henhaobio
Sign in

bioRxivNeurosciencePreprint

Development of py-mea-span for comprehensive analysis of multi-electrode array data and its application to LGI2-deficient neurons

Zhuang Zhuang HaniD, Janet Elwyn DeaneiD

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

Abstract

Multi-electrode arrays (MEAs) generate mid- to high-throughput electrophysiological data widely used to characterise neural network activity in vitro. Downstream analysis of recordings from Axion Biosystems instruments, however, has long depended on the manufacturers closed-source pipeline (AxIS Navigator, NeuralMetric Tool, AxIS Metric Plotting Tool), limiting customisation and integration into flexible Python workflows. We present py-mea-span, an open-source Python package providing end-to-end analysis of Axion Biosystems spk binary files. Our package enables spike extraction, inter-spike interval burst detection, network burst detection and spike synchrony. We also provide tools for preparation of publication-ready visualisations including raster plots with spike-detection-rate histograms, plate-layout heatmaps, time-course trajectories, single timepoint violin plots and statistical comparisons. Outputs reproduce the manufacturers NeuralMetric Tool reference with near perfect accuracy in all benchmarked metrics. A mea-span command-line interface enables batch processing without writing Python code. As a proof of concept, we applied py-mea-span to human iPSC-derived cortical neurons with CRISPRi-mediated LGI2 knockdown. This revealed a reproducible network hyperexcitability phenotype in which burst and network burst frequencies were elevated while spike synchrony remained unchanged. Availability and implementation The package is distributed on PyPI (pip install py-mea-span), and version-controlled at https://github.com/Molecularbiologyworld/py-mea-span. User feedback can be provided via the GitHub issues tab.

In-depth summary

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (25 Oct 2026) · CC BY 4.0

Open-source Python package parses Axion spk files directly and matches NeuralMetric Tool outputs, then reveals burst and network burst hyperexcitability, without synchrony change, in LGI2-knockdown human iPSC neurons.

Axion Biosystems recordings have been analysed with the manufacturer's closed-source NeuralMetric Tool, and the open-source MEA tools the paper lists need its tabular exports first. py-mea-span reads the spk binary files directly and adds statistics and plotting in one Python or command-line workflow.

Key results

  1. Shownpy-mea-span reproduced NeuralMetric Tool values for mean firing rate, number of active electrodes, per-electrode burst frequency and network burst frequency across the full DIV 12-44 range.

    R = 1.0000 for each of the four metrics, pooled across 946 well-observations covering all three conditions (lower n for burst and network-burst metrics, which include only wells where either tool detected such events). · py-mea-span reproduces NeuralMetric Tool outputs…; DISCUSSION

  2. ShownLGI2 knockdown (lines KD4 and KD5) raised burst and network burst frequency relative to scrambled control (SCRM) from DIV 24, and the elevation persisted to DIV 44.

    Divergence was abrupt from DIV 24 in both lines. At DIV 28, both lines were significantly higher than SCRM for burst and network burst frequency (Tukey HSD, n = 12 wells per condition; * p < 0.05, *** p < 0.001). The effect was seen in every batch on both plates. · LGI2 knockdown produces a progressive network…; Figure legends

  3. ShownSpike synchrony, measured as the spike time tiling coefficient (STTC), did not differ between knockdown and control, so the phenotype is raised excitability without changed synchrony structure.

    STTC increased at the same pace in SCRM, KD4 and KD5, levelling off near 0.9 from DIV 33. No difference between lines was seen even during DIV 24-30, before the ceiling was reached. · LGI2 knockdown produces a progressive network…

  4. SuggestedThe mean firing rate change was weaker than the burst phenotype and was clearest in KD5, the line with the more effective knockdown.

    Mean firing rate diverged gradually, most clearly in KD5. At DIV 28 the increase was smaller and more variable than for burst metrics. qPCR showed LGI2 mRNA reduced 85-90% in stem cells and 65-80% by DIV 14. · py-mea-span workflow and an LGI2 knockdown test…; LGI2 knockdown produces a progressive network…

  5. ShownThe package covers the pipeline from spike extraction to figures and statistics, with a command-line interface for batch use.

    Twenty-seven metrics; six modules (io, spike, burst, network, stats, viz); three pairwise tests (Tukey HSD, Mann-Whitney U with Bonferroni, Kruskal-Wallis with Dunn post hoc); 649 automated tests. · INTRODUCTION; Software implementation

Methods, data and reproducibility

The authors wrote a Python package that parses Axion spk binaries and computes spike, ISI-threshold burst, combined-ISI network burst and STTC metrics. They checked it against NeuralMetric Tool, then used it on 24-well MEA recordings of control and LGI2-knockdown CRISPRi iPSC-derived cortical neurons.

  • MEA recordings (24-well CytoView plates, 16 electrodes per well) of SCRM, LGI2 KD4 and LGI2 KD5 iPSC-derived cortical neurons, DIV 12-44 · human · n = 946 well-observations; 12 wells/condition; 3 biological replicates
  • qPCR of LGI2 mRNA in knockdown vs SCRM lines (stem cells, DIV 14, DIV 28); 5 knockdown lines generated, 2 taken forward · human · n = 3 technical (stem cells, DIV 14); 3 biological (DIV 28)
Methods & tools
py-mea-span, NeuralMetric Tool, AxIS Navigator, Maestro Edge, CytoView 24-well MEA plates, STTC, CRISPick, numpy, scipy, pandas, pingouin, statsmodels
Software & environment
Python 3.10 or later; depends on numpy, scipy, pandas, matplotlib, pingouin and statsmodels; 649 automated tests; installable with pip install py-mea-span.
Reproducibility gaps
No repository or accession is given for the raw LGI2 MEA recordings or the qPCR data.
Limitations (2)
  • The accuracy check compares against NeuralMetric Tool on four metrics (mean firing rate, active electrodes, burst frequency, network burst frequency) from the authors' own dataset; STTC and the other metrics are not compared with an independent reference. · noted in this summary
  • The LGI2 phenotype comes from two knockdown lines of a single iPSC background, with wells treated as independent observations for the statistics; no rescue experiment is reported, and the link to epilepsy models is described only as possibly consistent. · noted in this summary

Read this if you record Axion Maestro MEA data and want a scriptable Python pipeline that reproduces NeuralMetric Tool metrics and adds statistics and plots.

Citations
0

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
THISGelfman '18Wolff '25Bridges '18Hernandes '24Ren '26Mossink '21
71% semantic matchNeuroscience
Sahar Gelfman, Quanli Wang, Yifan Lu+14 more·PLoS Computational Biology·Oct 1, 2018· 35 citations