly acknowledge this limitation in Section 4.5 and restrict our interpretation to the evolution and local compatibility of the selected poses. Their retention within the pocket does not establish that they are unique or globally preferred binding arrangements.

Changes and specific locations:

·       Main manuscript, Section 2.4, page 5, lines 226–235: structural-admissibility criterion, XP GScore ranking, and supporting roles of Emodel and pose-level MM-GBSA.

·       Main manuscript, Section 2.5, page 6, lines 255–262: independent replicas generated using distinct initial velocity seeds and replica-level analysis.

·       Main manuscript, Section 4.5, pages 26–27, lines 1129–1144: explicit acknowledgment of the common starting pose, restricted sampling of alternative orientations, and distinction between local compatibility and a globally preferred binding mode.

Comment 7: “Seeking convergence with other docking methods could strengthen these results.”

Thank you for proposing this complementary validation strategy. We agree that agreement between poses generated by independent docking methods would strengthen confidence in the predicted binding arrangements. Our study used a hierarchical Glide SP–XP workflow; these are precision stages of the same docking platform and do not constitute validation using independent algorithms.

We have not performed a cross-method docking comparison. Section 4.5 therefore identifies comparison of poses generated using independent docking algorithms and receptor ensembles as a future validation step, followed by their evaluation through longer replicated or enhanced-sampling simulations. This acknowledges the remaining uncertainty without implying that agreement between independent methods has been demonstrated.

Changes and specific locations:

·       Main manuscript, Section 2.3, page 5, lines 202–219: description of the Glide SP–XP workflow and its sampling scope.

·       Main manuscript, Section 4.5, page 27, lines 1142–1144: recommendation to compare independently generated poses and receptor ensembles using additional replicated or enhanced-sampling simulations.

Comment 8: “Additionally, the positive Emodel value should be revised, as it does not indicate an optimal condition.”

We appreciate the reviewer’s attention to this value and agree that Emodel should not be interpreted as evidence of pose optimality. During revision, we identified an error in the DHA input structure, which contained two additional carbon atoms. After correcting it to the 22-carbon DHA scaffold, we updated the affected DHA docking and MD results throughout the manuscript and Supplementary Materials. The corrected Emodel value for PPARγ/DHA is −54.033 kcal·mol⁻¹, replacing the previously reported +18.940 kcal·mol⁻¹.

We clarified that Emodel serves as a secondary within-ligand pose-discrimination descriptor, not an estimate of binding affinity. Because Glide returned only one PPARγ/DHA XP pose, Emodel could not discriminate among alternative poses. The corrected value is therefore reported descriptively, and the pose was retained as a pocket-compatible MD starting configuration without claiming optimality.

Changes and specific locations:

·       Main manuscript, Section 2.2, pages 4–5, lines 189–200: correct DHA identity, 22-carbon molecular formula, double-bond geometry, and ionization state.

·       Main manuscript, Table 2, page 11, and Section 3.1, page 12, lines 520–523: corrected PPARγ/DHA Emodel value and explanation of the single returned XP pose.

·       Main manuscript, Section 2.4, page 5, lines 229–235: definition and restricted interpretation of Emodel, supported by references [47–49].

·       Main manuscript, Section 4.5, page 26, lines 1134–1140: single-pose limitation and distinction between pocket compatibility and pose optimality.

·       Supplementary Materials, Table S6 and its note, page 14: corrected Emodel value, XP pose count, and within-ligand interpretation.

Comment 9: “On the other hand, why are the docking-derived binding modes not shown in 3D?”

Thank you for highlighting the value of three-dimensional views for interpreting the initial binding arrangements. We revised Figure 1F–H to show receptor surfaces and enlarged pocket-level views for RORα, PPARα, and PPARγ, respectively. These views superimpose the highest-ranked structurally admissible XP poses of MaR1 and DHA with the corresponding crystallographic reference ligand and its redocked pose.

MaR1 and DHA are displayed in green and orange, respectively, and selected pocket residues are labeled to facilitate structural orientation. The revised caption explains these superpositions and clarifies that the labeled residues are illustrative. Complete two-dimensional interaction diagrams remain available in Figure S1.

Changes and specific locations:

·       Main manuscript, Figure 1F–H and its caption, page 10, lines 472–486: three-dimensional receptor and pocket views, ligand superpositions, and selected residue labels.

·       Main manuscript, Section 3.1, page 9, lines 406–412: references to the pocket-level comparisons and supplementary interaction diagrams.

·       Supplementary Materials, Figure S1, page 33: complete two-dimensional interaction diagrams of the selected XP poses used to initiate MD.

Comment 10: “It would be informative to compare the initial poses with the representative MD structures from each replica, to understand how the binding mode evolves over the course of each simulation.”

We are grateful for this suggestion to make changes in ligand positioning during MD easier to inspect. The selected initial docking poses and representative MD structures are presented separately in Figures 1F–H and 7, respectively. Figure 7 shows one observed trajectory frame per complex, selected from the global-minimum bin of its replica-balanced apparent FEL. Table S11 identifies the corresponding replica, frame, and simulation time.

To characterize variation across all four replicas, Figure S3B provides ligand heavy-atom RMSD profiles relative to the initially minimized bound pose after receptor alignment. At the same time, Figure S6 presents replica-resolved interaction summaries. These analyses complement the representative structural views.

However, the current revision does not include three-dimensional superpositions of the initial pose with a representative MD structure from every replica. The requested visual comparison is therefore only partially addressed, and Figure 7 should not be interpreted as providing that complete comparison.

Changes and specific locations:

·       Main manuscript, Figure 1F–H, page 10: selected initial docking poses.

·       Main manuscript, Section 3.6, pages 21–23, lines 883–889 and 942–965, and Figure 7, page 22 (caption, pages 22–23, lines 927–940): representative MD structures and their relationship to replica-level contact analyses.

·       Supplementary Materials, Table S11, page 28: selection procedure and replica, frame, and simulation-time identifiers.

·       Main manuscript, Section 2.6, page 6, lines 266–272, and Supplementary Figure S3B, page 36: ligand RMSD methodology and replica-resolved displacement profiles.

·       Supplementary Materials, Figure S6, caption on page 41: interaction summaries for each independent replica.

Comment 11: “In summary, given how strongly conventional MD depends on the initial structure of the complex, the central challenge lies in solidly justifying that choice. Both the receptor structure, extensively represented in the PDB, and ligands, which can adopt multiple conformations, require this justification”

We appreciate the reviewer’s emphasis on this central methodological issue. As detailed in our responses to Comments 1–8, we strengthened the receptor-selection rationale, added a comparative structural screening record, and documented ligand preparation, docking outputs, and pose-selection criteria. We also clarified that the simulations characterize interactions sampled from selected pocket-compatible starting configurations. The uncertainty associated with alternative starting poses remains explicitly acknowledged in Section 4.5, and the revised manuscript does not claim identification of a unique or globally preferred binding mode.

Changes and specific locations:

·       Main manuscript, Section 2.1, page 4, lines 143–161: receptor-selection criteria, template justification, and references to the structural screening and validation records.

·       Main manuscript, Sections 2.2–2.4, pages 4–5, lines 189–235: ligand preparation, docking workflow, sampling documentation, and representative-pose selection.

·       Main manuscript, Section 4.5, pages 26–27, lines 1129–1144: limitations of single-pose initialization and the need to evaluate alternative starting configurations.

Comments and suggestions

Reply

Comment 1: Page 1, lines 19-21: the sentence should be “Maresin 1 (MaR1), a pro-resolving mediator derived from docosahexaenoic acid (DHA), differs from DHA in its C7/C14 hydroxylation pattern and double-bond arrangement.”

Otherwise, “their” in the second sentence only applies to Maresin 1 and therefore it should be “its”.

Thank you for identifying this grammatical ambiguity. We have adopted the suggested wording, explicitly identifying DHA as the comparator. This also makes clear that “their” in the following sentence refers to both MaR1 and DHA.

Changes and specific locations:

·       Main manuscript, Abstract, Background/Objectives, page 1, lines 19–21: revised opening sentence specifying that MaR1 differs from DHA in its C7/C14 hydroxylation pattern and double-bond arrangement.

Comment 2: Page 5, lines 209-212: the phrasing “rather than an estimate of binding affinity” is misleading, as Emodel is not designed to estimate binding affinity in any context. Emodel is an internal pose-ranking term within Glide, not a predictor of ΔGbind. I suggest rephrasing to indicate that Emodel was used solely for pose assessment, consistent with its intended purpose. The same remark applies to page 10, line 447 and to page 10, lines 457-458.

We appreciate the reviewer’s clarification of Emodel’s intended purpose. We revised the text to define Emodel as Glide’s internal within-ligand pose-discrimination term and state that it was examined solely as a secondary pose-assessment descriptor, supported by references [47–49]. The ΔEmodel values in Table 2 are reported descriptively and were not used to rank chemically distinct ligands. We also clarified that, when only one XP pose was returned, Emodel could not discriminate among alternative poses. The corresponding wording is consistent across the Methods, Results, table notes, and Supplementary Materials.

Changes and specific locations:

·       Main manuscript, Section 2.4, page 5, lines 229–235: definition of Emodel and its restricted role in within-ligand pose assessment.

·       Main manuscript, Table 2 note, page 12, lines 508–512: descriptive reporting of ΔEmodel and clarification of its role.

·       Main manuscript, Section 3.1, page 12, lines 520–523, and Section 4.5, page 26, lines 1135–1138: inability to discriminate among alternative poses when only one XP pose was returned.

·       Supplementary Materials, Table S3 note, page 8, and Table S6 note, page 14: consistent description of Emodel’s intended within-ligand purpose.

Comment 3: Page 5, line 218, Molecular Dynamics and Simulation Setup: the MD timestep is not explicitly stated. Although RESPA sub-steps of 2 fs and 6 fs are mentioned, the fundamental integration timestep should be reported for clarity and reproducibility.

Thank you for requesting a clearer statement of the integration timestep. We now explicitly report a fundamental timestep of 2 fs using the RESPA multiple-time-step algorithm [56]. Bonded and short-range nonbonded interactions were evaluated every 2 fs, whereas long-range interactions were evaluated every 6 fs. This clarification makes the integration settings explicit for reproducibility.

Changes and specific locations:

·       Main manuscript, Section 2.5, page 6, lines 251–254: explicit fundamental integration timestep and evaluation intervals for bonded, short-range, and long-range interactions.

Comment 4:  Page 7, lines 297-298: the phrase “Because the trajectories were unbiased” is unclear as written. The authors should explicitly state what they mean by “unbiased” (e.g., absence of positional restraints or external biasing forces during production MD) and clarify that the FELs derived from RMSD and Rg represent relative populations sampled over the simulated timescale rather than converged thermodynamic free-energy surfaces. Given that production trajectories were run as standard equilibrium MD without external forces or positional restraints, it should be straightforward for the authors to specify the sense in which the trajectories are considered unbiased with respect to the structural descriptors used in the FEL analysis.

We appreciate this distinction between the absence of applied bias and thermodynamic convergence. We now explicitly state that production simulations used conventional NPT MD without positional restraints or external biasing potentials. No bias was applied along Cα RMSD, Rg, or any other collective variable.

We also clarified that RMSD and Rg are global structural descriptors and that the resulting FELs represent apparent relative free-energy projections of populations sampled over the simulated timescale. Their interpretation is descriptive and does not establish converged thermodynamic surfaces or quantitative transition barriers.

Changes and specific locations:

·       Main manuscript, Section 2.8, page 7, lines 324–326: explicit production conditions and absence of positional restraints or applied bias.

·       Main manuscript, Section 2.8, page 7, lines 327–331: descriptive interpretation of the RMSD–Rg projections and limits concerning convergence and transition barriers.

·       Main manuscript, Section 4.5, page 27, lines 1152–1154: acknowledgment that the FELs describe sampled distributions rather than converged thermodynamic free-energy surfaces.

Comment 5: Page 7, line 303: PyMOL is used for RMSD and Rg calculations (via the GeoMeasures plugin), but no citation is provided. PyMOL has its own canonical references (DeLano 2002; Schrödinger version), which should be included in the manuscript.

We thank the reviewer for noting this omission. We added the version-specific reference for The PyMOL Molecular Graphics System, Version 3.1 (Schrödinger LLC, 2025) as reference [59]. Section 2.8 now cites the Geo-Measures plugin [58] and PyMOL [59] separately when describing the RMSD and Rg calculations.

Changes and specific locations:

·       Main manuscript, Section 2.8, page 7, lines 331–334: identification of the Geo-Measures plugin, PyMOL version 3.1, and their respective citations.

·       Main manuscript, References [58–59], page 35, lines 1450–1452: bibliographic entries for the plugin and the PyMOL software.

Comment 6: Page 8, lines 343-344: the phrasing “rather than experimental binding affinities” is misleading, because MM-GBSA does not yield experimental affinities even in principle. MM-GBSA provides qualitative, relative energetic trends based on an endpoint approximation. I suggest rephrasing to indicate that the results are interpreted as relative energetic descriptors rather than quantitative binding free energies.

We are grateful for this clarification of the method’s interpretation. We removed the comparison with experimental binding affinities. Section 2.9 now describes MM-GBSA as an approximate endpoint method with incomplete treatment of conformational entropy and states that the calculated values were used only to describe relative energetic trends among the modeled complexes, not quantitative thermodynamic binding free energies [61,62].

Changes and specific locations:

·       Main manuscript, Section 2.9, page 8, lines 365–366 and 374–377: revised terminology identifying relative energetic estimates and explicit explanation of their interpretive limits, supported by references [61,62].

Comment 7:  Page 10, lines 439-449: in Table 2, ΔXPScore and ΔEmodel values are not reported for one of the ligands, although the table title indicates a comparative analysis. An explanation or the missing values should be provided for consistency.

We appreciate the reviewer’s request to clarify these comparative columns. We revised the table note to define ΔXP GScore and ΔEmodel as MaR1 − DHA within each receptor. Each Δ represents a single difference between the two ligands and is therefore displayed once per receptor. The note specifies the units and explains that negative values indicate a lower numerical value for MaR1, whereas positive values indicate a lower value for DHA. ΔEmodel is reported descriptively and was not used for interligand ranking.

Changes and specific locations:

·       Main manuscript, Table 2, page 11, and its note, pages 11–12, lines 502–510: definition of the differences, explanation of one Δ value per receptor, units, sign convention, and descriptive interpretation of ΔEmodel.

Comment 8:  Page 11, line 472, and Figure S2: the manuscript presents RMSD plots for the protein backbone (Cα) in each protein-ligand complex, but does not show. RMSD for the ligand alone. Ligand RMSD is typically required to evaluate binding-mode stability and to justify the MM-GBSA sampling region. Please consider including these data.

Thank you for recommending this additional structural diagnostic. We added ligand heavy-atom RMSD profiles for all six complexes, showing each of the four independent 250 ns replicas in Figure S3B. For each frame, receptor Cα atoms were aligned to the corresponding initially minimized complex before calculating ligand RMSD relative to the initially minimized bound pose. This procedure measures ligand positional and conformational displacement relative to the receptor.

The profiles are now cited in Section 3.2 and provide structural context for the production interval analyzed by MM-GBSA. Section 2.9 explicitly identifies that interval as 2,500 production frames per replica after excluding the initial reference frame. We do not use the RMSD profiles to claim thermodynamic convergence of these estimates.

Changes and specific locations:

·       Main manuscript, Section 2.6, page 6, lines 266–272: ligand RMSD calculation, receptor alignment, and structural interpretation.

·       Main manuscript, Section 3.2, page 12, lines 540–543: description of the profiles and citation to Figure S3B.

·       Supplementary Materials, Figure S3B and its caption, page 36: ligand RMSD profiles for all 24 trajectories and calculation procedure.

·       Main manuscript, Section 2.9, page 8, lines 367–369: production-frame ensemble used for MM-GBSA.

Comment 9:  Page 14, line 590: distances are reported in nanometres (e.g., RMSD = 0.193 nm, Rg = 1.899 nm), whereas earlier sections use Å units. For consistency and ease of interpretation, it would be helpful to use a single unit throughout or briefly note the unit change.

We appreciate the reviewer’s suggestion to clarify the unit convention. We retained nm for FEL RMSD–Rg coordinates and radius-of-gyration profiles, and Å for redocking RMSD, protein Cα RMSD/RMSF, ligand heavy-atom RMSD, and residue–ligand distances. We now explicitly state this convention and the conversion factor (1 nm = 10 Å) in Section 3.4 and in an opening note in the Supplementary Materials.

Changes and specific locations:

·       Main manuscript, Section 3.4, page 17, lines 660–663: explicit explanation of the quantities reported in nm and Å, including their conversion.

·       Supplementary Materials, “Unit convention” note, page 2: corresponding clarification covering structural descriptors, protein and ligand RMSD, RMSF, and residue–ligand distances.

Comment 10:  Page 16, line 715: the manuscript refers to the MM-GBSA values as “ΔGbind”. Since MM-GBSA does not include conformational entropy and does not yield absolute binding free energies, it would be more accurate to describe these values as MM-GBSA interaction energies or relative energetic estimates rather than ΔGbind.

Thank you for identifying this terminology issue. In Section 3.5, we have replaced “ΔGbind” with “total MM-GBSA score” and describe the calculations as providing “relative endpoint energetic estimates.” Section 2.9 also explicitly acknowledges the incomplete treatment of conformational entropy and limits interpretation to relative energetic trends, without assigning the scores a quantitative thermodynamic binding-free-energy interpretation.

Changes and specific locations:

·       Main manuscript, Section 3.5, page 20, lines 802–811: revised terminology describing the results as relative endpoint energetic estimates and total MM-GBSA scores.

·       Main manuscript, Section 2.9, page 8, lines 374–377: explicit methodological limitation concerning conformational entropy and thermodynamic interpretation.

Comment 11:  Page 17, lines 725-727: the phrasing suggests that MM-GBSA values could correspond to experimental binding affinities if statistical testing were applied. MM-GBSA does not yield experimental ΔGbind. This sentence should be modified.

We appreciate this clarification. We have removed the comparison with experimental binding affinities from the sentence concerning statistical testing. It now reads: “We did not apply inferential statistical tests to these numerical differences and interpreted them only as receptor-specific relative energetic trends.” The methodological limitations are stated separately, so the interpretation of MM-GBSA scores no longer appears conditional on whether statistical tests were performed.

Changes and specific locations:

·       Main manuscript, Section 3.5, page 20, lines 810–811: revised sentence limiting the numerical comparisons to descriptive energetic trends.

·       Main manuscript, Section 3.5, page 21, lines 877–879: separate clarification that the component values were not used to infer quantitative thermodynamic binding free energies, receptor activation, or ligand efficacy.

Comment 12:  Page 17, line 737, Figure 5: the title refers to “MM-GBSA binding free-energy estimates”. MM-GBSA does not yield binding free energies; it provides relative interaction-energy scores. A more precise title would avoid implying thermodynamic ΔGbind values.

Thank you for recommending a more precise figure title. The MM-GBSA figure, now numbered Figure 6, is entitled “MM-GBSA relative energetic estimates and component decomposition for the six receptor–ligand complexes.” Its caption identifies the total MM-GBSA score and individual component contributions, without describing them as thermodynamic binding free energies.

Changes and specific locations:

·       Main manuscript, Figure 6 caption, pages 20–21, lines 844–852: revised title and terminology describing relative energetic estimates, total MM-GBSA scores, and component contributions.

Comment 13:  The same remark applies to page 18, lines 748 (Table 5) and the use of the term ΔGbind in the table and on line 750.

We appreciate the request to align the table terminology with the methodological scope. Table 5 now refers to “MM-GBSA relative energetic estimates and component decomposition,” and the total-value column is labeled “Total MM-GBSA score.” The table note explicitly identifies these values as relative energetic descriptors and explains that the incomplete treatment of conformational entropy precludes their interpretation as quantitative thermodynamic binding free energies.

Changes and specific locations:

·       Main manuscript, Table 5, page 21, lines 854–862: revised title, total-score column heading, and explanatory note.

Comment 14:  Page 23, line 980: the phrase “total ΔGbind estimates” again implies that MM-GBSA yields binding free energies. As noted earlier, MM-GBSA provides relative interaction-energy scores rather than thermodynamic ΔGbind values. This sentence should be modified.

Thank you for identifying this additional occurrence. Section 4.3 now uses “lower numerical total scores” and presents MM-GBSA component decomposition as complementary energetic context. The paragraph explicitly limits interpretation to model-dependent relative energetic trends and does not describe the numerical differences as quantitative differences in binding affinity.

Changes and specific locations:

·       Main manuscript, Section 4.3, page 25, lines 1079–1088: revised terminology and interpretation of total MM-GBSA scores and component contributions.

Comment 15:  Page 24, lines 1063-1066: the sentence “In addition, we treated MM-GBSA estimates as relative energetic descriptors rather than experimental binding affinities, given their sensitivity to methodological choices and incomplete treatment of conformational entropy [54].” suggests that MM-GBSA data could be interpreted as binding affinities. The sentence should be modified so that this interpretation cannot be made.

We appreciate the reviewer’s attention to this distinction. The statement now reads: “MM-GBSA component scores were therefore treated exclusively as relative endpoint energetic descriptors; their sensitivity to methodological choices and incomplete treatment of conformational entropy preclude interpretation as quantitative thermodynamic binding free energies [61,62].” This wording removes the comparison with experimental binding affinities and directly states the methodological limitation.

Changes and specific locations:

·       Main manuscript, Section 4.5, page 27, lines 1163–1166: rewritten statement defining the interpretation of MM-GBSA scores, supported by the updated methodological references [61,62].

Comments and suggestions

Reply

Comment 1:  However, while reading the manuscript, I sometimes struggled to identify the central question that the authors were trying to answer as the overall objective is not always immediately obvious. I personally think the manuscript would benefit from a clearer statement of the main hypothesis and of what the authors ultimately want the reader to take away from the study.

We thank the reviewer for highlighting the need to make the central question explicit. The Introduction now asks whether the combined structural differences between MaR1 and DHA produce a conserved recognition motif or receptor-dependent interaction patterns. We explicitly state the hypothesis that C7/C14 hydroxylation and altered double-bond position and geometry redistribute polar interactions and energetic contributions in a receptor-dependent manner. The Abstract and Conclusions identify the principal finding: receptor-specific recognition patterns that define testable interaction sites, without establishing a conserved direct C7-OH/C14-OH motif.

Changes and specific locations:

·       Main manuscript, Introduction, page 3, lines 124–138: explicit knowledge gap, hypothesis, comparative objective, and analytical focus.

·       Main manuscript, Abstract, page 1, lines 38–42: concise statement of the principal finding and its experimental relevance.

·       Main manuscript, Conclusions, pages 28–29, lines 1199–1217: synthesis of receptor-specific recognition and priorities for experimental testing.

Comment 2:  The English is generally very good, however, in some parts of the text a simpler wording would improve readability without sacrificing scientific accuracy. For example, expressions such as “The atom-resolved analysis did not support...” found in page 22 could be replaced by more natural alternatives such as “Atom-level analyses did not support...”. I would encourage the authors to try and simplify the writing to make the reading more fluent.

We appreciate these suggestions for improving readability. We revised wording and sentence structure while retaining technical distinctions and supporting citations. The suggested expression has been adopted: Section 4.1 now begins “Atom-level analyses did not support a conserved direct C7-OH/C14-OH recognition motif across the three receptors.” The receptor-specific discussion was also streamlined to present each interaction pattern and its experimental context more directly.

Changes and specific locations:

·       Main manuscript, Section 4.1, page 24, lines 988–998: adoption of “Atom-level analyses” and revised presentation of the principal recognition patterns.

·       Main manuscript, Section 4.2, pages 24–25, lines 1018–1055: simplified receptor-specific discussion with supporting citations retained.

Comment 3:  In addition, the authors repeatedly remind the reader that it is a computational work and that does not demonstrate activation, agonism, binding affinity or biological efficacy. While I completely agree, and it is understandable that the authors want to make clear that the computational results should not be overinterpreted, those statements appear to frequently. A more concise and assertive treatment of these limitations in a strategic point of the text would improve the narrative.

Thank you for pointing out these repeated qualifications. We reduced recurring statements about activation, affinity, and efficacy in the receptor-specific discussion, which now focuses on the observed interaction patterns and their biological context. The detailed interpretative boundaries are consolidated in Section 4.5. Method-specific qualifications remain where necessary to explain pose selection, sampling diagnostics, and MM-GBSA interpretation.

Changes and specific locations:

·       Main manuscript, Section 4.2, pages 24–25, lines 1018–1055: streamlined discussion centered on receptor-specific interactions and relevant experimental evidence.

·       Main manuscript, Section 4.5, pages 26–27, lines 1129–1169: consolidated treatment of pose-selection, sampling, functional-inference, and energetic limitations.

Comment 4:  “Methodology wise, I would encourage the authors to clarify the role of the Emodel parameter. Emodel is discussed throughout the manuscript and is clearly important in the selection and evaluation of docking poses. However, readers who are not familiar with Glide may not understand exactly what Emodel represents or how it should be interpreted. Although the authors explain that Emodel is not a measure of binding affinity, they do not clearly explain what information it actually provides. In its current form, the text assumes the readers know how Glide shows the results.”

“I would therefore recommend briefly explaining in the methodology what Emodel measures, why it was used for pose selection, and how it differs from docking scores or binding-energy estimates.”

We are grateful for this request to make the methodology more accessible. Section 2.4 now defines Emodel as a composite term combining GlideScore, ligand–receptor molecular-mechanics interaction energy, and ligand strain energy [47–49]. It was examined as a secondary within-ligand pose-assessment descriptor. Representative-pose selection instead relied on pocket compatibility and XP GScore ranking. We also distinguish Emodel from the separately calculated Prime MM-GBSA relative energetic estimates.

Changes and specific locations:

·       Main manuscript, Section 2.4, page 5, lines 226–235: definition of Emodel, explanation of its secondary role, and distinction from XP GScore and Prime MM-GBSA.

·       Main manuscript, Table 2 note, page 12, lines 506–510: clarification of representative-pose selection and descriptive reporting of Emodel differences.

Comment 5:  This appears in the discussion of the PPARγ/DHA complex, where the aothors select a pose with a positive Emodel value, although acknowledging that the selection is not ideal and providing the justification. However, the weight that Emodel should be given to these results is not clear.

Thank you for drawing attention to the original PPARγ/DHA result. During revision, we identified an error in the original DHA input structure, which contained two additional carbon atoms. DHA was corrected to its 22-carbon structure, and the corresponding docking and molecular dynamics analyses were updated; the MaR1 datasets were unchanged. The corrected PPARγ/DHA Emodel value is −54.033 kcal·mol−1. Because Glide returned only one XP pose for this complex, Emodel could not discriminate among alternatives. The pose was retained on structural-admissibility grounds, and its Emodel value was not interpreted as evidence of optimality.

Changes and specific locations:

·       Main manuscript, Section 2.2, pages 4–5, lines 189–200: explicit DHA chemical identity, molecular formula, charge, and double-bond assignments.

·       Main manuscript, Table 2, page 11 (title and note, pages 11–12, lines 500–512), and Section 3.1, page 12, lines 520–523: corrected docking values and explanation of the single-pose starting configuration.

·       Main manuscript, Section 4.5, page 26, lines 1135–1138: clarification that Emodel cannot establish pose optimality when alternative poses are unavailable.

Comment 6:  Then, the manuscript should place less emphasis on the Emodel values and more emphasis on the interaction patterns, Calculated binding energies and simulation results.

We appreciate the recommendation to refocus the narrative on the principal findings. Emodel is now presented as a secondary pose-assessment descriptor. The Results emphasize replica-resolved contact occupancies, residue–ligand oxygen distances, conformational sampling, and MM-GBSA relative energetic estimates. Sections 4.2 and 4.3 integrate these findings with relevant structural and experimental evidence, making receptor-specific recognition patterns the central focus of the discussion.

Changes and specific locations:

·       Main manuscript, Section 2.4, page 5, lines 229–235: explicit secondary role of Emodel.

·       Main manuscript, Section 3.3, pages 13–14, lines 572–607: quantitative presentation of receptor-specific contact and distance profiles.

·       Main manuscript, Section 3.5, page 20, lines 802–817: presentation of total MM-GBSA scores and component-level trends.

·       Main manuscript, Sections 4.2–4.3, pages 24–25, lines 1018–1088: discussion centered on interaction patterns, conformational sampling, and relative energetic profiles.