{"id":"567f49cd-6009-4565-beb4-3d90eb2c5444","arxiv_id":"1908.07039","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A Hindmarsh-Rose neuron network on the mouse connectome produces persistent chimera-like states, some of which resemble epileptiform EEG activity, in a specific region of coupling-strength space.","lead":"This paper simulates a network of model neurons wired according to the mouse brain's actual connection map and finds patterns of mixed synchronized and desynchronized activity, called chimera states, across a range of connection strengths. These patterns look like the electrical traces seen during seizures, suggesting that the same mathematical mechanism could underlie some types of epilepsy.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The chimera-like index χ is never validated against a spatial definition of chimera states; the low-coupling island could reflect inter-community heterogeneity rather than coexisting synchrony and asynchrony.","rationale":"The reader's weakest assumption identifies the chimera-like index χ as the critical unvalidated detector. I agree: the central claim depends on χ faithfully indicating chimera states across a wide region of parameter space, but the paper provides only two illustrative time series and no direct spatial validation or null baseline. The concrete spatial-order-parameter test would settle whether χ corresponds to actual coexisting coherent and incoherent domains. Since the reader already made the verdict CONDITIONAL with this concern explicitly listed, my analysis reinforces rather than changes that verdict. The paper does have genuine supporting features: the code is publicly available, the model reproduces a prior published result, and the qualitative EEG-like traces are shown. The concern is about the strength of inference from the index to the claimed large chimera region, not about internal inconsistency. I therefore recommend no change to the conditional assessment.","tokens_in":12558,"tokens_out":4874,"duration_ms":52083,"concrete_test":"Select the highest-χ points in the low-coupling island (Fig. 8) and in the boundary band, and for each compute the per-neuron local order parameter R_j(t) = |(1/|N_j|) Σ_{k∈N_j} exp(iφ_k(t))| over the same saved interval, then map the time-averaged R_j across the 213 subcortices. A genuine chimera state requires spatially contiguous regions with high R_j and low R_j coexisting in the same network at the same time. If high-χ points do not exhibit such coexisting domains, χ is not a valid detector and the 'large region' claim collapses. Equivalently, a surrogate control with randomized edges (preserving degree and community strengths) would show whether the island is connectome-specific.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the Hindmarsh-Rose network 'produces ... persistent chimera states in a large region' of the (α,β) plane. The only quantitative detector used is the chimera-like index χ (Eq. 2), defined as the time-averaged variance of the order parameter r_c(t) across the 13 coarse cortical areas. By construction, χ is large whenever communities have persistently different mean levels of synchrony, irrespective of whether any community actually contains coexisting synchronized and desynchronized domains. The 13 coarse areas in the mesoscale mouse connectome differ substantially in size, degree, and connection strength (Fig. 4), so static topological heterogeneity alone could generate a high χ without any chimera state. The paper does not calibrate χ against a direct spatial measure (e.g., local order parameter fields or contiguous coherent/incoherent regions), and it does not include a null or surrogate-network control. The only support is two qualitative time-series figures (Figs. 6 and 9). In particular, the low-coupling island in Fig. 8 is presented as a 'surprising' feature, but with no error bars or null baseline it could be an artefact of the measure rather than a genuine dynamical regime. This is the weakest link in the chain from simulation output to the headline claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a modified Hindmarsh-Rose neural-mass network whose coupling graph is the mesoscale mouse connectome (213 fine areas grouped into 13 coarse areas). The authors sweep two coupling parameters, α (intra-cortical) and β (inter-cortical), and measure the chimera-like index χ and metastability index m defined by Shanahan (2010). They report that the physical region of parameter space contains persistent chimera-like states, with high χ in a low-coupling island near (α,β) = (0.04,0.02) and along a band near the aphysical boundary, and they argue that some simulated traces qualitatively resemble epileptiform EEG. The central claim is that this network 'produces superficially epileptiform activity converging on persistent chimera states in a large region' of the (α,β) plane.","tokens_in":12803,"tokens_out":3737,"duration_ms":39210,"significance":"If the central claim is supported, the paper would be a useful contribution to the literature connecting chimera states to seizure dynamics, extending prior cat-brain work to a modern mouse connectome. Strengths include the clear description of the numerical implementation, the availability of the simulation code, the explicit parameter sweeps in Table 2, and the verification against the Santos et al. (2017) model. The paper is also honest about its qualitative EEG comparison and about treating connection strengths as constant. However, the main result depends entirely on the chimera-like index χ, which is not validated against a direct spatial definition of chimera states, and no null-network baseline is provided; these gaps are load-bearing for the headline claim.","major_comments":[{"comment":"The chimera-like index χ is the time-averaged variance of community order parameters r_c(t). As the authors themselves show in Fig. 4, the 13 coarse areas differ substantially in size, degree, and mean connection strength, so static topological heterogeneity alone can produce persistently different mean synchrony levels across communities and hence a large χ without any community actually containing coexisting synchronized and desynchronized domains. The manuscript does not calibrate χ against a direct spatial measure of chimera states (e.g., local order-parameter fields or contiguous coherent/incoherent regions), and it does not include a null-network or surrogate control. Because χ is the only quantitative detector used for the claim of 'persistent chimera states in a large region' in the abstract, this is a load-bearing gap that should be addressed before the central claim can be accepted.","section":"Eq. (2) and Section 3.3"},{"comment":"The low-coupling island is presented as a 'surprising' feature and the authors state that it is not a calculation error, but the evidence is only visual. The high-resolution sweep (α,β)∈(0,0.2)×(0,0.2) is not repeated in Table 2; only the [0,1]×[0,1] sweep is listed as having 10 runs. Fig. 8B shows a 'variance of the chimera-like index' but no error bars or confidence intervals are given, and the quantity being averaged over is not clearly stated (variance across runs, across time, or across communities). Without repeated runs or a null baseline, the island could be an artifact of the measure or of the particular initial conditions; the authors should provide error bars or repeated-run statistics for Fig. 8, as they did for the larger sweep in Fig. 10.","section":"Fig. 8 and Table 2"},{"comment":"The aphysical region is excluded from the analysis based on the criterion that neurons do not fire, and the slope of the boundary is explained using average connection strengths in Fig. 4. However, the threshold K_j in Eq. (13) is not estimated or tested, and the argument for the boundary slope remains qualitative. Since the definition of the physical region determines which parts of parameter space are available for the chimera claim, the authors should provide a quantitative test of Eq. (13) or at least a sensitivity analysis of the firing criterion (e.g., varying the threshold 'increased past 1' or the definition of a firing event). This would also help clarify whether the high-χ band near the boundary is a dynamical chimera regime or an artifact of proximity to the aphysical region.","section":"Section 4.2 and Eq. (13)"}],"minor_comments":[{"comment":"The reference [Ljungberg et al.(2009)] contains a corrupted author string '/quotesingle.ts1Arcangelo'; please fix it.","section":"References"},{"comment":"'crebellar cortex' should be 'cerebellar cortex'.","section":"Fig. 2 caption"},{"comment":"The term 'aphysical' is used consistently; if this is a deliberate term, please define it at first use, otherwise replace with 'unphysical' or 'non-physical'.","section":"Throughout"},{"comment":"The sentence 'The highly chimeric portion of the landscape appears to be mostly below the β=α line' appears in Section 4.1 before Fig. 7 is presented; consider moving it to Section 4.3 where the chimera landscape is discussed.","section":"Section 4.1 vs 4.3"},{"comment":"The metastability index m is computed and displayed in Fig. 10C, but it is not discussed in the main text; a brief interpretation would help the reader understand what m adds to the chimera-like index results.","section":"Fig. 10C"},{"comment":"The code availability is mentioned only in a footnote; if the journal has a data-availability policy, consider adding a formal statement in the main text.","section":"Section 3.3"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Henry,\n\nYou'll want to know this paper before you cite it. It takes the Hindmarsh-Rose network model from Santos et al. (2017) and runs it on the mesoscale mouse connectome (Oh et al. 2014), computing the Shanahan chimera-like index χ and metastability m over a two-parameter (α,β) sweep. The genuinely new result is a low-coupling island (α around 0.04, β around 0.02) with high χ, plus a linear aphysical boundary. The authors show the island is not a single-run fluke (Fig. 9), and the code is on GitHub. That part is solid and worth a look.\n\nThe soft spots are real. First, the abstract says 'large region' but the high-χ region is actually a thin band near the aphysical boundary plus a small island in the corner. That overstates the finding. Second, χ as defined in Eq. 2 is the time-averaged variance of order parameters across the 13 coarse areas. It will be high whenever communities have persistently different mean synchrony, which can arise purely from topological heterogeneity (different internal connection densities) without any genuine chimera dynamics. The authors don't include a surrogate-network or null baseline, and they don't validate χ against a spatial measure of coexisting coherent/incoherent domains beyond two illustrative time series. This is the load-bearing gap. Third, the error bars on the main landscape (Fig. 7) are absent; they show a variance image only for the small island (Fig. 8B and Fig. 10B), and the number of runs for the other sweeps is unclear (only the [0,1]×[0,1] sweep was repeated 10 times). The exclusion of the aphysical region is handled reasonably, but the boundary slope analysis is informal.\n\nI think the stress-test note is half right. The concern that χ can be inflated by static inter-community differences is valid and needs a control. But the claim that χ doesn't detect coexisting synchrony and asynchrony is a bit too strong: Shanahan's index is explicitly a community-level chimera detector, and the 13 coarse areas are the communities. Still, without a null network, the low-coupling island could be an artefact of the measure on this particular connectome.\n\nIf I were refereeing this, I'd ask for three things: a degree-preserving randomized network control, error bars or at least a statement of runs per parameter point on the main sweep, and a direct spatial check (e.g., local order parameter maps) at representative points in the island and the boundary band. The paper is worth engaging with; the model is cheap, the code is available, and the island is a potentially interesting feature. It deserves peer review, though the central claim needs to be toned down and the evidence sharpened.\n\nReading group: maybe.","headline":"A well-executed application of a standard chimera index to a new connectome, but the headline 'large region' claim needs a null baseline and better statistical support before it can be taken at face value.","tokens_in":13348,"tokens_out":3057,"would_cite":false,"duration_ms":31965,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C20","37N25","34C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A modified Hindmarsh-Rose network on the mouse connectome produces persistent chimera states—coexisting synchrony and asynchrony—across a large region of coupling space, matching seizure-like EEG activity.","keywords":["chimera states","Hindmarsh-Rose model","mouse connectome","neural mass model","epileptiform activity","synchrony","metastability","seizure dynamics"],"falsifier":"For a parameter pair inside the low-coupling island (e.g., $\\alpha\\approx 0.04$, $\\beta\\approx 0.02$), record the instantaneous phases of all 213 subregions and map them spatially; the claim is falsified if no coarse area simultaneously contains a phase-locked cluster and a phase-scattered cluster, or if shuffling the intra-areal connections leaves $\\chi$ just as high.","tokens_in":12350,"feed_emoji":"🧠","tokens_out":9156,"duration_ms":79157,"temperature":0.7,"pith_summary":"This paper tries to show that a network of modified Hindmarsh-Rose neurons, wired according to the mesoscale mouse connectome, spontaneously generates persistent chimera states—where some brain regions fire in synchrony while others remain desynchronized—and that the resulting activity looks superficially like epileptic seizures on an EEG. The claim matters because chimera states are a leading mathematical framework for epilepsy, and if a connectome-realistic model produces them without being designed to do so, it strengthens the case that chimeras are a generic, perhaps universal, feature of brain-network dynamics. The paper maps a two-parameter space of coupling strengths—$\\alpha$ within cortices and $\\beta$ between cortices—and reports high chimera-like index values in a low-coupling island ($\\alpha$ below about 0.1, $\\beta$ below about 0.05) and along a band near the boundary of the model's unphysical region. These states persist for simulation times up to 4000 time units, and their EEG-like traces show seizure-like spike runs in regions such as the medulla, hypothalamus, and cortical subplate.","feed_headline":"Chimera states persist in a mouse-brain model's weak-coupling zone","feed_subtitle":"A connectome-based model yields seizure-like EEG traces via persistent chimera states, supporting models of epilepsy.","key_machinery":"The central machinery is the modified Hindmarsh-Rose neural-mass model (Eqs. 6–10): each of 213 masses obeys a 3-variable spiking equation with a sigmoidal activation function, coupled twice—once within cortices with strength $\\alpha$ through matrix $G'$ and once between cortices with strength $\\beta$ through matrix $G''$. The two coupling matrices come from the mouse connectome, with connection strengths binned into four levels (Eq. 11). The detectors are the chimera-like index $\\chi$ (time-averaged variance of the order parameter across the 13 coarse areas, Eq. 2) and the metastability index $m$ (average within-area variance of the order parameter over time, Eq. 3), following Shanahan. The model is integrated with a 4th-order Runge-Kutta scheme for 5000 time units, discarding the first 1000 as transients, and phases are extracted from spike times via Eq. 12. The two-parameter sweep in $\\alpha$ and $\\beta$, with the physical–aphysical boundary delimiting the valid region, is what turns the connectome into a phase diagram for chimeric seizure-like activity.","core_discovery":"On the paper's own terms, the core discovery is that a network of 213 Hindmarsh-Rose neural masses, grouped into 13 coarse cortical areas according to the mouse connectome, exhibits persistent chimera states in a large region of the $(\\alpha,\\beta)$ coupling plane rather than only in a narrow parameter sliver. The chimera-like index $\\chi$, computed as the time-averaged variance of the order parameter across the 13 areas, reaches its highest values in a surprising low-coupling patch at $\\alpha\\lesssim 0.1$ and $\\beta\\lesssim 0.05$, and again along a roughly diagonal band near the boundary where the model becomes aphysical (some neurons stop firing). The paper argues this is not a calculation artifact: the patch persists across repeated runs, has low run-to-run variance, and matches a qualitative phase portrait showing synchronized domains in the medulla, hypothalamus, and isocortex alongside asynchronous areas. It also finds that $\\alpha$ has a slightly larger influence than $\\beta$ on the physical–aphysical boundary, consistent with mean intra-cortical connection strengths exceeding mean inter-cortical strengths once zero-connection areas are set aside.","pith_inferences":["A direct spatial test not reported in the paper: map instantaneous phases of all 213 subregions inside the low-coupling island; the chimera interpretation requires a synchronized cluster and a desynchronized cluster to coexist within the same coarse area at the same time.","Because $\\chi$ is computed across only 13 coarse areas, it may conflate true chimeras with plain inter-area differences in firing rate; recomputing the index on sliding spatial windows or on the 213 fine areas would separate those alternatives.","The model's assumption of identical nodes is a natural extension point: adding heterogeneity in the input current $I_j$, which the paper notes disrupts chimeras, should shrink or shift the low-coupling island; mapping that shift would give an experimentally controllable lever on seizure-like dynamics."],"forward_implications":["If the model's persistent chimera states are real, then seizure-like EEG traces can emerge from the mouse connectome without any seizure-specific parameter tuning, supporting the idea that chimeras are a generic brain-network phenomenon.","The low-coupling island ($\\alpha\\lesssim 0.1$, $\\beta\\lesssim 0.05$) and the band along the physical–aphysical boundary become candidate regions for experimental or clinical searches for seizure precursors.","The sharp drop in $\\chi$ near $\\alpha\\approx 0.1$, where coupling becomes comparable to intrinsic dynamics, implies a regime threshold: below it, weak intra-areal coupling suffices to sustain chimera dynamics; above it, synchrony fades.","Because the hypothalamus and hippocampal formation show the strongest synchrony, the model singles out these areas as likely seizure foci, matching clinical observations and giving a concrete prediction for future connectome-based seizure models.","The physical–aphysical boundary's slope near $-1$ in $(\\alpha,\\beta)$ space, explained by mean intra- vs inter-cortical connection strengths, provides a simple topological criterion for where the model ceases to produce meaningful dynamics."],"supporting_citations":[{"why":"Supplies the modified Hindmarsh-Rose model and the $\\alpha$–$\\beta$ parameter-sweep methodology this paper extends to the mouse connectome.","marker":"[Santos et al.(2017)]"},{"why":"Provides the mesoscale mouse connectome, the 213 fine areas, 13 coarse areas, and connection strengths that define $G'$ and $G''$.","marker":"[Oh et al.(2014)]"},{"why":"Defines the chimera-like index $\\chi$ and metastability index $m$ used to detect chimeric states.","marker":"[Shanahan(2010)]"},{"why":"Establishes chimera-like states in modular neural networks and grounds the claim that small-world topology facilitates nonlocal coupling.","marker":"[Hizanidis et al.(2016)]"},{"why":"Supports the persistence of chimeras over long simulation times despite their transient nature.","marker":"[Wolfrum & Omel'chenko(2011)]"},{"why":"Provides the foundational two-population chimera model that motivates the intra- vs inter-community coupling distinction.","marker":"[Abrams & Strogatz(2004)]"},{"why":"Draws the analogy between chimera state collapse and epileptic seizures that frames the paper's physiological interpretation.","marker":"[Andrzejak et al.(2016)]"}],"fun_headline_variants":["Chimera states persist in weak-coupling zone of mouse brain","Seizure-like chimeras emerge in wide parameter region of mouse model","Persistent chimeras in mouse brain model defy narrow-parameter expectation","Weak coupling promotes lingering chimera states in mouse brain model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result rests on the assumption that the chimera-like index $\\chi$—a single number tracking how much synchrony varies across the 13 brain regions—truly detects coexisting synchronized and desynchronized domains, rather than merely picking up activity differences between regions or artifacts caused by the nearby unphysical boundary.","fun_headline_variants_meta":{"raw":{"variants":["Chimera states persist in weak-coupling zone of mouse brain","Seizure-like chimeras emerge in wide parameter region of mouse model","Persistent chimeras in mouse brain model defy narrow-parameter expectation","Weak coupling promotes lingering chimera states in mouse brain model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000901,"raw_usage":{"total_tokens":3854,"prompt_tokens":897,"completion_tokens":2957,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":2881}},"tokens_in":513,"tokens_out":2957,"duration_ms":21472,"temperature":1.0,"reasoning_tokens":2881,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:26:23.446576+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a parameter pair inside the low-coupling island (e.g., $\\alpha\\approx 0.04$, $\\beta\\approx 0.02$), record the instantaneous phases of all 213 subregions and map them spatially; the claim is falsified if no coarse area simultaneously contains a phase-locked cluster and a phase-scattered cluster, or if shuffling the intra-areal connections leaves $\\chi$ just as high.","supporting_citations":[],"review_version":1}