henhaobio
Sign in

bioRxivNeurosciencePreprint

Fast-ripples are emergent properties of neuronal networks

Laurent SheybaniiD, Yichen QiuiD, Prince Kumar SinghiD, Umesh VivekanandaiD, Neil BurgessiD, Beate DiehliD, Andrew W. McEvoyiD, Anna MiserocchiiD, James A. BisbyiD, Tawfeeq Shekh‐AhmadiD, Gabriele LignaniiD, Daniel BushiD, Matthew C. WalkeriD

PDFDOICode
In-depth full-text digest available · Jump to digest ↓

Abstract

Fast-ripples (250-500 Hz) have been proposed as a promising biomarker in epilepsy, but their specificity remains unclear. In particular, it is uncertain whether they reflect chance coincident neural activity or distinctly generated pathological entities. We combined in silico simulations, neuronal cultures, the intra-hippocampal kainate rat model of hippocampal epilepsy, and human microwire recordings to investigate whether fast-ripples occur more frequently than expected by the chance aggregation of action potentials. Our simulations showed that chance aggregation can generate fast-ripples and that their incidence changes depending on brain state, an observation that we confirmed in our rodent data. The likelihood of exceeding chance depended on biological system complexity and vigilance state: fast-ripples in neuronal cultures did not surpass chance levels, whereas those in awake - but not sleeping - rodents did. Similarly, the incidence of fast-ripples in awake human recordings was ~30% greater than expected by chance. As such, our findings suggest that most fast-ripples reflect stochastic network activity rather than distinctly generated pathological entities. This work does not rule out FRs as potential indicators of epileptogenic tissue, but it does challenge prevailing assumptions about their generation and specificity. Their higher prevalence in epileptogenic tissue is likely primarily due to increased excitation and/or neural synchronization, rather than peculiar abnormalities in network behavior.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (10 Sept 2026) · CC BY 4.0

Most fast-ripples (250-500 Hz) in rat and human recordings are consistent with chance clustering of action potentials; in awake patients, distinct events make up an estimated 37.5% of detections.

Fast-ripples are proposed as markers of seizure-generating tissue, but it is unresolved whether they are distinct pathological events or coincident firing merely more common in epileptic networks. Prior work linked them to action-potential sequences; this study tests the chance explanation with simulations of randomly inserted action potentials and, in neuronal cultures, kainate-treated rats and human microwires, with a chance baseline counted in recordings whose high-frequency band is shuffled.

Key results

  1. ShownIn awake human mesial temporal microwire recordings, fast-ripples (FRs) were more frequent in original than in band-shuffled signals, but the excess implies only about 37.5% of detected FRs are distinct from chance; sampling frequency did not alter this.

    Linear mixed model, 10 patients, data resampled at 3,000 to 15,000 Hz: signal type F(1,770.4)=20.62, p<0.001; sampling frequency F(1,13.2)=0.47, p=0.51; interaction F(1,770.4)=0.05, p=0.82. Original signals had ~1.6 times the FRs of shuffled ones (3 distinct per 5 chance). Simpler shuffling methods found no significant excess. · The majority of FRs also occur by chance during…; Patient recordings and preprocessing; Statistical analyses; Figure legends

  2. ShownIn kainate-treated rats, the ratio of FR rate in original to shuffled electrocorticography (ECoG) exceeded 1 after kainate only at delta-power troughs (presumed wake), and did not exceed 1 before kainate at either delta phase.

    8 epileptic rats; kainate x delta interaction F(1,7)=17.1, p=0.004. After kainate: 1.6 [1.3-2] at troughs (adjusted p=0.006) vs 1.1 [0.7-1.5] at peaks (adjusted p>0.99); troughs vs peaks p=0.0003. Before kainate: 1.5 [1-2] at troughs (adjusted p=0.09), 1.3 [0.8-1.9] at peaks (adjusted p=0.41). · The relative proportion of distinct and chance…; ECoG data acquisition and Analysis; Figure legends

  3. ShownAfter kainate, absolute FR rate in rats rose only at delta troughs (presumed wake), while the original:shuffled ratio at troughs did not change; the authors read this as increased network excitability rather than new distinct FRs.

    Kainate x delta interaction on FR rate F(1,7)=14.7, p=0.006; increase at troughs 12.6 min^-1 [9.0 to 16.3] (adjusted p<0.001) vs 4.2 min^-1 [0.5 to 7.9] at peaks (adjusted p=0.12). Ratio change at troughs -0.11 [-0.25 to 0.03], adjusted p=0.41. · Diurnal variation of FR rate as a confound in…; The relative proportion of distinct and chance…

  4. SuggestedRandomly inserted action potentials produced FRs in silico, with counts structured by neuron number, firing rate and synchrony; constraining only firing (lower) and synchrony (higher) during sleep reproduced a weak but significant daily FR rhythm matching rat data.

    Gradient boosted tree over 125,000 parameter sets: predicted vs simulated FR counts r^2=0.73, p<0.0001; permuted parameters R^2 0.02 ± 0.001 (p=0.002); 80/20 cross-validation beat shuffled outputs (p<0.0001). Rat fit: median r^2 0.01 [0.004-0.04] vs 0.0009 [0.0008-0.001] for shuffled predictions (p=0.016); high/low incidence AUC 0.58 ± 0.05 (p=0.003). · FRs can occur as emergent phenomena in silico; Emergent FRs index neuronal population, firing…; FR rate is predicted by neural firing rate and…; Gradient boosted tree

  5. SuggestedIn mouse cortical cultures on multi-electrode arrays, FR counts did not exceed those of shuffled signals, either at baseline or 24 h after the GABA-A receptor antagonist picrotoxin.

    Two-way ANOVA: shuffling F(1,15)=0.99, p=0.33; picrotoxin F(1,15)=1.53, p=0.24; interaction F(1,15)=1.11, p=0.31. The number of wells or plates analysed is not stated in the text. · FR rate is at chance level in neuronal cultures; Mouse primary neuronal culture recording with…

Methods, data and reproducibility

Action potentials were inserted into a human microwire background across 125,000 neuron-count, firing-rate and synchrony settings, and a gradient boosted tree related these to fast-ripple (FR) counts. Delphos-detected FRs in neuronal cultures on multi-electrode arrays, week-long rat electrocorticography and human microwires were compared with counts in wavelet-IAAFT surrogates shuffling only >200 Hz content.

  • Simulated signals: action potentials (mean waveform from Keles et al.) inserted into one human microwire recording over a 50 x 50 x 50 grid of neuron count, firing rate and synchrony · human (background microwire signal) · n = 125,000 simulations
  • Embryonic (E18-19) cortical neuronal cultures on Axion 24-well multi-electrode arrays (16 electrodes per well), recorded at baseline and 24 h after 50 μM picrotoxin · mouse (C57BL/6J)
  • Continuous wireless epidural ECoG (2048 Hz) over right dorsal hippocampus, before and after kainic acid-induced status epilepticus · rat (male Sprague-Dawley) · n = 10 implanted, 8 epileptic analysed; c. 11,905 hours of recording
  • Awake intracranial Behnke-Fried microwire recordings from mesial temporal lobe of presurgical epilepsy patients (30 kHz, Blackrock Neuroport) · human · n = 10 patients; median 4.5 electrodes; median recording 10 min 17 s
Methods & tools
Delphos detector, wavelet-IAAFT surrogate shuffling, MATLAB (fitrensemble LSBoost, pwelch, cwt), Axion Mastro Edge MEA system and Axion Navigator, Neural Metric Tool (Axion), Open-Source Instruments A3049T5 ECoG transmitter, Neuroarchiver, Blackrock Neuroport, Behnke-Fried microwire electrodes, Jamovi 2.6.22 with GAMLj3 3.4.2, picrotoxin, kainic acid
Software & environment
Mixed models in Jamovi 2.6.22 with GAMLj3 3.4.2; MEA acquisition with Mastro Edge 3.7.1.16; other analyses in MATLAB, version not stated. No container or dependency file described.
Reproducibility gaps
Culture (MEA) and rat data available on reasonable request; clinical data on request within the ethics terms. Analysis code was partly written with ChatGPT and Copilot; one surrogate method (AR) was written in-house.
Limitations (6)
  • The authors state that conclusions could depend on the temporal shuffling method; they tested several alternatives, and simpler shuffles not restricted to the FR band found no excess of FRs in the original data. · stated by the authors
  • The relationship of FRs to the epileptogenic zone was not assessed, so the work does not speak for or against their clinical use as biomarkers. · stated by the authors
  • Rat FR detections (c. 11,905 hours) could not be checked by eye, and sleep and wake were inferred from delta power on the assumption that high delta reflects sleep (cross-checked with the beta/delta ratio). · stated by the authors
  • Human data were short awake recordings (median 10 min 17 s); the authors note that, given how fast-ripples vary over time, longer nocturnal recordings in humans are necessary. · stated by the authors
  • Internal inconsistencies: the abstract reports a ~30% excess of human FRs over chance and an intra-hippocampal kainate model, whereas the Results give ~1.6 times and the Methods a single intraperitoneal kainic acid injection. · noted in this summary
  • The claim that only epileptic rats exceed chance rests on one-sample tests either side of significance (1.5, adjusted p=0.09 before vs 1.6, adjusted p=0.006 after, at delta troughs); the direct before/after comparison at troughs was not significant (adjusted p=0.41). · noted in this summary

Read this if you detect high-frequency oscillations for epilepsy surgery or research and want a surrogate-shuffling baseline for how many fast-ripples exceed chance action-potential clustering.

Citations
2

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISCurot '22Gavenas '24Maslarova '25Lippmann '22Tong '21Wen '26
75% semantic matchNeuroscience
Jonathan Curot, Emmanuel J. Barbeau, Elodie Despouy+5 more·Brain·Sep 11, 2022· 33 citations