{"id":"6edd7f87-a8b3-45ab-8bad-6ec9ec02f989","arxiv_id":"2507.23678","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A conceptual review proposing that brain pathology, vulnerability, and recovery be analyzed through complex network structure, with the disease definition determining which network properties are relevant.","lead":"This is a review essay arguing that brain diseases can be understood as damage to the brain's network structure, and that the definition of disease determines which network properties matter. It offers a conceptual framework for talking about vulnerability, resilience, and recovery in network terms, without presenting new experiments.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Most load-bearing risk: the 'relevant to brain function' clause in the network-disease definition is unoperationalized and unfalsifiable; Sec. 3.4 itself says structural equivalence does not imply functional equivalence.","rationale":"The paper is a thoughtful review that repeatedly concedes its limits (Sec. 3.4, Sec. 4.2, Concluding remarks 4-5). However, the central definition in Sec. 3.1 carries a hidden empirical premise: that some non-trivial network structure is 'relevant to brain function' and that this relevance can be identified. The reader's weakest assumption targets the earlier premise that the brain is a genuine network. I find the 'relevance' premise more load-bearing because it remains open even if genuine networkness is established. A network can be real yet its topological/geometric features can be functionally inert epiphenomena; the paper itself emphasizes degeneracy and the unknown status of structural-to-functional equivalence. The paper does not supply a falsifiable criterion for the relevance clause, so the proposal that specific diseases are network diseases is currently unfalsifiable. This does not warrant rejecting the review — the authors explicitly frame it as a research agenda — but it strengthens the condition in the CONDITIONAL verdict. The proposed perturbation test for rich-club connectivity in schizophrenia is a single, concrete way to operationalize the relevance clause for a flagship case; if the rich-club-specific lesion is functionally silent, the network-disease reading of that case collapses. Thus the reader's conditional stance is appropriate and no verdict change is needed.","tokens_in":48328,"tokens_out":7443,"duration_ms":78570,"concrete_test":"Take the well-documented rich-club connectivity deficit in schizophrenia (van den Heuvel et al. 2010). In a whole-brain mean-field model constrained by human connectome data, lesion the rich club by reducing the weights of rich-club edges; in a control condition, apply matched random rewiring that preserves the node degree sequence and modular organization but deliberately disrupts rich-club topology. Simulate both conditions and measure a functional readout such as global integration capacity or performance on a working-memory-like delayed-activity task. If the rich-club-specific perturbation and the matched random rewiring produce indistinguishable functional outcomes, then the rich-club alteration in schizophrenia is not a damage to network structure 'relevant to brain function' in the sense of Sec. 3.1, directly falsifying the network-disease interpretation for this property.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Sec. 3.1) defines a network disease as damage to 'non-trivial network structure relevant to brain function.' The framework's traction depends on the 'relevant to brain function' clause, but the paper provides no operational criterion for it. Sec. 3.4 states: 'topological, geometric or combinatorial equivalence does not necessarily entail functional equivalence and whether such equivalence classes underlie brain function is unknown.' It also says: 'that a system has a given structure does not entail that such a structure is functional.' Sec. 4.2 concedes that altered network properties in a disease 'does not per se guarantee that the former is a genuine network disease.' Thus, even granting that the brain is a genuine network (resolving the reader's concern), the network-disease definition remains unfalsifiable: any observed network metric change could be declared functionally relevant, or not, without a test. The paper's caveats about degeneracy and unknown equivalence relations mean the structure-to-function step is an empirical unknown, not an acceptable premise. Without a method to establish which structural properties are functionally relevant, the claim that network properties characterize pathology cannot be distinguished from an epiphenomenal story, and the proposed pathoconnectomic clinical program lacks a foundation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper is a conceptual review/discussion of how complex network representations can be applied to brain disease. The authors distinguish anatomical, dynamical, and functional network structure, propose that 'network diseases' be defined as conditions in which non-trivial network structure relevant to brain function is damaged, and use this definition to organize a taxonomy of pathologies (local vs. non-local, anatomical vs. dynamical), a discussion of resilience (elastic/plastic regimes, cognitive reserve, reorganisation), and a clinical outlook (pathoconnectomics, control-based interventions). The paper is heavily referenced and explicitly acknowledges open questions about whether the brain's network structure is genuinely functional. It contains no new data or quantitative analysis; its contribution is an organizing framework.","tokens_in":48554,"tokens_out":3762,"duration_ms":38865,"significance":"If the framework were made testable, it would be a useful conceptual resource for network neuroscience and for translational research: it systematically connects definitions of disease, function, and network structure, raises the important question of when a network metric change is functionally relevant, and makes the modest but defensible point that 'networkness' should be treated as graded and disease-dependent rather than all-or-nothing. The paper's strengths are its breadth, its explicit caveats, and its careful distinction between structure, dynamics, and function. However, the central definition is currently unfalsifiable, and the paper's own admissions (Secs. 3.4 and 4.2) undermine the 'relevant to brain function' clause. The significance is therefore conditional on the authors adding operational criteria or reframing the thesis as a hypothesis.","major_comments":[{"comment":"Section 3.1 defines a network disease as a condition where non-trivial network structure relevant to brain function is damaged. The phrase 'relevant to brain function' is never given an operational criterion. Section 3.4 then states that topological, geometric or combinatorial equivalence does not necessarily entail functional equivalence, and Section 4.2 concedes that altered network properties in a disease do not per se guarantee that it is a genuine network disease. Taken together, these passages make the central definition unfalsifiable: any network metric change could be declared functionally relevant or not without a test. I recommend replacing the definition with a falsifiable version, for example by specifying an intervention or prediction (e.g., lesioning the structure should reproduce the disease phenotype, or network-based biomarkers must outperform non-network features in prospective classification), or by explicitly re-labelling it as a working hypothesis with stated disconfirming conditions.","section":"Sec. 3.1"},{"comment":"The clinical program in Sections 4.1 and 7 rests on pathoconnectomics as a biomarker, but the paper gives no account of how 'functionally relevant' structure is identified. Section 4.1 itself notes that defining links in dynamical networks is complicated because no connectivity metric is explicitly based on neurophysiology. Without a method to establish which structural properties are functionally relevant, the claim that network properties characterize pathology is indistinguishable from an epiphenomenal description. The manuscript would be strengthened by a dedicated subsection with concrete proposals for establishing structure-to-function links, such as perturbation experiments, controllability analyses, or lesion-behavior mapping in patient cohorts.","section":"Secs. 4.1 and 7"},{"comment":"Section 6.2 introduces the solid-material metaphor (yield point, ultimate tensile strength, toughness) without any formal link to brain network properties. The metaphor is used to organize Sections 6.3–6.5, but the mapping is asserted rather than derived; for example, 'elastic range width' in Sec. 6.4.2 is identified with cognitive reserve without a quantitative definition. As it stands, this is more analogy than framework, and it supports the central thesis only rhetorically. Please state explicitly what mathematical objects correspond to stress, strain, and yield, or demote the metaphor to an illustrative device.","section":"Secs. 6.2 and 6.4.2"}],"minor_comments":[{"comment":"The abstract says 'We show how the way disease is defined is related to the way function is defined,' but the paper argues and illustrates rather than formally shows; 'we argue' or 'we propose' would be more accurate.","section":"Abstract"},{"comment":"The lengthy etymological discussion of disease-related terms, while interesting, is not used to derive any of the later definitions or distinctions; consider condensing it to one short paragraph.","section":"Sec. 2.1"},{"comment":"The statement that some systems 'may not' meaningfully be equipped with a network structure (e.g., purely feedforward systems) is presented without a citation or example; this claim deserves support.","section":"Sec. 3.2"},{"comment":"The concept of 'microscopic interventions' is introduced with references but its difference from the earlier 'control strategies' is not explained until the end of the section; clarify earlier.","section":"Sec. 6.6.2"},{"comment":"Several references are incomplete or listed as in press (e.g., Papo and Buldú, 2024b; Buldú et al., 2024); these need to be completed before publication.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The reader's concern about the unfalsifiable 'relevant to brain function' clause is, in my reading, valid and is explicitly acknowledged by the authors in Secs. 3.4 and 4.2. Because the paper is framed as a discussion piece, this is not an immediate reason to reject, but it is a load-bearing weakness for a journal that expects the central thesis to be more than an organizing metaphor. The revision should either add operational criteria or explicitly reposition the paper as a hypothesis-generating review. The manuscript is also quite long and could be trimmed in places without loss of content."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a careful, honest review, not a research paper. It offers a genuinely useful organizing frame: how you define disease, and how you define function, determines which network property you should care about. That is clarifying even though the ingredients — dysconnection, pathoconnectomics, dynamical disease — are established. The authors also sharply distinguish dynamics from function and structure from anatomy, a distinction often muddled in the field.\n\nThe best parts are the disease taxonomy by network role (focal vs. non-local, anatomical vs. dynamical) and the discussion of degeneracy, neutral networks, and cognitive reserve as elastic-range resilience. The paper is heavily referenced, and the concluding remarks are admirably explicit: current network theory's sensitivity and specificity are not yet clinically relevant, and the role of higher-order network properties is poorly understood.\n\nThe soft spot is the one the stress-test flagged. The central definition — a network disease is damage to 'non-trivial network structure relevant to brain function' — depends on an unoperationalized clause. As stated, any network metric change could be declared functionally relevant or not. But the authors repeatedly concede the gap: structure does not entail function; altered network properties in a disease do not guarantee a genuine network disease; equivalence classes do not necessarily map to functional ones. So it is a known limitation, not a hidden flaw. What is missing is any concrete proposal for testing functional relevance, e.g., perturbation experiments, causal interventions, or falsifiable predictions. That keeps the framework at the level of a research agenda, which the authors themselves acknowledge.\n\nThis is a review that network neuroscientists and clinicians wanting a conceptual map will find useful. It deserves a serious referee; it is not a desk reject. For my own work I would not cite it unless writing a review, but I would bring it to a reading group to debate the definitional issue.\n\nRecommendation: send to peer review at a suitable network-science or neuroscience venue. The definitional gap should be raised with the authors, but it is a legitimate open problem, not a fatal flaw.","headline":"A careful, honest review that frames how disease definitions determine which network properties matter, but the key 'network disease' definition is unfalsifiable as stated — a known limitation, not a hidden flaw.","tokens_in":49038,"tokens_out":2008,"would_cite":false,"duration_ms":23064,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proposes that a brain disease is a network disease when non-trivial network structure relevant to brain function is damaged, with the definition of disease, the definition of function, and the choice of relevant network property…","keywords":["complex networks","brain disease","network disease","pathoconnectomics","resilience","vulnerability","degeneracy","network neuroscience"],"falsifier":"Take a disorder the paper classifies as a network disease, such as Alzheimer's or schizophrenia, and test in a computational model whether the characteristic spatiotemporal pattern of atrophy and functional impairment is reproduced by damaging the identified network property on the human connectome, while a degree-matched randomized network fails to reproduce it. If the randomized network reproduces the disease just as well, the network structure is not doing the causal work and the network-disease claim for that disorder would be falsified.","tokens_in":48132,"feed_emoji":"🧠","tokens_out":8053,"duration_ms":73700,"temperature":0.7,"pith_summary":"The paper argues that at least some brain disorders are best understood as network diseases: conditions in which the brain's non-trivial relational structure is what gets damaged, and that damage produces functional impairment. It contends that deciding whether a pathology counts as a network disease is not purely empirical, because the answer depends on how disease and function are defined, and those definitions determine which network property is functionally relevant. If this framing holds, maps of abnormal brain networks (pathoconnectomics) could serve as biomarkers for disease, predict vulnerability and recovery, and guide network-based interventions. The paper also notes that demonstrating genuine network diseases would be evidence that the brain genuinely behaves as a complex network rather than merely being represented as one.","feed_headline":"Network disease: damage to the structure behind brain function","feed_subtitle":"How disease is defined decides which network property matters—and could turn connectome maps into clinical tools.","key_machinery":"The carrying object is the complex-network representation of the brain: a graph whose nodes are brain regions, states, genes, or proteins and whose links are anatomical, dynamical, or functional relations. On top of this the paper places a three-way dependency among disease definition, function definition, and the network property judged relevant, plus two auxiliary structures: equivalence classes and neutral networks, which are sets of configurations that map to the same function, and the stress-strain metaphor for resilience, with an elastic range, yield point, and ultimate tensile strength. These machinery pieces translate vague notions like vulnerability, cognitive reserve, and reorganization into statements about which network properties are preserved, bent, or broken.","core_discovery":"On the paper's own terms, the central proposal is a definition: a network disease is a condition where non-trivial network structure relevant to brain function is damaged. The paper does not attempt to characterize the network structure of any single pathology; instead, it builds a general framework in which disease can be an ontological perturbation of an intact structure, a dynamical regime reached under certain parameters, or a process unfolding on a network. The way disease is conceived is tied to the way brain function is conceived, and that pair of choices fixes which network property, anatomical or dynamical, topological or geometric, local or global, is the functionally relevant one. From there the paper derives a taxonomy of brain diseases by how they damage network structure, and organizes vulnerability, cognitive reserve, and recovery through a material-science metaphor of elastic resilience, plastic reorganization, and structural failure. It concludes that current network measures are not yet clinically specific enough, but pathoconnectomics could become a clinical tool if the network-disease framing holds.","pith_inferences":["If the three-way dependency is correct, contradictory findings in network neuroscience about which metric matters for a disorder may trace back to implicit disagreements about what counts as function; comparing classifications under different function definitions would test this directly.","The stress-strain metaphor suggests a testable prediction: disease progression should show an elastic regime where network properties recover, then plastic reorganization to equifunctional structures, then failure marked by critical slowing down, eigenvector localization, and decreased topological dimension, patterns already known from hierarchical materials.","The neutral-network view of cognitive reserve implies that interventions aimed at increasing redundancy or degeneracy, such as training or cognitive enrichment, should expand the neutral space and delay symptom onset, which could be tested by measuring network redundancy before and after such interventions.","If pathoconnectomics becomes a biomarker, the same framework implies that treatment efficacy should be evaluated by whether the network property identified as functionally relevant returns toward the healthy range, not just by symptom scores."],"forward_implications":["If the network-disease definition is adopted, disorders can be classified by what they damage: nodes versus links, anatomy versus dynamics, topology versus geometry, and local versus non-local consequences.","Abnormal connectome maps, called pathoconnectomics, could become biomarkers that detect disease before behavioral symptoms, gauge severity, and define patient theratypes separating therapy responders from non-responders.","Vulnerability and recovery become measurable network properties, so cognitive reserve and symptom onset can be studied as extension of the elastic range of a networked system.","If network structure is essential to both function and dysfunction, it could be acted upon through network control, targeted stimulation, seizure-network surgery, or microscopic interventions that keep the system in a healthy regime.","Demonstrating at least one genuine network disease would support the broader claim that the brain genuinely behaves as a complex network and not merely that network language is a convenient description."],"supporting_citations":[{"why":"Supplies the standard network-neuroscience framing that the paper extends to disease.","marker":"(Bullmore and Sporns, 2009)"},{"why":"Defines complex networks and the non-trivial properties distinguishing them from lattices or random graphs.","marker":"(Albert and Barabási, 2002)"},{"why":"Founds the dynamical-disease approach the paper uses for disease as a regime of the system.","marker":"(Mackey and Glass 1977)"},{"why":"Provides the quenched-disorder mechanism by which anatomical network structure can drive critical dynamics without fine tuning.","marker":"(Moretti and Muñoz, 2013)"},{"why":"Evidence that neurodegenerative disease spreads along fibre pathways, grounding disease-as-process-on-network.","marker":"(Seeley et al., 2009)"},{"why":"Models protein spreading as network diffusion, used to predict atrophy patterns in dementias.","marker":"(Raj et al., 2012)"},{"why":"Defines degeneracy, the key mechanism the paper invokes for robustness and cognitive reserve.","marker":"(Edelman and Gally, 2001)"},{"why":"Maps the connectomics-of-brain-disorders program that pathoconnectomics extends.","marker":"(Fornito et al., 2015)"}],"fun_headline_variants":["Brain disease as network damage: a new framework","How disease definition shapes the relevant network measure","Viewing brain pathologies through a network lens","Network damage may define brain disease and recovery"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The framework rests on the assumption that the brain genuinely has non-trivial network structure that is functionally relevant, because if that structure is only a convenient way to draw anatomy and activity, the network-disease ontology collapses.","fun_headline_variants_meta":{"raw":{"variants":["Brain disease as network damage: a new framework","How disease definition shapes the relevant network measure","Viewing brain pathologies through a network lens","Network damage may define brain disease and recovery"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000255,"raw_usage":{"total_tokens":1540,"prompt_tokens":880,"completion_tokens":660,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":604}},"tokens_in":496,"tokens_out":660,"duration_ms":6442,"temperature":1.0,"reasoning_tokens":604,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T10:27:39.899346+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a disorder the paper classifies as a network disease, such as Alzheimer's or schizophrenia, and test in a computational model whether the characteristic spatiotemporal pattern of atrophy and functional impairment is reproduced by damaging the identified network property on the human connectome, while a degree-matched randomized network fails to reproduce it. If the randomized network reproduces the disease just as well, the network structure is not doing the causal work and the network-disease claim for that disorder would be falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Models protein spreading as network diffusion, used to predict atrophy patterns in dementias."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Maps the connectomics-of-brain-disorders program that pathoconnectomics extends."}],"review_version":1}