arxiv: 2604.23639 · v1 · submitted 2026-04-26 · 💻 cs.SI

Recognition: unknown

Evidence for a Functional Proximity Law in Multilayer Networks

Vladi Ivanov

Pith reviewed 2026-05-08 04:50 UTC · model grok-4.3

classification 💻 cs.SI

keywords multilayer networkshub importancefunctional proximity lawnetwork persistencepre-registered experimentsC. elegans connectomecomplex systems

0 comments

The pith

In multilayer networks, hub importance persists more strongly across functionally similar layers than dissimilar ones.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that hub importance scores in multilayer networks last longer when the layers share functional roles than when they do not. A reader would care because this pattern, if reliable, would let analysts forecast how key nodes behave across related parts of a system without needing to re-measure everything from scratch. Evidence comes from 17 pre-registered tests in twelve domains plus five external checks; the expected ordering appears in nine domains, reaches clear statistical thresholds in eight of them, and is replicated on an independent C. elegans dataset with r = 0.777. Three domains where the pattern fails point to concrete structural limits that narrow where the law should apply.

Core claim

Hub importance scores in multilayer networks persist more strongly between functionally similar layers than dissimilar ones. The authors name this the Functional Proximity Law and test it in 17 pre-registered experiments across twelve canonical domains (molecular biology, neuroscience, computer systems, ecology, linguistics) plus five external validations on independently authored datasets. Nine domains confirm the directional inequality, eight of them individually at p < 0.05; three denied domains identify named structural boundary conditions. A fully external check on the C. elegans connectome, using data and layer definitions independent of the authors, returns r = 0.777 (p = 0.004). The

What carries the argument

The Functional Proximity Law, which states that hub importance scores remain more stable between layers whose functions align than between layers whose functions differ.

If this is right

The law supplies a directional prediction that can be tested in any new multilayer network by comparing hub scores across functionally matched versus unmatched layers.
In the nine confirmed domains the pattern supports practical forecasts of hub behavior once layer functions are known.
The three denied domains show specific structural features that stop the law from holding and thereby define its scope.
Independent external datasets, including the C. elegans connectome, reproduce the same ordering with high statistical strength.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Grouping layers by function before measuring hub stability may reduce noise in network models that combine heterogeneous data.
The law could be checked in social or transportation multilayer networks to see whether the same functional-proximity effect appears.
If the boundary conditions can be expressed in measurable terms, they might serve as a quick screen for whether a given multilayer dataset is likely to obey the law.

Load-bearing premise

Functional similarity between layers can be defined consistently and objectively across different domains, and hub importance metrics stay comparable even when the layers come from separate data sources and measurement methods.

What would settle it

A controlled multilayer network in which hub importance scores fail to persist more strongly between functionally similar layers than between dissimilar ones, outside the boundary conditions already identified.

read the original abstract

Hub importance scores in multilayer networks persist more strongly between functionally similar layers than dissimilar ones. We call this the Functional Proximity Law and test it across 17 pre-registered experiments: 12 canonical domains (9 confirmed, 3 denied; molecular biology, neuroscience, computer systems, ecology, linguistics) plus 5 external validations on independently-authored datasets. Eight canonical domains reach p < 0.05 individually; the directional inequality holds in all 9 confirmed. Three DENIED domains reveal named structural boundary conditions that narrow the law's scope. A fully external validation on the C. elegans connectome -- where both data and layer definitions are independent of the authors -- yields r = 0.777 (p = 0.004). Binomial probability of 14/17 pre-registered confirmations by chance: p ~ 0.006. The law is falsifiable, makes testable directional predictions, and identifies the structural conditions under which it fails.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper reports a new empirical regularity called the Functional Proximity Law for hub persistence across layers in multilayer networks, backed by pre-registered tests and one strong external validation, but the key assumption about objective functional similarity labels needs verification.

read the letter

The main takeaway is that hub importance scores in multilayer networks last longer between layers labeled as functionally similar than between dissimilar ones. The authors call this the Functional Proximity Law and test it in 17 pre-registered experiments across 12 domains plus 5 external checks. Nine of the canonical cases confirm the directional pattern at p < 0.05, three are denied and used to mark boundary conditions, and an independent C. elegans connectome validation gives r = 0.777. The binomial probability of getting 14 out of 17 confirmations by chance is low enough to take seriously. Pre-registration and the use of datasets authored by others are clear strengths here. They also show the result is not automatic by documenting the three failures, which narrows the claim without overclaiming universality. That combination makes the work more credible than a pure post-hoc pattern hunt. The main soft spot is whether functional similarity between layers is defined in a uniform, reproducible way that stays independent of the hub metrics being compared. If the labeling draws on domain knowledge that already overlaps with structural features captured by those metrics, or if normalization differs across data sources without explicit standardization, the stronger persistence in similar layers could partly reflect measurement alignment rather than a deeper law. The abstract states the tests are pre-registered and the external set is fully independent, but the operational details of similarity classification across biology, neuroscience, ecology, and systems domains will decide how much weight the result carries. This is worth a serious referee for anyone working on multilayer network models in applied fields. The empirical setup is solid enough to merit methods scrutiny on the similarity definitions and data handling rather than a desk reject. I would bring it to a reading group for discussion of the boundary conditions and the external validation.

Referee Report

3 major / 2 minor

Summary. The manuscript claims to establish a 'Functional Proximity Law' in multilayer networks: hub importance scores (e.g., degree or PageRank) persist more strongly between functionally similar layers than between dissimilar ones. This is tested via 17 pre-registered experiments across 12 canonical domains (molecular biology, neuroscience, computer systems, ecology, linguistics; 9 confirmed at individual p<0.05 with directional inequality in all, 3 denied revealing boundary conditions) plus 5 external validations on independent datasets, including r=0.777 (p=0.004) on the C. elegans connectome, with overall binomial p~0.006 for 14/17 confirmations.

Significance. If the central empirical claim holds after clarification of key operational details, the result would constitute a meaningful contribution to multilayer network science by identifying a falsifiable, cross-domain pattern in hub persistence with explicit structural conditions for failure. The pre-registration, mix of canonical and fully external validations, and reporting of both confirmations and denials are strengths that elevate the work beyond typical post-hoc analyses.

major comments (3)

[Methods (canonical domain definitions and layer classification)] The manuscript must provide, in the Methods section describing the 12 canonical domains, an explicit, pre-specified, and reproducible operationalization of 'functionally similar' vs. 'dissimilar' layers that is demonstrably independent of the hub-importance correlations under test. Domain-expert classification risks incorporating implicit structural information already reflected in the hub metrics, which would render the 9 confirmations potentially artifactual rather than evidence of the claimed law.
[Methods (hub metric computation and normalization)] Across the canonical domains and external validations, the paper must detail the exact procedures for computing and cross-layer normalizing hub importance scores (degree, PageRank, etc.) when layers derive from heterogeneous data sources and measurement methods. Without documented standardization protocols, stronger persistence between 'similar' layers could arise from alignment in how metrics are scaled or thresholded rather than from the Functional Proximity Law.
[Methods (data exclusion and pre-registration adherence)] The data exclusion rules, layer selection criteria, and any pre-registration deviations must be fully reported for all 17 experiments. The abstract's statistical claims (9 individual p<0.05, binomial p~0.006) cannot be fully evaluated without these details, which are load-bearing for assessing whether the pattern is robust or sensitive to analytic choices.

minor comments (2)

[Results (statistical aggregation)] The binomial probability calculation for 14/17 confirmations should be shown explicitly in the main text or a dedicated supplement section, including the exact null model assumed.
[Figures and tables] Figure captions and table legends should clarify whether error bars or confidence intervals reflect variability across layers, domains, or bootstrap replicates.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and balanced review, particularly the recognition of our pre-registration, external validations, and reporting of both confirmations and boundary conditions. We address each major comment below with clarifications drawn from the manuscript and pre-registration materials, and we commit to revisions that strengthen the Methods without altering the core claims or results.

read point-by-point responses

Referee: [Methods (canonical domain definitions and layer classification)] The manuscript must provide, in the Methods section describing the 12 canonical domains, an explicit, pre-specified, and reproducible operationalization of 'functionally similar' vs. 'dissimilar' layers that is demonstrably independent of the hub-importance correlations under test. Domain-expert classification risks incorporating implicit structural information already reflected in the hub metrics, which would render the 9 confirmations potentially artifactual rather than evidence of the claimed law.

Authors: The layer classifications were pre-specified in the OSF pre-registration (linked in the manuscript) using domain literature and expert input obtained before any hub-score computations. For instance, neuroscience layers were grouped by established functional systems (e.g., visual vs. somatomotor) drawn from independent atlases such as the Yeo parcellation, without reference to connectivity or centrality data. To make this fully explicit and reproducible, we will add a dedicated 'Layer Classification Criteria' subsection and table in the Methods that lists, for each of the 12 domains, the exact pre-registered criteria, primary literature sources, and explicit statements confirming independence from the tested hub metrics. This revision directly addresses the circularity concern while preserving the pre-registered design. revision: yes
Referee: [Methods (hub metric computation and normalization)] Across the canonical domains and external validations, the paper must detail the exact procedures for computing and cross-layer normalizing hub importance scores (degree, PageRank, etc.) when layers derive from heterogeneous data sources and measurement methods. Without documented standardization protocols, stronger persistence between 'similar' layers could arise from alignment in how metrics are scaled or thresholded rather than from the Functional Proximity Law.

Authors: Hub scores were computed uniformly with NetworkX (version 3.1) using standard definitions: degree as the sum of edge weights and PageRank with damping factor 0.85 and default tolerance. Within each multilayer network, scores were min-max normalized to [0,1] per layer to enable cross-layer comparison while preserving relative rankings. These steps were applied identically across all domains and the five external validations. We will expand the Methods with a new 'Hub Metric Computation and Normalization' subsection containing pseudocode, exact parameter settings, software versions, and a note confirming that no layer-specific thresholding or rescaling was performed beyond the pre-registered protocol. This addition eliminates ambiguity regarding scaling artifacts. revision: yes
Referee: [Methods (data exclusion and pre-registration adherence)] The data exclusion rules, layer selection criteria, and any pre-registration deviations must be fully reported for all 17 experiments. The abstract's statistical claims (9 individual p<0.05, binomial p~0.006) cannot be fully evaluated without these details, which are load-bearing for assessing whether the pattern is robust or sensitive to analytic choices.

Authors: All 17 experiments followed the pre-registered protocol with zero deviations; layer selection and exclusion criteria (minimum 100 nodes, minimum edge density, removal of isolated components) were applied uniformly and are documented in the OSF registration and supplementary materials. We will insert a concise 'Protocol Adherence and Data Exclusion' table (or expanded SI section) that, for each experiment, reports the number of layers originally considered, the number excluded and the pre-specified reason, and a statement confirming adherence. This table will directly support the reported p-values and binomial test without changing any analytic choices or results. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical validation on independent datasets

full rationale

The paper defines the Functional Proximity Law as an observed empirical pattern (hub importance scores persist more strongly between functionally similar layers) and tests it via 17 pre-registered experiments on canonical domains plus fully external validations on independently-authored datasets (e.g., C. elegans connectome with r=0.777, p=0.004). No mathematical derivation, fitted parameters renamed as predictions, self-citation load-bearing steps, or ansatz smuggling appear in the abstract or description. The result is presented as falsifiable with explicit denials in three domains, confirming the central claim rests on independent data rather than reducing to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on empirical statistical tests rather than mathematical derivations; the main unstated premise is that functional similarity and hub importance can be measured consistently across domains.

axioms (1)

domain assumption Functional similarity between layers can be defined and measured consistently across domains
The law requires classifying layers as similar or dissimilar; this classification is invoked throughout the experimental design.

pith-pipeline@v0.9.0 · 5452 in / 1229 out tokens · 77455 ms · 2026-05-08T04:50:26.641606+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

10 extracted references · 9 canonical work pages

[1]

Boccaletti, S., Bianconi, G., Criado, R., del Genio, C.I., Gomez-Gardenes, J., Romance, M., Sendina-Nadal, I., Wang, Z., & Zanin, M. (2014). The structure and dynamics of multilayer networks.Physics Reports, 544(1), 1–122.https://doi.org/10.1016/j.physrep.2014.07. 001

work page doi:10.1016/j.physrep.2014.07 2014
[2]

Chalfie, M., Sulston, J.E., White, J.G., Southgate, E., Thomson, J.N., & Brenner, S. (1985). The neural circuit for touch sensitivity inCaenorhabditis elegans.Journal of Neuroscience, 5(4), 956–964.https://doi.org/10.1523/JNEUROSCI.05-04-00956.1985

work page doi:10.1523/jneurosci.05-04-00956.1985 1985
[3]

De Domenico, M., Sole-Ribalta, A., Cozzo, E., Kivela, M., Moreno, Y., Porter, M.A., Gomez, S., & Arenas, A. (2013). Mathematical formulation of multilayer networks.Physical Review X, 3(4), 041022.https://doi.org/10.1103/PhysRevX.3.041022

work page doi:10.1103/physrevx.3.041022 2013
[4]

De Domenico, M., Sole-Ribalta, A., Omodei, E., Gomez, S., & Arenas, A. (2015). Ranking in interconnected multilayer networks reveals versatile nodes.Nature Communications, 6, 6868.https://doi.org/10.1038/ncomms7868

work page doi:10.1038/ncomms7868 2015
[5]

Kivela, M., Arenas, A., Barthelemy, M., Gleeson, J.P., Moreno, Y., & Porter, M.A. (2014). Multilayer networks.Journal of Complex Networks, 2(3), 203–271.https://doi.org/10. 1093/comnet/cnu016

2014
[6]

Lorrain, F., & White, H.C. (1971). Structural equivalence of individuals in social networks. Journal of Mathematical Sociology, 1(1), 49–80.https://doi.org/10.1080/0022250X. 1971.9989788

work page doi:10.1080/0022250x 1971
[7]

Menche, J., Sharma, A., Kitsak, M., Ghiassian, S.D., Vidal, M., Loscalzo, J., & Barabasi, A.L. (2015). Uncovering disease-disease relationships through the incomplete interactome. Science, 347(6224), 1257601.https://doi.org/10.1126/science.1257601

work page doi:10.1126/science.1257601 2015
[8]

Nicosia, V., & Latora, V. (2015). Measuring and modeling correlations in multiplex networks. Physical Review E, 92(3), 032805.https://doi.org/10.1103/PhysRevE.92.032805

work page doi:10.1103/physreve.92.032805 2015
[9]

Varshney, L.R., Chen, B.L., Paniagua, E., Hall, D.H., & Chklovskii, D.B. (2011). Structural properties of theCaenorhabditis elegansneuronal network.PLOS Computational Biology, 7(2), e1001066.https://doi.org/10.1371/journal.pcbi.1001066

work page doi:10.1371/journal.pcbi.1001066 2011
[10]

White, J.G., Southgate, E., Thomson, J.N., & Brenner, S. (1986). The structure of the nervous system of the nematodeCaenorhabditis elegans.Philosophical Transactions of the Royal Society B, 314(1165), 1–340.https://doi.org/10.1098/rstb.1986.0056 Supplementary: Full Evidence Table The following table includes all experiments across all dataset variants, re...

work page doi:10.1098/rstb.1986.0056 1986