{"id":"381b377e-8f67-4b5d-80ca-d8e2769438c7","arxiv_id":"2506.21792","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Coarse-grained simulations predict that FUS-LC droplet assembly initiates at core-3 (residues 155-190), followed by core-2 and core-1, with formation times inversely correlated to stability, consistent with Ostwald's rule of stages.","lead":"Computer simulations plus AlphaFold predict that a specific C-terminal region of the protein FUS folds into structured beta-sheet early during droplet formation, before two other regions. This ordering follows Ostwald's rule of stages and may explain how liquid droplets mature into disease-linked fibrils.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corrected MFPTs that set the core-3 < core-2 < core-1 order violate Eq. (2): with 42/200 transitions and T=280 ms the maximum possible MLE for core-1 is 1333 ms, yet 1554 ms is reported; core-2 similarly exceeds its bound.","rationale":"The paper has real strengths: the N* sampling of monomer ensembles, the independent observation of core-3-mediated contacts and beta-strand propensity in droplets, and the agreement of coexisting densities with experiments. These provide partial support for early core-3 involvement independent of AlphaFold. However, the paper's central claim is the ternary kinetic order, and the numbers that establish it are the 'corrected' MFPTs from Eq. (2). The reader identified the unvalidated AlphaFold core-3 reference as the weakest assumption, which is a legitimate external-validity concern. I instead found a more immediate, manuscript-internal problem: the published Eq. (2) cannot produce the reported core-1 and core-2 MFPTs from the published NS, NO, and T. This is an arithmetic consistency check rather than a modeling preference, and it directly undermines the 'unambiguous' ordering as written. Because the raw core-2 versus core-3 comparison is not significant, the second step of the order depends entirely on this correction. The concern is potentially fixable by releasing raw FPTs and correcting T or the reported values, so a REJECT verdict would be too harsh; the appropriate disposition remains CONDITIONAL, with the additional condition that Eq. (2) values be reproducible and that the corrected ordering be accompanied by significance estimates. I therefore leave the reader's verdict unchanged. My disagreement is only about where the weakest link lies: the internal censoring correction is more load-bearing than the external validity of the AF3 core-3 models, though both are worth addressing.","tokens_in":20153,"tokens_out":11948,"duration_ms":131481,"concrete_test":"Request the per-trajectory FPT data or recompute from the stated counts: for each core, evaluate Eq. (2) with NS=200 and the published T (240 or 280 ms), and compare against the reported d<tau_c.i>. The maximum attainable MLE for core-1 is (200/42)*280 = 1333 ms and for core-2 is (200/87)*280 = 644 ms; if the reported 1554 and 685 ms cannot be reproduced with any internally consistent T, the corrected MFPT table is invalid. If the authors supply corrected T values or raw trajectories, rerun the pairwise Welch tests on the corrected MFPTs; if core-2 versus core-3 remains not significant, the claimed ternary order is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the Results section on Ostwald's rule of stages, the authors apply Eq. (2), <tau>=(sum tau_k + NE*T)/NO, and report corrected MFPTs of 1554±16 ms for core-1 and 685±10 ms for core-2, then state these 'unambiguously show' the order core-3, core-2, core-1. With NS=200, NO=42 for core-1, NE=158, and T=280 ms as stated in the main text (Methods says 240 ms), the upper bound of Eq. (2) is (NS/NO)*T = 4.76*280 = 1333 ms, so the reported 1554 ms is impossible; with T=240 ms the bound is 1143 ms. For core-2, NO=87 gives bounds of 644 ms (T=280) or 552 ms (T=240), below the reported 685 ms. Thus the corrected numbers that produce the central ordering are not attainable from the stated data and Eq. (2) as written. This matters because the raw FPTs do not separate core-2 from core-3 (Welch p>0.6); the core-3-before-core-2 step rests entirely on the censoring correction. The AlphaFold dependence of the core-3 reference is a genuine external-validity concern, but this internal arithmetic inconsistency is more directly load-bearing: the headline numerical result is not reproducible as stated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses SOP-IDP coarse-grained simulations of FUS-LC and variants to address three related questions: the relative kinetics of formation of three fibril-like cores (core-1, core-2, and a predicted core-3), the thermodynamics and growth mechanism of FUS-LC droplets, and the role of core-3 in initiating assembly. The central claim is that the mean first-passage times to the fibril-like N* states satisfy core-3 < core-2 < core-1, and that these times are inversely correlated with the relative stabilities of the cores, in accord with Ostwald's rule of stages. Multichain simulations are used to argue that droplets form by nucleation and coarsen by an Ostwald-ripening-like mechanism, that chemical potentials in the coexisting phases are nearly equal, and that simulated phase densities agree with experiments. The paper also reports AlphaFold2/3 structural models for core-3 and uses them to identify residues 164–176 as an early interchain contact hub. The manuscript is clearly written in broad strokes and is rich in comparisons to experimental data, but several load-bearing numerical and textual issues need to be resolved before the main ordering claim can be accepted.","tokens_in":20429,"tokens_out":4121,"duration_ms":47833,"significance":"If the predicted kinetic ordering and the inverse stability-kinetics relation hold, the work would be a valuable example of Ostwald's rule of stages in a physiologically relevant IDP and would provide a concrete structural interpretation of the early events in FUS-LC assembly. The paper deserves credit for anchoring the model to experimental fibril structures for core-1 and core-2, for testing predictions against experimental dense-phase concentrations and conformation-sensitive measurements (SAXS, FRET, DEER), and for explicitly attempting to correct censored first-passage data with a maximum-likelihood formula that has been validated on solvable models. The inferred central role of residues 164–176 is supported by existing hydrogel truncation experiments. However, the headline numerical result is currently not reproducible as reported, and the core-3 reference structures that determine the kinetics are explicitly disclaimed by the authors as potentially incorrect.","major_comments":[{"comment":"The corrected mean first-passage times reported in the Results are not attainable from the stated data and Eq. (2). For core-1 the text states NS=200, NO=42, NE=158, and T=280 ms (the Methods states T=240 ms). Since every observed FPT is nonnegative, Eq. (2) is bounded above by (NS·T)/NO, which evaluates to 1333 ms for T=280 ms (or 1143 ms for T=240 ms), yet the paper reports a corrected value of 1554±16 ms. For core-2, the bound is 644 ms for T=280 ms (or 552 ms for T=240 ms), while the reported value is 685±10 ms. The raw first-passage times do not significantly separate core-2 from core-3 (Welch p>0.6), so the core-3-before-core-2 step of the central ordering rests entirely on this censoring correction. The authors must recompute the corrected MFPTs from the raw trajectory data, report the actual estimates with their bounds, and state which T value was used; as written, the headline ordering is supported by numbers that are mathematically impossible under the paper's own formula.","section":"Results, 'Ostwald's rule of stages'; Eq. (2); Methods 'True Mean FPT'"},{"comment":"The paper explicitly states, 'We do not claim that any of the predicted structures of core-3 are correct in reality.' This disclaimer is in direct tension with the load-bearing use of those structures: the first-passage times to the core-3 U-bend and S-bend N* states are defined through structural overlap with AlphaFold-generated reference fibrils, and the statement that core-3 forms before core-2 and core-1 depends on those times. The paper notes that AlphaFold predictions for IDRs and fold-switching proteins have been criticized, and the internal check against experimental structures is only available for core-1 and core-2. The authors should either provide a concrete external validation of the core-3 model (for example, a comparison to future ssNMR or cryo-EM data, or a mutation-based kinetic test of the predicted 155–190 contact pattern) or substantially weaken the kinetic-ordering conclusion, since the predicted MFPTs would not correspond to real FUS-LC assembly if the core-3 reference structures are wrong.","section":"Discussion, 'Core-3 structure prediction'; Results, 'Structural model for core-3'"},{"comment":"This section contains a self-contradictory sentence that undermines the phase-equilibrium claim. After reporting a relative chemical potential difference of 0.003±0.03 and a Jensen-Shannon divergence of 0.01, the text states: 'Given that the system size that can be reliably simulated is small (maximum of 200 chains), the conclusion that the chemical potentials between the coexisting phases are similar is unreasonable.' If the word 'unreasonable' is intended, then the subsequent calculation of coexisting densities from the equality of chemical potentials is not justified; if it is a typo for 'reasonable', the sentence should be corrected. Either way, the manuscript must be revised so that the stated conclusion is consistent with its own text.","section":"Results, 'Chemical potentials in the two phases are similar'"}],"minor_comments":[{"comment":"The trajectory duration is given as 280 ms in the Results section but as 240 ms in the Methods; the discrepancy must be reconciled because the censoring correction in Eq. (2) depends directly on T.","section":"Methods, 'Transition to fibril-like monomer N* states' vs Results"},{"comment":"The symbols NS, NE, and NO used in Eq. (2) are not defined when the equation is first invoked; the definitions appear only later in the Methods, and the hat notation d⟨τc.i⟩ is never explicitly defined as an estimate.","section":"Results, 'Ostwald's rule of stages' and Methods, 'True Mean FPT'"},{"comment":"The text contains a typographical error, 'FUC-LC', which should read 'FUS-LC'.","section":"Results, 'Core 3 residues drive FUS-LC condensation'"},{"comment":"The phrase 'polymer chains chains exhibit' contains a duplicated word and should read 'polymer chains exhibit'.","section":"Results, 'Chemical potentials in the two phases are similar'"},{"comment":"The possessive 'Ostwalds's rule' contains an extra apostrophe and should be 'Ostwald's rule'.","section":"Discussion, 'Core-3 structure prediction'"}],"recommendation":"major_revision","confidential_remarks":"The arithmetic inconsistency in the corrected MFPTs is the most serious issue: reported values exceed the maximum allowed by the paper's own Eq. (2). If a recount yields corrected values that no longer separate core-2 from core-3, the central ordering claim will need to be revised or removed. The self-contradictory sentence about the reasonableness of the chemical-potential equality also needs to be fixed. The AlphaFold dependence of core-3 is a genuine external-validity limitation, but it is less immediately decisive than the internal numerical inconsistency."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper's central assertion—that FUS-LC cores form in the order core-3, core-2, core-1—is not supported by the numbers as printed. The corrected MFPTs in the Results and Fig. 1I violate the authors' own Eq. (2). With NS=200, T=280 ms (the main text value; Methods says 240 ms), the maximum value Eq. (2) can produce for core-1 with 42 observed transitions is 1333 ms (1143 ms for T=240), yet they report 1554 ± 16 ms. Core-2 similarly exceeds its upper bound: 200*280/87 = 644 ms, below the reported 685 ms. The raw MFPTs, which the paper says do not significantly distinguish core-2 from core-3 (Welch p > 0.6), become ordered only after this impossible correction. So the headline ordering claim is arithmetically unreproducible as stated.\n\nThat said, the paper is not a throwaway. The multichain simulations of 200 chains give a plausible account of nucleation, Ostwald ripening, and near-equality of chemical potentials, and the computed dense-phase concentration (16.24 mM) agrees well with experiments. The identification of residues 164–176 as the dominant early interchain contact is a concrete, testable prediction. The authors are also honest about the AlphaFold provenance of core-3 structures, with an explicit disclaimer.\n\nThe arithmetic inconsistency is load-bearing and must be fixed. If the corrected MFPTs cannot be reproduced from the stated data, the ordering claim collapses. The self-contradictory sentence in the chemical-potential section—calling the near-equality conclusion 'unreasonable' right after reporting JSD = 0.01 and Δμ = 0.003—needs to be resolved. The AlphaFold dependence is a secondary but real limitation, though the explicit caveat helps. The accelerated simulation conditions (100x reduced viscosity, T = 209 K for multichain) could affect the temporal order of events; the higher-viscosity check only uses 64 chains for 150 μs, which does not test the MFPT ordering.\n\nThis paper is for the condensate and IDP community. It deserves a serious referee because the question is important and the simulations produce testable outputs, but it needs major revision before the kinetic ordering can be accepted. I would send it to review with an explicit instruction to the authors to verify Eq. (2), release the raw FPT data and code, and resolve the chemical-potential contradiction.","headline":"The headline core-ordering claim is invalidated by an internal arithmetic contradiction in the corrected MFPTs, though the multichain simulations and testable core-3 contacts are worth engaging.","tokens_in":21155,"tokens_out":3030,"would_cite":false,"duration_ms":32315,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["87.15.hm","64.75.Gh"],"model":"deepseek-v4-flash","headline":"Simulations predict FUS-LC fibril-like cores form least-stable-first, following Ostwald's rule, with the same C-terminal core initiating droplet formation.","keywords":["FUS-LC","phase separation","Ostwald's rule of stages","fibril formation kinetics","intrinsically disordered proteins","coarse-grained simulations","N* states","AlphaFold structure prediction"],"falsifier":"Residue-resolved kinetics of FUS-LC maturation would settle it: time-resolved solid-state NMR or single-molecule FRET pairs reporting on residues 39-95 (core-1) versus 155-190 (core-3) should detect ordered $\\beta$-structure at core-3 within hours and at core-1 only after days, and if the first ordered region to appear is core-1 instead, the hierarchy is wrong. Independently, an experimental structure of the core-3 fibril (e.g., from FUS 141-214, which forms core-3 fibrils in vitro) that places $\\beta$-strands differently than the AlphaFold models would invalidate the core-3 reference and the mean first-passage times measured against it.","tokens_in":19929,"feed_emoji":"🧬","tokens_out":18490,"duration_ms":151753,"temperature":0.7,"pith_summary":"This paper predicts the order in which fibril-like structures appear in FUS-LC, the 214-residue low-complexity domain of the RNA-binding protein Fused in Sarcoma, which phase-separates into droplets and slowly matures into fibers. Using coarse-grained simulations and rare fibril-shaped N* conformations of the isolated monomer as reference states, the authors find that the least stable core, core-3 (residues 155-190), forms first, followed by core-2 (residues 112-150), and finally the most stable core-1 (residues 39-95). Because formation speed runs opposite to stability, the result is an instance of Ostwald's rule of stages, and the same simulations show that core-3 residues 164-176 make the earliest inter-chain contacts, initiating droplet formation, which then coarsens by Ostwald ripening. If the prediction holds, it explains why the core-2 fibril seen in truncated FUS-LC never appears in the full-length domain, and it tells experimenters which region to probe in the earliest stages of assembly.","feed_headline":"FUS-LC fibrils assemble weakest core first, simulations show","feed_subtitle":"The least stable core seeds FUS droplets; the most stable forms last, matching Ostwald's rule.","key_machinery":"The argument is carried by three pieces of machinery. The first is the N* state framework: long equilibrium simulations of the isolated monomer enumerate rare, fibril-like excited conformations, identified by a structural overlap parameter $\\chi$ that compares each sampled conformation against a reference fibril structure, namely the experimental solid-state NMR structure for core-1 (PDB 5W3N), the cryo-EM structure for core-2 (PDB 6XFM), and AlphaFold2/3-predicted models for core-3. The second is first-passage analysis: Brownian dynamics trajectories record the first time each monomer reaches an N* state from the random-coil ground state, and because most trajectories never observe that transition within the simulation window, a maximum-likelihood correction that assumes exponential waiting times converts the truncated observed distributions into true mean first-passage times. The third is multichain simulation of 200 chains, which yields nucleation-like droplet formation, Ostwald-ripening coarsening, and phase equilibrium verified by near-equal chemical potentials in the two phases; quenching the droplet to its inherent structure shows $\\beta$-strand content concentrated in the cores, with the highest propensity in core-3 before core-1 forms. The identity that carries the Ostwald claim is a simple inversion: the mean first-passage times satisfy $\\langle\\tau_{c.1}\\rangle > \\langle\\tau_{c.2}\\rangle > \\langle\\tau_{c.3}\\rangle$ while the stabilities run in the opposite direction, so the fastest-forming core (core-3) is the least stable and the slowest-forming core (core-1) is the most stable.","core_discovery":"On the paper's own terms, the central discovery is a kinetic hierarchy among the three fibril-forming cores of FUS-LC. Fibril-like N* conformations of core-3 (residues 155-190) are reached first, with corrected mean first-passage times of $\\langle\\tau\\rangle \\approx 405$ ms (S-bend) and 426 ms (U-bend), followed by core-2 (residues 112-150) at about 685 ms and core-1 (residues 39-95) last at about 1554 ms. The stabilities run in the opposite direction, with core-1 the most stable and core-3 the least ($\\Delta F_{1,3S} = -2.70 \\pm 0.10$ kcal/mol), so formation time and stability are inversely correlated, as Ostwald's rule of stages requires. Multichain SOP-IDP simulations of 200 chains support the thermodynamic side of the picture: chemical potentials in the dense and dilute phases are equal within noise ($\\Delta\\mu/\\langle\\mu_c\\rangle = 0.003 \\pm 0.03$), coexisting concentrations of about 16 mM and 0.28 mM match experiments, and the earliest, most persistent inter-chain contacts involve residues 164-176, the segment with the highest monomer $\\beta$-strand propensity. The paper concludes that core-3 initiates FUS-LC assembly and that the order of fibril-like structure formation is core-3, then core-2, then core-1.","pith_inferences":["A decisive test the paper does not run: residue-resolved kinetics (time-resolved solid-state NMR, or FRET pairs reporting on residues 39-95 versus 155-190) should see $\\beta$-rich contacts at core-3 within hours and at core-1 only after days; if the first ordered region is core-1 instead, the hierarchy is wrong.","Because the core-3 reference structures are predictions, an experimental structure of the core-3 fibril (e.g., from FUS 141-214, which forms core-3 fibrils in vitro) would either confirm or overturn the computed core-3 formation time, making it the single most decisive check.","The deletion results suggest a graded control the paper does not map: variants that stabilize core-3 (for example ALS-linked mutations near residues 164-176) should further delay core-1 formation, while destabilizing core-3 should accelerate it.","Applied more broadly, the N*-state plus corrected-first-passage protocol could rank nucleation-prone segments in other low-complexity domains (TDP-43, hnRNPA1) and predict which segment seeds their fibers; that generalization is the paper's stated ambition but not yet demonstrated."],"forward_implications":["The kinetic ordering is a falsifiable prediction: in full-length FUS-LC, ordered $\\beta$-structure should first appear at residues 155-190 and only later at residues 39-95, en route to the mature cross-$\\beta$ fibril.","Because core-1 and core-2 are nearly tied in stability yet core-1 forms much more slowly, the core-2 fibril found in the truncated FUS-LC-C construct is kinetically preempted in the full-length protein.","Removing core-3 speeds up core-1 and core-2 formation by roughly 3-fold and 11-fold, making the C-terminal core a kinetic gate; perturbations in residues 155-190 should shift the onset of full FUS-LC fiber formation.","The simulated coexisting concentrations (about 16 mM dense, 0.28 mM dilute) and the near-equality of chemical potentials put the model's phase equilibrium on the same footing as the measured values.","The same N*-state plus corrected-first-passage protocol should transfer to other low-complexity domains such as TDP-43, where the paper expects the same Ostwald-type staging to appear."],"supporting_citations":[{"why":"Supplies the experimental solid-state NMR structure of core-1 (PDB 5W3N), the reference for core-1 N* states, and the 5-7 day maturation timescale.","marker":"[20]"},{"why":"Experimental demonstration that C-terminal residues within core-3 control hydrogel binding and that core-1 and core-3 are mutually exclusive; the main experimental anchor for core-3's role.","marker":"[21]"},{"why":"Supplies the cryo-EM structure of core-2 (PDB 6XFM) and the solubility-based stability comparison used to benchmark the core-1/core-2 free energy difference.","marker":"[23]"},{"why":"Establishes that fibril-like N* states are populated in the FUS monomer ensemble, the premise for the first-passage-time analysis.","marker":"[24]"},{"why":"Parameterization of the SOP-IDP coarse-grained model used in all monomer and multichain simulations.","marker":"[26]"},{"why":"The N* concept itself: propensity to form amyloid fibrils is encoded as rare excitations in the monomer free energy landscape.","marker":"[30]"},{"why":"AlphaFold2 predictions, one of the two sources of core-3 reference structures used to define core-3 N* states.","marker":"[31]"},{"why":"AlphaFold3 predictions, whose core-3 fibril models (U-bend and S-bend) set the reference structures for the core-3 first-passage times.","marker":"[32]"},{"why":"Precedent that monomer energy landscapes of A-beta peptides are sculpted in accord with Ostwald's rule of stages, the rule the paper applies to FUS-LC.","marker":"[60]"}],"fun_headline_variants":["Weakest FUS-LC core forms first, seeds fibril assembly","Ostwald's rule: FUS-LC assembles least stable core first","FUS-LC fibrils build weakest core first, then stable ones","Core-3 seeds FUS-LC fibrils; stable core forms last","Inverse stability order: FUS-LC forms weak core early"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the AlphaFold-predicted structures used as the reference for core-3 are accurate enough proxies for the real core-3 fibril; the paper explicitly says it does not claim those structures are correct, and if they are wrong, the computed formation time for core-3, and the predicted core-3 then core-2 then core-1 ordering built on it, would not describe real FUS-LC.","fun_headline_variants_meta":{"raw":{"variants":["Weakest FUS-LC core forms first, seeds fibril assembly","Ostwald's rule: FUS-LC assembles least stable core first","FUS-LC fibrils build weakest core first, then stable ones","Core-3 seeds FUS-LC fibrils; stable core forms last","Inverse stability order: FUS-LC forms weak core early"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000269,"raw_usage":{"total_tokens":1742,"prompt_tokens":1188,"completion_tokens":554,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":804,"completion_tokens_details":{"reasoning_tokens":456}},"tokens_in":804,"tokens_out":554,"duration_ms":5362,"temperature":1.0,"reasoning_tokens":456,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:18:59.938679+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Residue-resolved kinetics of FUS-LC maturation would settle it: time-resolved solid-state NMR or single-molecule FRET pairs reporting on residues 39-95 (core-1) versus 155-190 (core-3) should detect ordered $\\beta$-structure at core-3 within hours and at core-1 only after days, and if the first ordered region to appear is core-1 instead, the hierarchy is wrong. Independently, an experimental structure of the core-3 fibril (e.g., from FUS 141-214, which forms core-3 fibrils in vitro) that places $\\beta$-strands differently than the AlphaFold models would invalidate the core-3 reference and the mean first-passage times measured against it.","supporting_citations":[{"cited_title":"Structure of fus protein fibrils and its relevance to self-assembly and phase separation of low-complexity domains","cited_arxiv_id":null,"evidence_quote":"Supplies the experimental solid-state NMR structure of core-1 (PDB 5W3N), the reference for core-1 N* states, and the 5-7 day maturation timescale."},{"cited_title":"The low-complexity domain of the fus rna binding protein self-assembles via the mutually exclusive use of two distinct cross- β cores","cited_arxiv_id":null,"evidence_quote":"Experimental demonstration that C-terminal residues within core-3 control hydrogel binding and that core-1 and core-3 are mutually exclusive; the main experimental anchor for core-3's role."},{"cited_title":"Molecular structure and interactions within amyloid-like fibrils formed by a low- complexity protein sequence from fus","cited_arxiv_id":null,"evidence_quote":"Supplies the cryo-EM structure of core-2 (PDB 6XFM) and the solubility-based stability comparison used to benchmark the core-1/core-2 free energy difference."},{"cited_title":"Sequence determines the switch in the fibril forming regions in the low-complexity fus protein and its variants","cited_arxiv_id":null,"evidence_quote":"Establishes that fibril-like N* states are populated in the FUS monomer ensemble, the premise for the first-passage-time analysis."},{"cited_title":"Sequence effects on size, shape, and structural heterogeneity in intrinsically disordered proteins","cited_arxiv_id":null,"evidence_quote":"Parameterization of the SOP-IDP coarse-grained model used in all monomer and multichain simulations."},{"cited_title":"Propensity to form amyloid fibrils is encoded as excitations in the free energy landscape of monomeric proteins","cited_arxiv_id":null,"evidence_quote":"The N* concept itself: propensity to form amyloid fibrils is encoded as rare excitations in the monomer free energy landscape."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"AlphaFold2 predictions, one of the two sources of core-3 reference structures used to define core-3 N* states."},{"cited_title":"Ballard, Joshua Bambrick, Sebastian W","cited_arxiv_id":null,"evidence_quote":"AlphaFold3 predictions, whose core-3 fibril models (U-bend and S-bend) set the reference structures for the core-3 first-passage times."},{"cited_title":"Energy landscapes of a β monomers are sculpted in accordance with ostwald’s rule of stages","cited_arxiv_id":null,"evidence_quote":"Precedent that monomer energy landscapes of A-beta peptides are sculpted in accord with Ostwald's rule of stages, the rule the paper applies to FUS-LC."}],"review_version":1}