{"id":"b7340ecd-f917-43a8-9b18-b0148c60419d","arxiv_id":"2411.16423","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Adding non-interacting peptides with strong affinity to saturated lipids can drive formation of large ordered (Lo/gel) domains in a lattice model of DPPC/DOPC/cholesterol membranes, including metastable raft-sized domains near the critical point.","lead":"This paper uses computer simulations to study what happens when small proteins are added to a model membrane made of three kinds of lipids and cholesterol. It finds that, if the proteins are strongly attracted to saturated lipids, they can pull those lipids together into large ordered patches, which may help explain how lipid rafts form in real cells.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (2) order-parameter classifier is unvalidated for protein sites; all Lo/gel claims, phase partition, and raft-size conclusions are read through this ad hoc lens, so a misclassification would collapse the central claims.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the order-parameter classifier in Eq. (2) is the lens through which every Lo/gel statement is made, and its protein-specific modification (protein score=0, Wi reweighting) is not validated. The paper's previous model was validated for lipid-only mixtures (refs 58, 59), but the protein extension changes the classifier in an ad hoc way, and no independent check—such as a composition-based order parameter or local packing analysis—is provided. All central claims, including protein partitioning, induced Lo/gel separation, and raft-sized metastable domains, depend on this classification. Using the Gi histogram as evidence of Lo phase is circular because the same Gi defines the phase. The authors may well be correct; the model is simple, parameter-lean (one new parameter ε25), and built on prior work. But the current evidence does not exclude a classification artifact, and the concrete test we propose would settle the question. The reader's additional concern about the qualitative nature of the metastability and raft-size claims also supports the conditional verdict, but the order-parameter issue is more fundamental and is the one we stress-test here. Therefore no change to the reader's conditional verdict is needed.","tokens_in":14968,"tokens_out":9240,"duration_ms":92656,"concrete_test":"Re-analyze the stored trajectories (or rerun the code) and identify ordered domains with an independent physical criterion that does not use the scores of Eq. (2), e.g., connected clusters of sites in which at least 3 of 6 neighbors are ordered-DPPC or Chol, or a local excess of ordered-DPPC over disordered-DPPC. Then compare the large protein-rich domain in the 12DPPC ε25=1.95ε and 18DPPC ε25≥1.3ε cases: does it remain a single large ordered domain with similar size and composition? Also vary the score set (e.g., protein=+1, or threshold Gi≥0.5) and require the phase boundary at ε25≈ε22 to be unchanged. If the domain disappears or is no longer Lo/gel under an independent order parameter, the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Every quantitative conclusion—protein partition fraction φ (Figs. 2d, 3e), domain compositions (Table II), the Lo/gel histograms (Figs. 5d, 7), and the identification of the large protein-rich domain itself—is produced by the order-parameter algorithm of Eq. (2) with the hand-assigned scores (void=0, ordered DPPC=2, disordered DPPC=-0.5, Chol=1, DOPC=-1, protein=0) and threshold Gi<0 vs Gi≥0. The modification for proteins (Wi=6/(6-N_i^p), protein score=0) is introduced without validation against any independent structural measure. The paper's evidence that the large domain is Lo/gel is the bimodal Gi histogram, but that is circular: the same Gi gives the label. If the protein-containing clusters are actually disordered aggregates with only a few ordered DPPC neighbors, the algorithm would still score them as ordered. Since the central claim that ε25 ≳ ε22 induces a protein-rich Lo/gel phase rests entirely on this classification, a systematic misclassification would invalidate the main results.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends a previously calibrated lattice Monte Carlo model of DPPC/DOPC/Chol mixtures by adding non-interacting triangular peptide trimers. It investigates how the peptide affinity to ordered DPPC chains (parameter ε25) controls protein partitioning between liquid-disordered and liquid-ordered regions, and whether proteins can induce the formation of large ordered domains in one-phase and near-critical mixtures. In the two-phase (35DPPC) case, the authors report a sharp crossover in protein partitioning as ε25 increases. In the one-phase (12DPPC and 18DPPC) cases, they observe that sufficiently strong ε25 leads to a single large protein-rich domain that is classified as Lo or gel, and they discuss the domain size relative to biological rafts. The main claims are that proteins with sufficiently strong attraction to saturated lipids can drive phase separation and that near the critical composition they produce metastable dynamic ordered domains of raft-comparable size.","tokens_in":15236,"tokens_out":7201,"duration_ms":69067,"significance":"If the findings are correct, the paper offers a useful minimal demonstration that small, non-interacting proteins can act as nucleation centers and shift the phase behavior of model membranes, which is directly relevant to the lipid-raft debate. The simulations are extensive and well described, including multiple compositions, protein numbers, time-resolved movies, and a clear presentation of protein partitioning. The paper builds on a previously calibrated model and explicitly lists its parameters, which is commendable. However, the central quantitative conclusions rest on a heuristic order parameter that has not been independently validated for protein-containing systems, and one internal inconsistency in the gel definition needs to be resolved.","major_comments":[{"comment":"The stated definition of the gel region is inconsistent with the score values given in the same section. With the scores listed (ordered DPPC = 2, Chol = 1, disordered DPPC = -0.5, DOPC = -1, void = 0, protein = 0) and Wi = 1, Eq. (2) gives a maximum Gi of 4 (an ordered DPPC site surrounded by six ordered DPPC neighbors), and the protein modification Wi = 6/(6 - N_i^p) does not change this upper bound. Since the gel phase is identified as sites with Gi = 14 in Figs. 5(d) and 7, the reported gel fractions are not interpretable unless the score values or the threshold are corrected.","section":"Section II, Eq. (2)"},{"comment":"The modification of the order parameter for proteins (Wi = 6/(6 - N_i^p), protein score 0) is introduced ad hoc and is not validated against any independent structural observable. All phase labels, domain compositions in Table II, and the conclusion that the large protein-rich domain is Lo or gel are derived from this same Gi classifier, which makes the identification partly circular. The authors should verify the classification by reporting a direct measure of chain order within the large domain, such as the fraction of DPPC chains in the ordered state (s = 2), or by using a separate structural estimator such as a bond-orientational order parameter.","section":"Section II and Sections III.B-III.C"},{"comment":"The inference of a protein-driven phase transition is based on the appearance of a single large cluster in finite systems with 100-500 proteins. Although the growth of the cluster with protein number is suggestive, finite-size effects could produce similar clustering without a true thermodynamic transition. The paper does not provide finite-size scaling of the lattice size or a free-energy comparison for the competing states. The language should be tempered (e.g., 'apparent phase separation in the simulated finite system') or the analysis should be supplemented with such tests.","section":"Section III.B, Figs. 4-5"}],"minor_comments":[{"comment":"The word 'distirubtion' should be 'distribution'.","section":"Abstract"},{"comment":"In the text near Fig. 2, the phase 'L0' should be 'Lo' to match the notation used elsewhere.","section":"Section III.A"},{"comment":"The phrases 'energy energy' and 'of of' should be corrected.","section":"Section III.B"},{"comment":"Please add axis labels to the histogram (e.g., 'order parameter Gi' and 'count').","section":"Fig. 5(d)"},{"comment":"Please specify the number of independent runs used to compute the standard deviations reported in Table II and the points in Fig. 2(d).","section":"Section III and Table II"},{"comment":"The parameters a and b are fit values; the text should state explicitly that the dashed curve is a descriptive fit rather than a predictive test, and the conclusion that the crossover occurs when ε25 is approximately ε22 should be presented as an empirical observation from these simulations.","section":"Eq. (4) and Fig. 2(d)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the scope of a soft-matter physics journal and the central idea is interesting. The main problems are the unvalidated order-parameter classifier, the internal inconsistency in the gel definition (Gi = 14 appears impossible with the stated score values), and the finite-size evidence for a phase transition. These are fixable with additional analysis and careful rewording, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this is a clean, simple extension of the authors' own validated lattice model, and it makes a plausible case that non-interacting peptides can nucleate Lo/gel domains when their attraction to ordered DPPC exceeds the DPPC-DPPC attraction. The crossover in protein partitioning around ε25 ≈ ε22 is clearly demonstrated, and the snapshots showing a single large protein-rich ordered domain at 12DPPC with 100, 200, and 500 proteins are visually convincing. The stress-test worry about the order-parameter classifier is legitimate but not fatal: the classification is ad hoc and unvalidated for protein sites, but the large domain is visible by eye, and the Gi histogram is just quantifying what the eye already sees. I do not think a misclassification is collapsing the main story.\n\nWhat is genuinely new is modest but useful: adding trimer peptides to the DPPC/DOPC/Chol lattice model, mapping the partition curve as a function of one new parameter, and showing that proteins can convert a one-phase mixture into a phase-separated one. The systematic sweep and the finite-size check with different protein numbers are good. The paper is also honest about metastability—it calls the dynamic domains metastable and does not overstate the raft analogy.\n\nThe soft spots are real but proportionate. The order-parameter modification for proteins is introduced without any independent structural validation, so a referee should ask for that. More importantly, the claim about raft-sized domains rests on visual inspection of snapshots and movies, not on a quantitative lifetime or size-distribution analysis. The paper says data is available on request but ships no code or data, which hampers reproducibility. The partition curve fit to Eq. (4) is explicitly a fit, not a prediction, which is fine given that it is not load-bearing.\n\nWho should read this? Researchers using lattice or ultra-CG models to understand protein-lipid phase behavior, and experimentalists looking for a simple mechanism for protein-stabilized rafts. It deserves a serious referee. I would send it to review, with the request that the authors provide the order-parameter validation and a quantitative metastability analysis. The central mechanism is likely sound.","headline":"A simple lattice-model study showing non-interacting peptides can nucleate Lo/gel domains when peptide-DPPC attraction beats DPPC-DPPC; the central claim is believable despite an unvalidated order-parameter classifier.","tokens_in":15724,"tokens_out":2059,"would_cite":true,"duration_ms":21837,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["87.16.dj","05.10.Ln"],"model":"deepseek-v4-flash","headline":"Adding a few non-interacting peptides can force lipid membranes into raft-sized ordered domains.","keywords":["lipid rafts","lattice model","Monte Carlo simulation","liquid-ordered domains","protein partitioning","DPPC/DOPC/cholesterol","phase separation","membrane domains"],"falsifier":"Simulate the same mixtures but classify phases by an independent measure, such as the local fraction of ordered DPPC chains or a cluster criterion based on chain tilt, instead of the Eq. (2) score; if no distinct large Lo/gel domain appears under the alternative criterion for ε25 > ε22, the reported phase separation is an artifact of the scoring algorithm. Alternatively, in GUV experiments with short transmembrane peptides of graded hydrophobic length, if increasing peptide affinity to ordered lipids does not enlarge ordered domains beyond the protein-free case, the nucleation mechanism is not operative in real bilayers.","tokens_in":14685,"feed_emoji":"🔬","tokens_out":7879,"duration_ms":66339,"temperature":0.7,"pith_summary":"This paper extends a lattice model of the ternary lipid mixture DPPC/DOPC/cholesterol to include small protein-like peptides, and uses Monte Carlo simulation to ask what the proteins do to the mixture's phase behavior. The authors find that when the peptide's attraction to ordered saturated lipid chains (parameter ε25) exceeds the attraction between saturated lipids themselves (ε22), even a small number of non-interacting peptides can drive the system to separate into a distinct protein-rich liquid-ordered or gel domain. In mixtures near the critical composition, peptides with slightly weaker affinity produce metastable, dynamic ordered domains whose sizes match those of biological lipid rafts (tens of nanometers). The paper argues that protein-lipid affinity alone, without direct protein-protein interactions, can nucleate and stabilize liquid-ordered domains, offering a concrete mechanism for how proteins may control raft size in real membranes.","feed_headline":"Small proteins can force lipid membranes into raft-sized domains","feed_subtitle":"A lattice-model simulation shows attraction to saturated lipids alone drives ordered-domain growth to biological raft scale.","key_machinery":"The central mechanism is the competition between two nearest-neighbor attractions in the lattice Hamiltonian (Eq. 1): the protein–ordered-DPPC interaction ε25 and the DPPC–DPPC interaction ε22 (with ε22 = 1.3ε). When ε25 exceeds ε22, ordered DPPC chains prefer to sit next to a peptide rather than next to each other, so peptides act as nucleation centers that recruit saturated lipids and cholesterol into a dense ordered cluster; because the proteins themselves do not interact (ε55 = 0), the driving force is entirely lipid-mediated. Phase regions are identified by the order parameter Gi of Eq. (2), a site score plus weighted neighbor average, with scores 2 for ordered DPPC, 1 for cholesterol, −0.5 for disordered DPPC, −1 for DOPC, 0 for proteins and voids; negative Gi marks Ld, non-negative marks Lo, and Gi = 14 marks gel. Domain sizes and fluctuations are measured with the Hoshen–Kopelman algorithm and radius of gyration Rg (Eq. 3).","core_discovery":"The central claim is that the phase behavior of a ternary DPPC/DOPC/Chol mixture is not fixed by the lipids alone: adding a small population of non-interacting peptides (2% of the lattice area) with a tunable attraction to ordered DPPC chains qualitatively changes the phase diagram. In the two-phase region (35 mol% DPPC), proteins partition between Ld and Lo according to a sharp Boltzmann crossover centered near ε25 ≈ ε22, with the fraction inside the Lo phase fitted by φ = exp[(aε25 − b)/kBT]/(1 + exp[(aε25 − b)/kBT]) with a ≃ 7.2 and b ≃ 5.8. In the one-phase 12DPPC mixture, increasing ε25 first grows metastable Lo domains around the proteins and then, for ε25 > ε22, coalesces them into a single large stable domain that contains most of the proteins and recruits up to about 90% of the DPPC; depending on protein number, the domain is Lo or converts to gel. In the near-critical 18DPPC mixture, ε25 ≲ ε22 yields metastable dynamic Lo domains with gyration radii in the 16–23 nm range and large size fluctuations, precisely the length scale of biological rafts. The paper concludes that protein-mediated nucleation, driven solely by the competition between ε25 and ε22, is sufficient to create and stabilize raft-sized domains.","pith_inferences":["If the mechanism is generic, experimental systems with peptides engineered to have controlled hydrophobic matching should show raft-size domains that grow with peptide affinity; this is testable in giant unilamellar vesicles or supported bilayers with single-molecule tracking.","The model suggests a route to reconciling Type I and Type II mixtures: the location of the mixture relative to the critical point, combined with protein affinity, sets whether protein addition nucleates a stable phase or only stabilizes dynamic nanodomains; the authors' preliminary Type I data (their Fig. 8) hint at this.","A parameter-free prediction is that the crossover in protein partitioning should shift to higher ε25 at lower DPPC content, since the average number of ordered-DPPC neighbors per protein decreases; this is already visible in comparing Figs. 2 and 3 and could be quantified.","The order-parameter classification could be validated against an independent measure, such as local chain tilt or deuterium order profiles; if discrepancies appear, some of the sites labeled 'gel' may actually be ordinary Lo sites."],"forward_implications":["A small density of non-interacting proteins can induce macroscopic Lo/gel phase separation in a one-phase mixture when ε25 > ε22, so lipid-only phase diagrams may understate membrane heterogeneity in protein-rich environments.","Near the critical composition, protein-lipid affinities slightly below the lipid-lipid value create metastable Lo domains with raft-comparable sizes (Rg ~ 16–23 nm) and large fluctuations, suggesting raft size can be tuned by protein affinity without requiring protein clustering.","The fraction of proteins partitioning into Lo follows a two-state Boltzmann form, so protein enrichment in ordered domains is a sharp switch rather than a gradual response to ε25.","Increasing protein content shifts the separated domain from liquid-ordered toward gel, so local protein density controls whether a raft-like or gel-like domain forms.","The same single-parameter extension reproduces these effects across three compositions, implying that the ε25/ε22 ratio is a control variable for domain formation in the model."],"supporting_citations":[{"why":"Supplies the base lattice model of DPPC/Chol and the calibrated values of Ω, ε22, and ε23 used in the Hamiltonian.","marker":"[58]"},{"why":"Extends the model to DPPC/DOPC/Chol, sets ε24 = 0, and introduces the order-parameter algorithm (Eq. 2) that the paper adapts for proteins.","marker":"[59]"},{"why":"Provides the experimental phase diagram of DPPC/DOPC/Chol that the simulated compositions and two-phase region are compared against.","marker":"[37]"},{"why":"The Hoshen–Kopelman algorithm is used to identify and measure the ordered domains and their radii of gyration.","marker":"[98]"},{"why":"Earlier ultra-coarse-grained and lattice simulations are cited as having demonstrated protein-induced domain formation, the phenomenon the paper reproduces and extends to raft sizes.","marker":"[94]"},{"why":"Experimental NMR study of DPPC/DOPC/Chol phase coexistence that anchors the Type II mixture behavior.","marker":"[24]"}],"fun_headline_variants":["Peptides with lipid affinity nucleate raft-sized lipid domains","Protein attraction to saturated lipids grows raft-scale domains","Small proteins can make lipid domains as big as rafts","Adding peptides to lipid mix triggers raft-sized domain growth"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification of lattice sites into liquid-disordered, liquid-ordered, and gel regions relies on an order-parameter formula (Eq. 2) with hand-assigned scores for each component, and the conclusions about phase separation and raft-sized domains would collapse if this classification misidentifies the true phase of a region.","fun_headline_variants_meta":{"raw":{"variants":["Peptides with lipid affinity nucleate raft-sized lipid domains","Protein attraction to saturated lipids grows raft-scale domains","Small proteins can make lipid domains as big as rafts","Adding peptides to lipid mix triggers raft-sized domain growth"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000286,"raw_usage":{"total_tokens":1771,"prompt_tokens":1126,"completion_tokens":645,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":742,"completion_tokens_details":{"reasoning_tokens":581}},"tokens_in":742,"tokens_out":645,"duration_ms":6568,"temperature":1.0,"reasoning_tokens":581,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:08:13.157674+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the same mixtures but classify phases by an independent measure, such as the local fraction of ordered DPPC chains or a cluster criterion based on chain tilt, instead of the Eq. (2) score; if no distinct large Lo/gel domain appears under the alternative criterion for ε25 > ε22, the reported phase separation is an artifact of the scoring algorithm. Alternatively, in GUV experiments with short transmembrane peptides of graded hydrophobic length, if increasing peptide affinity to ordered lipids does not enlarge ordered domains beyond the protein-free case, the nucleation mechanism is not operative in real bilayers.","supporting_citations":[{"cited_title":"Podewitz , author Y","cited_arxiv_id":null,"evidence_quote":"Supplies the base lattice model of DPPC/Chol and the calibrated values of Ω, ε22, and ε23 used in the Hamiltonian."},{"cited_title":"Sarkar \\ and\\ author O","cited_arxiv_id":null,"evidence_quote":"Extends the model to DPPC/DOPC/Chol, sets ε24 = 0, and introduces the order-parameter algorithm (Eq. 2) that the paper adapts for proteins."},{"cited_title":"Choucair , author M","cited_arxiv_id":null,"evidence_quote":"Provides the experimental phase diagram of DPPC/DOPC/Chol that the simulated compositions and two-phase region are compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The Hoshen–Kopelman algorithm is used to identify and measure the ordered domains and their radii of gyration."},{"cited_title":"Hinderliter , author P","cited_arxiv_id":null,"evidence_quote":"Earlier ultra-coarse-grained and lattice simulations are cited as having demonstrated protein-induced domain formation, the phenomenon the paper reproduces and extends to raft sizes."}],"review_version":1}