Pith. sign in

REVIEW 2 major objections 5 minor 60 references

Component-removing blind source separation can destroy the causal signal in calcium traces.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Component-removing blind source separation can collapse causal graph recovery to zero in synthetic calcium traces and dramatically densify connectivity estimates from real v2a-RSN traces, so BSS denoising is not a neutral preprocessing step.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection Useful cautionary study with a genuine synthetic benchmark; the headline overreaches slightly because only one component-selection family is tested. the 2 major comments →

arxiv 2608.00655 v1 pith:ZS4S6TJ7 submitted 2026-08-01 stat.AP

Blind Source Separation Can Distort Behavior and Connectivity Analyses of Calcium Transients

Adedayo S. A This is my paper
classification stat.AP
keywords blind source separationcalcium imagingcausal structure learningGranger causalitydenoisingbehavior decodingcomponent selectionconnectivity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper argues that denoising calcium traces by blind source separation (BSS)—decomposing the traces into components and discarding some—is not a neutral cleanup step. In a synthetic benchmark with known lagged causal graphs, raw corrupted traces still allowed graph recovery, but every component-removing BSS variant collapsed median graph recovery to zero across four estimator families. On four larval zebrafish recordings of v2a reticulospinal neurons, BSS sometimes improved behavior decoding, but the gains were inconsistent across fish, methods, and retained component clusters, and matched PCA or low-pass controls often matched or beat them. The paper's central conclusion is that BSS should be treated as an intervention on the measured process, and its effect on the specific downstream analysis—especially causal structure learning—should be validated before trusting the cleaned traces.

Core claim

The central discovery is that removing components after blind source separation can delete the temporal dependencies that causal structure learning needs, while leaving behavior decoding roughly intact or even improved. In the synthetic benchmark, median graph F1 for raw corrupted traces was 0.585 (c-GC), 0.500 (c-GC*), and 0.658 (PCMCI+/JPCMCI+), while all component-removing BSS variants returned median F1 of 0, with zero true and false positives and 17–18 false negatives. The paper further shows that on v2a-RSN recordings, BSS-cleaned traces produce much denser c-GC and c-GC* connectivity graphs (49–88% off-diagonal directed edges) than raw traces (4.5–6.7%), meaning preprocessing dominate

What carries the argument

The load-bearing mechanism is the component-removing reconstruction: BSS decomposes the trace matrix into source time courses and loadings, clusters components by spectral features (Welch power spectra, k-means), retains only a subset of clusters, and reconstructs traces from the retained components alone. This selection step deletes coordinates that may carry behavior-relevant and graph-relevant signal. The evaluation framework—fold-local leakage audits, matched PCA and low-pass controls, behavior scorecards, and causal-state diagnostics—makes the distortion visible, but the deletion itself is the intervention that erases causal evidence.

Load-bearing premise

The negative result rests on how components are selected for removal: spectral clustering into seven clusters with retention of top clusters by peak mean power, a rule that may not match how practitioners choose components; if expert or variance-based selection keeps more graph-relevant components, the distortion could be smaller.

What would settle it

Run the same synthetic benchmark with a component-selection rule that removes only components whose time courses correlate with an injected artefact probe above a threshold (an oracle-like artefact removal). If median graph F1 stays near or above the raw-trace level (e.g., ≥0.5 for c-GC), then the collapse to zero is an artifact of the spectral-cluster selection rule, not of component-removing BSS in general.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Researchers who clean calcium traces with BSS before Granger causality or related causal analyses should report raw-trace connectivity as a sensitivity check, because cleaned graphs can be far denser.
  • Component-removing BSS should not be assumed safe for causal structure learning; its effect on graph recovery should be validated on synthetic data with known ground truth before use on real data.
  • BSS can improve behavior decoding, but because matched PCA and low-pass controls often achieve the same or better scores, such improvements do not by themselves justify BSS as a denoising step.
  • The separation between trace preservation and graph recovery (raw traces had median trace correlation 0.83 vs clean, BSS-removed 0.60) implies that trace similarity to raw data is a poor proxy for preserving causal evidence.

Where Pith is reading between the lines

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

  • Editorial inference: The same audit logic applies to any preprocessing that transforms the trace matrix—PCA, low-pass filtering, deconvolution, or deep-learning denoisers—so the paper's framework could serve as a general validation protocol for denoising before causal analysis.
  • Editorial inference: The density inflation after BSS suggests that component removal may suppress variance that Granger causality uses to reject spurious edges, or introduce correlations by reconstruction; this could be tested by comparing BSS-cleaned graphs to graphs from rank-matched random subspaces in the synthetic benchmark.
  • Editorial inference: A natural extension is to test component-selection rules based on explained variance or artefact-probe correlation rather than spectral clustering; the paper's negative result may be sensitive to that choice, and such rules could preserve more graph-relevant information.
  • Editorial inference: Because behavior decoding gains from BSS were often matched by smoothing controls, a plausible hypothesis is that BSS improves decoding by smoothing, not by recovering neural signal; this could be tested by comparing BSS to a low-pass filter matched on dynamic state diagnostics.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper asks whether blind source separation (BSS) applied to already-extracted calcium traces preserves the information needed for behavior decoding and causal structure learning. It introduces a synthetic benchmark with known lagged graphs and corrupted observations, and reports that raw corrupted traces retain recoverable graph information (median F1 ≈ 0.5–0.66), whereas component-removing BSS variants collapse median graph F1 to 0 across c-GC, c-GC*, PCMCI+, and JPCMCI+. On four larval zebrafish v2a-RSN recordings, the paper finds that BSS sometimes improves behavior-decoding scores, but matched PCA and low-pass controls often match or exceed BSS in fold-audited comparisons, and BSS-cleaned traces yield much denser c-GC/c-GC* graphs than raw traces. The authors conclude that BSS should be treated as an intervention on the measured process, not as a neutral cleanup step, and they provide a validation framework combining trace preservation, behavior scorecards, leakage audits, matched controls, causal-state diagnostics, and graph sensitivity.

Significance. If the central claim holds, the paper is a useful cautionary study for the calcium-imaging preprocessing community: component-removing BSS is not automatically safe for downstream causal analysis. The main strengths are the synthetic ground-truth benchmark with known lagged graphs, the use of four BSS methods and several causal estimators, the fold-local leakage audit, matched PCA/low-pass baselines, and the explicit, reproducible code repository. The authors are also transparent about many limitations, including post hoc best-of-grid selection and missing uncertainty intervals. However, the breadth of the headline negative result is constrained by the fact that only one component-selection family is tested, which limits the generality of the quantitative claims even though the qualitative advice is sound.

major comments (2)
  1. [Abstract, §3.1, Table 1] The central negative result — component-removing BSS collapses median graph F1 to 0 — is demonstrated only for the cluster-retention selection family (k-means spectral clustering into seven clusters, retaining/removing whole clusters; §2.6, Table S2). The artifact/protection scoring rule defined in S1.6, which drops a component only when artifact evidence exceeds a threshold and protection evidence stays below its maximum, is never run in the synthetic benchmark or in the v2a-RSN connectivity analysis (Table 4). Because that rule is explicitly designed to protect behavior-relevant components, the headline overgeneralizes. Either add experiments with the artifact/protection rule and at least one explained-variance-based selection rule, or qualify every occurrence of "component-removing BSS" as "cluster-retention BSS".
  2. [§3.4, Table 4] The empirical connectivity-density result is also tied to a single selection rule, cluster_keep_top_04, applied with all four BSS methods. There is no ground-truth graph, so the only readout is density; using one operationalization of "component removal" makes the claim "BSS cleaning makes v2a-RSN connectivity matrices denser" narrower than stated. The paper's own artifact/protection rule could plausibly yield different density shifts. Please report sensitivity to at least one additional selection rule (e.g., artifact-score threshold or variance-retained) or explicitly state in the main text and abstract that this is a single-selection-rule sensitivity result.
minor comments (5)
  1. [§2.10 and §3.4] The sentence "The lag τ=2 was an assumed value corresponding to approximately one-third of the acquisition frequency, and npasts=4 was chosen as it is a sufficient conditioning depth" appears nearly verbatim twice. State it once and refer back.
  2. [Figure 2 caption] The caption uses abbreviations cgc, cgc_star, jpcmciplus, and pcmci without expanding them, and panel (a) "Best BSS graph F1" includes all-component and oracle-selection outputs. The main text notes this is a diagnostic envelope, but the caption alone should make that clear.
  3. [Table 3 and Figure 5] Define abbreviations in captions/footnotes: "Ctrl.", "dyn.", "BA". Currently they are only clear from context after reading the Methods.
  4. [§3.3] The phrase "The best BSS variant improved several raw behavior readouts" should explicitly say "best over the audited grid" to avoid the implication of a prespecified successful variant; the paper's own post hoc caveat is in the text but could be moved closer to the results.
  5. [§2.2] For readers not familiar with the author's earlier work, a one-sentence definition of c-GC versus c-GC* would be helpful at first use; currently the distinction is only clear from context and reference [52].

Circularity Check

0 steps flagged

No circularity: the core graph-recovery and density comparisons are self-contained empirical contrasts; only minor hyperparameter self-citations appear and they are not load-bearing.

full rationale

The paper's derivation chain is empirical, not definitional. In the synthetic benchmark, a known lagged VAR graph is generated, corrupted traces are passed through BSS with spectral-cluster component selection, and graph recovery is scored against the ground-truth adjacency matrix. The component-selection rules (seven k-means spectral clusters, retain 1-7 clusters) are derived from spectral features, not from the graph target, so the median F1=0 result is a measured outcome rather than an identity. The empirical v2a-RSN analyses explicitly state that no edge-level ground truth is available and report graph densities as sensitivity outputs; the BSS-vs-raw density contrast uses identical estimators and conditioning, so it does not reduce to a fitted parameter. BPI is normalized so that raw traces equal 100 by construction, but the paper states this openly and uses BPI only as a descriptive scorecard, not as a predicted outcome. The only author self-citations are the c-GC estimator [52] and the npasts=4 conditioning-depth choice [55]; these set a method/hyperparameter, while the main synthetic result holds across external estimator families (PCMCI+/JPCMCI+ via Tigramite). The untested artifact/protection selection rule is a generalization-scope limitation, not circularity. All learned operations are placed inside fold boundaries and the paper withholds state-preservation claims where diagnostics are missing. No circular step is present; the score reflects only minor, non-load-bearing self-citations.

Axiom & Free-Parameter Ledger

10 free parameters · 5 axioms · 0 invented entities

The paper's central claim rests on a complex preprocessing pipeline with many hand-chosen hyperparameters, the most important being the retained component set under spectral clustering. The synthetic benchmark also relies on a specific generative model that is not validated against external data. There are no newly invented theoretical entities. The c-GC estimator settings are self-cited, which adds a mild circularity concern.

free parameters (10)
  • Number of spectral clusters = 7
    k-means cluster count used for component grouping (§2.6, Table S2); not derived from data.
  • Retained cluster set (cluster_keep_top_04) = top 4 of 7 clusters
    The component-removal rule used for the v2a-RSN connectivity table (Table 4) and for the synthetic F1 collapse; a hand-selected subset of the 7 spectral clusters.
  • PCA rank for matched control = 40
    Rank used for the matched PCA baseline in the audited grid (Table S2); chosen by hand, not data-driven.
  • Low-pass cutoff frequencies = 0.1, 0.2, 0.4 Hz
    One-sided causal low-pass controls used as baselines (§2.9, Table S2); hand-chosen range.
  • Granger causal lag tau = 2 frames (~1/3 acquisition frequency)
    Fixed lag for all connectivity estimation (§2.10); assumed value not selected from data.
  • Past conditioning depth npasts = 4
    Conditioning depth for tau=2, justified via self-citation Adedayo et al. [55] (§2.10); a free choice.
  • Ridge alpha for decoders = 10
    Regularization strength for the simple behavior decoder and causal-state transition models (§2.8, Table S2).
  • Latent state dimension = 3
    Dimension of the fold-local PCA encoder in the causal-state diagnostic (§2.8).
  • History window w = 15 frames
    History length for the causal-state diagnostic, producing overlapping 14-frame shifts (§2.8, Table S2).
  • Welch spectral settings = nperseg=250, noverlap=125, discard bins before index 30
    Parameters for spectral feature computation used in component clustering (Table S2).
axioms (5)
  • domain assumption Calcium fluorescence traces are a linear instantaneous mixture of latent neural sources plus additive artefacts.
    The BSS decomposition X = S A^T (§2.5) assumes linear mixing; nonlinear or convolutionally mixed sources would violate this.
  • standard math The true causal graph is recoverable from latent VAR activity under the standard observational CSL assumptions (Markov, faithfulness, finite memory, causal sufficiency).
    Used to interpret synthetic graph recovery as causal evidence (§2.2).
  • domain assumption Spectral features (Welch power ratios, entropy, narrowband dominance) cluster into components that separate artefact from neural signal.
    Component selection rule (§2.6) assumes spectral clustering is informative for distinguishing artefact components from signal.
  • ad hoc to paper Behavior targets are generated by delayed latent parents with fixed linear weights, as in the synthetic benchmark (Figure 1b).
    The synthetic benchmark's behavior generation is a design choice for this paper, not an externally validated model.
  • domain assumption c-GC and c-GC* with tau=2 and npasts=4 are valid estimators for the recorded v2a-RSN processes.
    The connectivity density results depend on these estimators; the settings are justified by the author's prior work [55] rather than by a data-driven selection (§2.10).

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Blind Source Separation Can Distort Behavior and Connectivity Analyses of Calcium Transients." pith.science (2026). https://pith.science/paper/ZS4S6TJ7

@misc{pith2026260800655,
  author       = {Pith},
  title        = {Pith review of: Blind Source Separation Can Distort Behavior and Connectivity Analyses of Calcium Transients},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZS4S6TJ7}},
  note         = {Machine review of arXiv:2608.00655}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Denoising is often treated as a technical prelude to calcium transients analysis, but it can redefine the variables used for behavior decoding and causal structure learning. An inspectable motorneuron recording with a visible artefact motivated this study: removing artefact-linked blind source separation (BSS) components also changed trace dynamics outside the targeted frame. We therefore tested whether BSS denoising preserves behavior and causal evidence on already extracted calcium transients. In a synthetic benchmark with known lagged graphs, raw corrupted traces retained graph recovery, whereas component-removing BSS variants collapsed median graph F1 to 0 across estimator families. We then evaluated four BSS methods: FastICA, Infomax, SOBI, and JADE, on four larval zebrafish recordings of v2a reticulospinal neurons (v2a-RSNs) with tail behavior. BSS sometimes improved behavior decoding, but gains depended on fish, method and retained clusters; matched PCA and low-pass controls often matched or exceeded BSS in fold-audited comparisons. Connectivity effects were more consistent: c-GC and c-GC* graphs inferred from BSS-cleaned traces were much denser than raw-trace graphs. BSS should therefore be treated as an intervention on the measured process, not a neutral cleanup step.

Figures

Figures reproduced from arXiv: 2608.00655 by Adedayo S. A.

Figure 1
Figure 1. Figure 1: Synthetic benchmark construction. (a) Artefact scaling for one replicate. (b) Delayed latent-parent behavior targets. Raw corrupted traces retained recoverable graph information. Median raw graph F1 was 0.585 for c-GC, 0.500 for c-GC* and 0.658 for J-/PCMCI+ ( [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Synthetic graph-recovery diagnostics. (a) Scenario-level upper envelope by BSS method and estimator; plotted best variants include all-component and oracle-selection outputs and should be read diagnostically. (b) Paired graph F1 change relative to raw traces; negative values indicate reduced graph recovery. The degradation was not a single false positive effect. Raw c-GC and c-GC* produced more erroneous d… view at source ↗
Figure 3
Figure 3. Figure 3: Infomax IC cluster diagnostics for the fish-1 v2a-RSN fluorescence recording. Left, IC feature clusters; right, cluster mean log power spectral density [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Whole recording BPI diagnostics. (a) Values are normalized to raw traces as retained spectral clusters are added. (b) Retaining more clusters generally increases trace preservation, but behavior decoding can peak at intermediate reconstructions. fish 1 fish 2 fish 4 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Vigor Pearson a Higher better fish 1 fish 2 fish 4 0.5 0.6 0.7 0.8 Bout balanced accuracy b Higher better fish… view at source ↗
Figure 5
Figure 5. Figure 5: condenses the audited grid to three representation classes: raw traces, the best BSS variant, and the best matched control. This avoids plotting the full candidate list while preserving the main contrast: BSS can help behavior readouts, but the control remains competitive and gives the lowest dynamic MSE. a. BPI as retained clusters are added. b. BPI against trace preservation [PITH_FULL_IMAGE:figures/ful… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references · 25 canonical work pages

  1. [1]

    Brain-wide neuronal dynamics during motor adaptation in zebrafish.Nature2012,485, 471–477

    Ahrens, M.B.; Li, J.M.; Orger, M.B.; Robson, D.N.; Schier, A.F.; Engert, F.; Portugues, R. Brain-wide neuronal dynamics during motor adaptation in zebrafish.Nature2012,485, 471–477. https://doi.org/10.1038/nature11057

    doi:10.1038/nature11057

  2. [2]

    Whole-brain functional imaging at cellular resolution using light-sheet microscopy.Nature Methods2013,10, 413–420

    Ahrens, M.B.; Orger, M.B.; Robson, D.N.; Li, J.M.; Keller, P .J. Whole-brain functional imaging at cellular resolution using light-sheet microscopy.Nature Methods2013,10, 413–420. https://doi.org/10.1038/nmeth.2434

    work page doi:10.1038/nmeth.2434

  3. [3]

    Whole-Brain Activity Maps Reveal Stereotyped, Distributed Networks for Visuomotor Behavior.Neuron2014,81, 1328–1343

    Portugues, R.; Feierstein, C.E.; Engert, F.; Orger, M.B. Whole-Brain Activity Maps Reveal Stereotyped, Distributed Networks for Visuomotor Behavior.Neuron2014,81, 1328–1343. https://doi.org/10.1016/j.neuron.2014.01.019

    work page doi:10.1016/j.neuron.2014.01.019 2014

  4. [4]

    Pan-neuronal calcium imaging with cellular resolution in freely swimming zebrafish.Nature methods2017,14, 1107–1114

    Kim, D.H.; Kim, J.; Marques, J.C.; Grama, A.; Hildebrand, D.G.; Gu, W.; Li, J.M.; Robson, D.N. Pan-neuronal calcium imaging with cellular resolution in freely swimming zebrafish.Nature methods2017,14, 1107–1114. https://doi.org/10.1038/nmeth.4429

    work page doi:10.1038/nmeth.4429

  5. [5]

    Structure of the zebrafish locomotor repertoire revealed with unsupervised behavioral clustering.Current Biology2018,28, 181–195

    Marques, J.C.; Lackner, S.; Félix, R.; Orger, M.B. Structure of the zebrafish locomotor repertoire revealed with unsupervised behavioral clustering.Current Biology2018,28, 181–195

  6. [6]

    Functional and effective connectivity: a review.Brain Connectivity2011,1, 13–36

    Friston, K.J. Functional and effective connectivity: a review.Brain Connectivity2011,1, 13–36. https://doi.org/10.1089/brain.20 11.0008

    work page doi:10.1089/brain.20

  7. [7]

    Estimation of directed effective connectivity from fMRI functional connectivity hints at asymmetries of cortical connectome.PLOS Computational Biology2016,12, e1004762

    Gilson, M.; Moreno-Bote, R.; Ponce-Alvarez, A.; Ritter, P .; Deco, G. Estimation of directed effective connectivity from fMRI functional connectivity hints at asymmetries of cortical connectome.PLOS Computational Biology2016,12, e1004762. https: //doi.org/10.1371/journal.pcbi.1004762

    doi:10.1371/journal.pcbi.1004762

  8. [8]

    https://doi.org/10.1201/9781420010138

    Carroll, R.J.; Ruppert, D.; Stefanski, L.A.; Crainiceanu, C.M.Measurement Error in Nonlinear Models: A Modern Perspective, 2 ed.; Chapman and Hall/CRC, 2006. https://doi.org/10.1201/9781420010138

    doi:10.1201/9781420010138 2006

  9. [9]

    Imaging Calcium in Neurons.Neuron2012,73, 862–885

    Grienberger, C.; Konnerth, A. Imaging Calcium in Neurons.Neuron2012,73, 862–885. https://doi.org/10.1016/j.neuron.2012.02. 011

    work page doi:10.1016/j.neuron.2012.02 2012

  10. [10]

    Ultrasensitive fluorescent proteins for imaging neuronal activity.Nature2013,499, 295–300

    Chen, T.W.; Wardill, T.J.; Sun, Y.; Pulver, S.R.; Renninger, S.L.; Baohan, A.; Schreiter, E.R.; Kerr, R.A.; Orger, M.B.; Jayaraman, V .; et al. Ultrasensitive fluorescent proteins for imaging neuronal activity.Nature2013,499, 295–300. https://doi.org/10.1038/ nature12354

  11. [11]

    NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data.Journal of Neuroscience Methods2017,291, 83–94

    Pnevmatikakis, E.A.; Giovannucci, A. NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data.Journal of Neuroscience Methods2017,291, 83–94. https://doi.org/10.1016/j.jneumeth.2017.07.031

    work page doi:10.1016/j.jneumeth.2017.07.031 2017

  12. [12]

    Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data.Neuron2016,89, 285–299

    Pnevmatikakis, E.A.; Soudry, D.; Gao, Y.; Machado, T.A.; Merel, J.; Pfau, D.; Reardon, T.; Mu, Y.; Lacefield, C.; Yang, W.; et al. Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data.Neuron2016,89, 285–299. https: //doi.org/10.1016/j.neuron.2015.11.037

    work page doi:10.1016/j.neuron.2015.11.037 2015

  13. [13]

    CaImAn an open source tool for scalable calcium imaging data analysis.elife2019,8, e38173

    Giovannucci, A.; Friedrich, J.; Gunn, P .; Kalfon, J.; Brown, B.L.; Koay, S.A.; Taxidis, J.; Najafi, F.; Gauthier, J.L.; Zhou, P .; et al. CaImAn an open source tool for scalable calcium imaging data analysis.elife2019,8, e38173. https://doi.org/10.7554/eLife.38173. Version August 4, 2026 submitted toEntropy 11 of 18

    work page doi:10.7554/elife.38173 2026

  14. [14]

    Analysis pipelines for calcium imaging data.Current opinion in neurobiology2019,55, 15–21

    Pnevmatikakis, E.A. Analysis pipelines for calcium imaging data.Current opinion in neurobiology2019,55, 15–21. https: //doi.org/10.1016/j.conb.2018.11.004

    work page doi:10.1016/j.conb.2018.11.004 2018

  15. [15]

    Fluorescence imaging of large-scale neural ensemble dynamics.Cell2022,185, 9–41

    Kim, T.H.; Schnitzer, M.J. Fluorescence imaging of large-scale neural ensemble dynamics.Cell2022,185, 9–41

  16. [16]

    Calcium imaging and the curse of negativity.Frontiers in neural circuits2021, 14, 607391

    Vanwalleghem, G.; Constantin, L.; Scott, E.K. Calcium imaging and the curse of negativity.Frontiers in neural circuits2021, 14, 607391

  17. [17]

    Spirtes, P .; Glymour, C.N.; Scheines, R.; Heckerman, D.Causation, prediction, and search; MIT press, 2000

    2000

  18. [18]

    Pearl, J.Causality; Cambridge university press, 2009

    2009

  19. [19]

    Peters, J.; Janzing, D.; Schölkopf, B.Elements of causal inference: foundations and learning algorithms; The MIT Press, 2017

    2017

  20. [20]

    Causal inference for time series.Nature Reviews Earth & Environment2023,4, 487–505

    Runge, J.; Gerhardus, A.; Varando, G.; Eyring, V .; Camps-Valls, G. Causal inference for time series.Nature Reviews Earth & Environment2023,4, 487–505

  21. [21]

    The effect of filtering on Granger causality based multivariate causality measures.NeuroImage2010,50, 577–588

    Florin, E.; Gross, J.; Pfeifer, J.; Fink, G.R.; Timmermann, L. The effect of filtering on Granger causality based multivariate causality measures.NeuroImage2010,50, 577–588. https://doi.org/10.1016/j.neuroimage.2009.12.050

    work page doi:10.1016/j.neuroimage.2009.12.050 2009

  22. [22]

    Behaviour of Granger causality under filtering: Theoretical invariance and practical application.Journal of Neuroscience Methods2011,201, 404–419

    Barnett, L.; Seth, A.K. Behaviour of Granger causality under filtering: Theoretical invariance and practical application.Journal of Neuroscience Methods2011,201, 404–419. https://doi.org/10.1016/j.jneumeth.2011.08.010

    work page doi:10.1016/j.jneumeth.2011.08.010 2011

  23. [23]

    Granger Causality Analysis in Neuroscience and Neuroimaging.Journal of Neuroscience2015, 35, 3293–3297, [https://www.jneurosci.org/content/35/8/3293.full.pdf]

    Seth, A.K.; Barrett, A.B.; Barnett, L. Granger Causality Analysis in Neuroscience and Neuroimaging.Journal of Neuroscience2015, 35, 3293–3297, [https://www.jneurosci.org/content/35/8/3293.full.pdf]. https://doi.org/10.1523/JNEUROSCI.4399-14.2015

    work page doi:10.1523/jneurosci.4399-14.2015 2015

  24. [24]

    Granger causality analysis of fMRI BOLD signals is invariant to hemodynamic convolution but not downsampling.NeuroImage2013,65, 540–555

    Seth, A.K.; Chorley, P .; Barnett, L.C. Granger causality analysis of fMRI BOLD signals is invariant to hemodynamic convolution but not downsampling.NeuroImage2013,65, 540–555. https://doi.org/10.1016/j.neuroimage.2012.09.049

    work page doi:10.1016/j.neuroimage.2012.09.049 2012

  25. [25]

    The influence of filtering and downsampling on the estimation of transfer entropy.PLOS ONE2017,12, e0188210

    Weber, I.; Florin, E.; von Papen, M.; Timmermann, L. The influence of filtering and downsampling on the estimation of transfer entropy.PLOS ONE2017,12, e0188210. https://doi.org/10.1371/journal.pone.0188210

    work page doi:10.1371/journal.pone.0188210

  26. [26]

    Spatio-temporal Granger causality: A new framework

    Luo, Q.; Lu, W.; Cheng, W.; Valdes-Sosa, P .A.; Wen, X.; Ding, M.; Feng, J. Spatio-temporal Granger causality: A new framework. NeuroImage2013,79, 241–263. https://doi.org/10.1016/j.neuroimage.2013.04.091

    work page doi:10.1016/j.neuroimage.2013.04.091 2013

  27. [27]

    Analysis of sampling artifacts on the Granger causality analysis for topology extraction of neuronal dynamics.Frontiers in Computational Neuroscience2014,8, 75

    Zhou, D.; Zhang, Y.; Xiao, Y.; Cai, D. Analysis of sampling artifacts on the Granger causality analysis for topology extraction of neuronal dynamics.Frontiers in Computational Neuroscience2014,8, 75. https://doi.org/10.3389/fncom.2014.00075

    arXiv 2014

  28. [28]

    Detectability of Granger causality for subsampled continuous-time neurophysiological processes.Journal of Neuroscience Methods2017,275, 93–121

    Barnett, L.; Seth, A.K. Detectability of Granger causality for subsampled continuous-time neurophysiological processes.Journal of Neuroscience Methods2017,275, 93–121. https://doi.org/10.1016/j.jneumeth.2016.10.016

    work page doi:10.1016/j.jneumeth.2016.10.016 2016

  29. [29]

    Causal discovery from temporally aggregated time series

    Gong, M.; Zhang, K.; Schölkopf, B.; Glymour, C.; Tao, D. Causal discovery from temporally aggregated time series. In Proceedings of the Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, 2017, p. 269

    2017

  30. [30]

    A constraint optimization approach to causal discovery from subsampled time series data.International Journal of Approximate Reasoning2017,90, 208–225

    Hyttinen, A.; Plis, S.; Järvisalo, M.; Eberhardt, F.; Danks, D. A constraint optimization approach to causal discovery from subsampled time series data.International Journal of Approximate Reasoning2017,90, 208–225. https://doi.org/10.1016/j.ijar.2017 .07.009

    work page doi:10.1016/j.ijar.2017 2017

  31. [31]

    Rate-Agnostic (Causal) Structure Learning

    Plis, S.; Danks, D.; Freeman, C.; Calhoun, V . Rate-Agnostic (Causal) Structure Learning. In Proceedings of the Advances in Neural Information Processing Systems, 2015, Vol. 28

    2015

  32. [32]

    Fast and Robust Fixed-Point Algorithms for Independent Component Analysis.IEEE Transactions on Neural Networks1999,10, 626–634

    Hyvarinen, A. Fast and Robust Fixed-Point Algorithms for Independent Component Analysis.IEEE Transactions on Neural Networks1999,10, 626–634. https://doi.org/10.1109/72.761722

    doi:10.1109/72.761722

  33. [33]

    An Information-Maximization Approach to Blind Separation and Blind Deconvolution.Neural Computation1995,7, 1129–1159

    Bell, A.J.; Sejnowski, T.J. An Information-Maximization Approach to Blind Separation and Blind Deconvolution.Neural Computation1995,7, 1129–1159. https://doi.org/10.1162/neco.1995.7.6.1129

    doi:10.1162/neco.1995.7.6.1129 1995

  34. [34]

    A blind source separation technique using second-order statistics

    Belouchrani, A.; Abed-Meraim, K.; Cardoso, J.F.; Moulines, E. A blind source separation technique using second-order statistics. IEEE Transactions on Signal Processing1997,45, 434–444. https://doi.org/10.1109/78.554307

    work page doi:10.1109/78.554307

  35. [35]

    Blind Beamforming for Non Gaussian Signals.IEE Proceedings F: Radar and Signal Processing1993, 140, 362–370

    Cardoso, J.F.; Souloumiac, A. Blind Beamforming for Non Gaussian Signals.IEE Proceedings F: Radar and Signal Processing1993, 140, 362–370. https://doi.org/10.1049/ip-f-2.1993.0054

    arXiv 1993

  36. [36]

    Independent Component Analysis: Algorithms and Applications.Neural Networks2000,13, 411–430

    Hyvarinen, A.; Oja, E. Independent Component Analysis: Algorithms and Applications.Neural Networks2000,13, 411–430. https://doi.org/10.1016/S0893-6080(00)00026-5

    doi:10.1016/s0893-6080(00)00026-5

  37. [37]

    Independent component analysis at the neural cocktail party.Trends in Neurosciences 2001,24, 54–63

    Brown, G.D.; Yamada, S.; Sejnowski, T.J. Independent component analysis at the neural cocktail party.Trends in Neurosciences 2001,24, 54–63. https://doi.org/10.1016/S0166-2236(00)01683-0

    work page doi:10.1016/s0166-2236(00)01683-0 2001

  38. [38]

    BSS and ICA in Neuroinformatics: From Current Practices to Open Challenges.IEEE Reviews in Biomedical Engineering2008,1, 50–61

    Vigario, R.; Oja, E. BSS and ICA in Neuroinformatics: From Current Practices to Open Challenges.IEEE Reviews in Biomedical Engineering2008,1, 50–61. https://doi.org/10.1109/RBME.2008.2008244

    arXiv 2008

  39. [39]

    Automated Analysis of Cellular Signals from Large-Scale Calcium Imaging Data

    Mukamel, E.A.; Nimmerjahn, A.; Schnitzer, M.J. Automated Analysis of Cellular Signals from Large-Scale Calcium Imaging Data. Neuron2009,63, 747–760. https://doi.org/10.1016/j.neuron.2009.08.009

    work page doi:10.1016/j.neuron.2009.08.009 2009

  40. [40]

    Granger causality analysis for calcium transients in neuronal networks, challenges and improvements.eLife2023,12, e81279

    Chen, X.; Ginoux, F.; Carbo-Tano, M.; Mora, T.; Walczak, A.M.; Wyart, C. Granger causality analysis for calcium transients in neuronal networks, challenges and improvements.eLife2023,12, e81279. https://doi.org/10.7554/eLife.81279

    work page doi:10.7554/elife.81279

  41. [41]

    Fallani, F.D.V .; Corazzol, M.; Sternberg, J.R.; Wyart, C.; Chavez, M. Hierarchy of neural organization in the embryonic spinal cord: Granger-causality graph analysis of in vivo calcium imaging data.IEEE Transactions on Neural Systems and Rehabilitation Engineering2014,23, 333–341. https://doi.org/10.1109/TNSRE.2014.2341632. Version August 4, 2026 submitt...

    arXiv 2014

  42. [42]

    Functional coupling of the mesencephalic locomotor region and v2a reticulospinal neurons driving forward locomotion.bioRxiv2022, pp

    Carbo-Tano, M.; Lapoix, M.; Jia, X.; Auclair, F.; Dubuc, R.; Wyart, C. Functional coupling of the mesencephalic locomotor region and v2a reticulospinal neurons driving forward locomotion.bioRxiv2022, pp. 2022–04. https://doi.org/10.1101/2022.04.01.486 703

    work page doi:10.1101/2022.04.01.486 2022

  43. [43]

    To Deconvolve, or Not to Deconvolve: Inferences of Neuronal Activities using Calcium Imaging Data.arXiv preprint arXiv:2103.021632021, [arXiv:q-bio.NC/2103.02163]

    Shen, T.; Lur, G.; Xu, X.; Yu, Z. To Deconvolve, or Not to Deconvolve: Inferences of Neuronal Activities using Calcium Imaging Data.arXiv preprint arXiv:2103.021632021, [arXiv:q-bio.NC/2103.02163]

    Pith/arXiv arXiv

  44. [44]

    Welch, P .D. The Use of Fast Fourier Transform for the Estimation of Power Spectra: A Method Based on Time Averaging Over Short, Modified Periodograms.IEEE Transactions on Audio and Electroacoustics1967,15, 70–73. https://doi.org/10.1109/TAU.19 67.1161901

    work page doi:10.1109/tau.19

  45. [45]

    Least Squares Quantization in PCM.IEEE Transactions on Information Theory1982,28, 129–137

    Lloyd, S.P . Least Squares Quantization in PCM.IEEE Transactions on Information Theory1982,28, 129–137. https://doi.org/10.110 9/TIT.1982.1056489

    arXiv 1982

  46. [46]

    BunDLe-Net: Neuronal Manifold Learning Meets Behaviour.bioRxiv2023

    Kumar, A.; Gilra, A.; Gonzalez-Soto, M.; Meunier, A.; Grosse-Wentrup, M. BunDLe-Net: Neuronal Manifold Learning Meets Behaviour.bioRxiv2023. https://doi.org/10.1101/2023.08.08.551978

    work page doi:10.1101/2023.08.08.551978 2023

  47. [47]

    Neuro-cognitive multilevel causal modeling: A framework that bridges the explanatory gap between neuronal activity and cognition.PLOS Computational Biology2024,20, e1012674

    Grosse-Wentrup, M.; Kumar, A.; Meunier, A.; Zimmer, M. Neuro-cognitive multilevel causal modeling: A framework that bridges the explanatory gap between neuronal activity and cognition.PLOS Computational Biology2024,20, e1012674. https: //doi.org/10.1371/journal.pcbi.1012674

    doi:10.1371/journal.pcbi.1012674

  48. [48]

    Probabilistic Principal Component Analysis.Journal of the Royal Statistical Society Series B: Statistical Methodology1999,61, 611–622

    Tipping, M.E.; Bishop, C.M. Probabilistic Principal Component Analysis.Journal of the Royal Statistical Society Series B: Statistical Methodology1999,61, 611–622. https://doi.org/10.1111/1467-9868.00196

    arXiv

  49. [49]

    Assessing and tuning brain decoders: Cross-validation, caveats, and guidelines.NeuroImage2017,145, 166–179

    Varoquaux, G.; Raamana, P .R.; Engemann, D.A.; Hoyos-Idrobo, A.; Schwartz, Y.; Thirion, B. Assessing and tuning brain decoders: Cross-validation, caveats, and guidelines.NeuroImage2017,145, 166–179. https://doi.org/10.1016/j.neuroimage.2016.10.038

    doi:10.1016/j.neuroimage.2016.10.038 2016

  50. [50]

    Machine learning algorithm validation with a limited sample size.PLOS ONE 2019,14, e0224365

    Vabalas, A.; Gowen, E.; Poliakoff, E.; Casson, A.J. Machine learning algorithm validation with a limited sample size.PLOS ONE 2019,14, e0224365. https://doi.org/10.1371/journal.pone.0224365

    doi:10.1371/journal.pone.0224365 2019

  51. [51]

    On Over-fitting in Model Selection and Subsequent Selection Bias in Performance Evaluation.Journal of Machine Learning Research2010,11, 2079–2107

    Cawley, G.C.; Talbot, N.L.C. On Over-fitting in Model Selection and Subsequent Selection Bias in Performance Evaluation.Journal of Machine Learning Research2010,11, 2079–2107

    2079

  52. [52]

    Re-examining Granger Causality with Causal Bayesian Networks and Reichenbachs Principles, 2025, [arXiv:stat.ML/2501.02672]

    Adedayo, S.A. Re-examining Granger Causality with Causal Bayesian Networks and Reichenbachs Principles, 2025, [arXiv:stat.ML/2501.02672]. arXiv:2501.02672, revised May 2026

    Pith/arXiv arXiv 2025

  53. [53]

    Detecting and quantifying causal associations in large nonlinear time series datasets.Science advances2019,5, eaau4996

    Runge, J.; Nowack, P .; Kretschmer, M.; Flaxman, S.; Sejdinovic, D. Detecting and quantifying causal associations in large nonlinear time series datasets.Science advances2019,5, eaau4996

  54. [54]

    Causal Discovery for Time Series from Multiple Datasets with Latent Contexts

    Günther, W.; Ninad, U.; Runge, J. Causal Discovery for Time Series from Multiple Datasets with Latent Contexts. In Proceedings of the Proceedings of the Thirty-Ninth Conference on Uncertainty in Artificial Intelligence; Evans, R.J.; Shpitser, I., Eds. PMLR, 2023, Vol. 216,Proceedings of Machine Learning Research, pp. 766–776

    2023

  55. [55]

    Conditioning-Depth Diagnostics for Hidden Memory in Temporal Causal Discovery.arXiv preprint arXiv:2606.012142026, [arXiv:stat.AP/2606.01214]

    Adedayo, S.A. Conditioning-Depth Diagnostics for Hidden Memory in Temporal Causal Discovery.arXiv preprint arXiv:2606.012142026, [arXiv:stat.AP/2606.01214]

    Pith/arXiv arXiv

  56. [56]

    Structural and effective brain connectivity in focal epilepsy.NeuroImage: Reports2025,5, 100274

    Jelsma, S.B.; Zijlmans, M.; Heijink, I.B.; Hoefnagels, F.W.A.; Raemaekers, M.; Otte, W.M.; van Klink, N.E.C.; van Blooijs, D. Structural and effective brain connectivity in focal epilepsy.NeuroImage: Reports2025,5, 100274. https://doi.org/10.1016/j.ynirp. 2025.100274

    arXiv 2025

  57. [57]

    INTENSE: Detecting and disentangling neuronal selectivity in calcium imaging data.arXiv preprint arXiv:2603.046222026, [arXiv:q-bio.NC/2603.04622]

    Pospelov, N.; Plusnin, V .; Rogozhnikova, O.; Ivanova, A.; Sotskov, V .; Toropova, K.; Ivashkina, O.; Avetisov, V .; Anokhin, K. INTENSE: Detecting and disentangling neuronal selectivity in calcium imaging data.arXiv preprint arXiv:2603.046222026, [arXiv:q-bio.NC/2603.04622]

    arXiv

  58. [58]

    NeuroGraph: Benchmarks for Graph Machine Learning in Brain Connectomics.Advances in Neural Information Processing Systems2023,36, 65073–65097, [arXiv:cs.LG/2306.06202]

    Said, A.; Bayrak, R.G.; Derr, T.; Shabbir, M.; Moyer, D.; Chang, C.; Koutsoukos, X. NeuroGraph: Benchmarks for Graph Machine Learning in Brain Connectomics.Advances in Neural Information Processing Systems2023,36, 65073–65097, [arXiv:cs.LG/2306.06202]

    Pith/arXiv arXiv

  59. [59]

    Jacobi Angles for Simultaneous Diagonalization.SIAM Journal on Matrix Analysis and Applications 1996,17, 161–164

    Cardoso, J.F.; Souloumiac, A. Jacobi Angles for Simultaneous Diagonalization.SIAM Journal on Matrix Analysis and Applications 1996,17, 161–164. https://doi.org/10.1137/S0895479893259546

    work page doi:10.1137/s0895479893259546 1996

  60. [60]

    Non-negative Matrix Factorization with Sparseness Constraints.Journal of Machine Learning Research2004,5, 1457– 1469

    Hoyer, P .O. Non-negative Matrix Factorization with Sparseness Constraints.Journal of Machine Learning Research2004,5, 1457– 1469. Version August 4, 2026 submitted toEntropy 13 of 18 Supplementary Materials S1. Motorneuron motivation and CSL sensitivity The motorneuron recording is used only as an inspectable motivation and sensitivity check, not as a mai...

    2026

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.