Screening brain organoids for their information-processing potential

Complex systems like our brains, when close to criticality (1), become versatile information processors (2). Can neural organoids become near-critical and thus useful models of biological information processing? If proven to be near-critical, neural organoids could become a long-term platform for studying biological computation. I propose an open-source benchmark with curated datasets and reproducible code along with a proof-of-concept distributed data acquisition pipeline to investigate criticality in neural organoids.

Background and Aims

Neural organoids are three-dimensional brain tissue primarily used as cytoarchitectural models of different brain regions (3). Recent work has begun studying neural network dynamics in neural organoids. This research suggests human neural organoids, despite lacking inputs, self‑organize into neuronal networks and exhibit electrical activity that may approach near-critical dynamics, a type of emergent property of complex systems (4,5). Many otherwise unrelated complex systems ranging from snow avalanches to human social networks to artificial and biological neural networks show emergent criticality. (see ref. 6 for a shorter, succinct review; ref.7 for a comprehensive survey of brain criticality). Complex systems when at or near their critical point, exhibit highly unusual properties like scale invariance (aka scale-free), extreme responsiveness, long-range correlations, maximal dynamic range in response to different inputs, maximal information transfer and storage amongst others. These properties are associated with favorable information-processing capabilities, as has been demonstrated in both artificial as well as brain neuronal networks (7). Over the past two decades, criticality has been demonstrated in many different species ranging from leeches to several mammals including humans (7). If organoid neural networks do become near- or quasi-critical as suggested by the recent work (4,5), it raises the prospect of using them to carry out useful computation. This concept has been called organoid intelligence (8), and has been suggested as an eventual energy-efficient alternative to current AI systems. Equally importantly, OI may also help clarify how biological intelligence differs from AI. Furthermore, studies using neural organoids as models of brain development might also benefit given criticality appears to be a favorable dynamical regime for brains, and has been proposed as a potential developmental endpoint (7). An open source, documented, reproducible, benchmarking workflow to establish near-criticality in neural organoids could facilitate these goals. Over the last two decades, the brain criticality field has identified pitfalls related to recording and analysis methodology which lead to controversial conclusions regarding criticality in brains (7). Therefore, our proposed workflow for organoids is based on state-of-the-art criticality measures emerging from the brain criticality field (6,7). In turn, in vitro organoids could bring their unique advantages to advance that field. For example, the existence of mechanisms that homeostatically control criticality has been demonstrated in the mammalian visual cortex (9) but are hard to investigate precisely because of homeostasis in vivo. Identifying bottleneck biological processes (control parameters) that underpin organoid criticality could be catalyzed by relatively rapid in vitro screening using our proposed criticality benchmark which, in turn, could lead to manipulating criticality in vivo.

Hypothesis

Mammalian neural organoid neuronal networks are capable of becoming near-critical as assessed using multiple criticality metrics and these dynamics can be potentially tuned through yet-to-be-discovered biological mechanisms.

Specific Aims

Aim 1: Benchmark data sufficiency and signal-quality requirements for robust estimation of criticality-related metrics using open-source code and curated public datasets. We will develop an open, versioned analysis framework that quantifies how recording duration, unit count, channel count, spike-sorting quality, subsampling, and temporal binning affect the stability and interpretability of major criticality-related metrics. Using publicly available in vivo cortical datasets and published neural organoid datasets, we will compare results across sorted spikes, pooled multi-unit activity, and population-level signals. Outputs will include documented code, example notebooks, and a practical decision rubric classifying datasets as sufficient, borderline, or insufficient for different categories of criticality claims.

Aim 2: Build and test a proof-of-concept standardized acquisition workflow for neural organoid criticality analysis. We will work with an electrophysiology technology partner, Diagnostic Biochips Inc., with whose recording technology (10) I have hands-on experience. We will define a standardized workflow for sample metadata, recording, cloud data return, quality control, and downstream open analysis. An executive of the company has indicated their interest in collaborating on this project after reading a version of this proposal. Where feasible within the project period, we will apply this workflow to at least one pilot organoid dataset obtained through a collaborating laboratory, compatible recording pipeline, or pre-existing compatible dataset. The goal is to produce a reusable acquisition-and-analysis pathway, not to make a definitive biological generalization from one preparation. Although we will collaborate with only one technology provider for this pilot, the distributed pipeline is intended to work with other providers and multiple collaborating laboratories.

Methods

Aim 1: we will curate public extracellular electrophysiology datasets spanning different recording durations, densities, and signal qualities. Candidate reference datasets include either the Allen Visual Coding (DANDI:000022) or the OpenScope datasets (DANDI:000253), along with a published neural organoid dataset (ref. 5; DANDI:001603). For each dataset, we will derive multiple signal representations where possible: spike-sorted units, pooled multi-unit activity, and population-level event summaries. We will then perform controlled perturbations in silico by varying recording duration, channel count, unit count, subsampling fraction, and temporal binning.

Aim 2: we will define a minimal standardized workflow covering sample metadata, recording conditions, raw data return, QC, and benchmark analysis. The workflow will be designed for portability and transparency. We will then test it on one pilot dataset to show how the benchmark can be applied to judge dataset adequacy. The point of this pilot is to demonstrate operational viability and interoperability between collaborators, not to complete a large consortium study within one year.

Procedures and Analyses

For each dataset and perturbation condition in Aim 1, we will estimate criticality-related metrics including avalanche size and duration statistics, branching-related measures, distance-from-criticality coefficient (DCC), and long-range temporal correlation measures. We will define adequacy using preregistered thresholds for metric stability under repeated subsampling and for agreement across signal representations where applicable. Stability will be quantified under repeated resampling and subsampling. The central analysis question is not simply whether a metric can be computed, but whether it remains interpretable and robust under realistic data limitations.

The main outputs of the statistical analysis will be:

a mapping from dataset properties to metric stability,
a practical adequacy rubric for criticality-related claims, and
an example application of this rubric to an organoid dataset.

The dataset in Aim 2 will be analyzed using the code and QC framework developed in Aim 1, ensuring that the acquisition workflow and analysis framework are linked.

Expected Outcomes and Impact

By the end of 12 months, this project will deliver:

an openly released framework for assessing whether extracellular neural recordings are adequate for criticality analysis,
documented code and example notebooks,
a standardized organoid acquisition and QC workflow, and
at least one pilot dataset or equivalent demonstration analyzed with the benchmarked framework.

Timeline

Total duration expected: 12 months.

Months 1 and 2: Finalize metric panel, curate public datasets, establish GitHub and cloud infrastructure, begin Benchmark v1 implementation.
Months 2 through 6: Run benchmark analyses across curated datasets; quantify effects of recording duration, signal class, spike-sorting quality, and subsampling; draft adequacy rubric. Months 6 through 9: Public release of Benchmark v1; finalize metadata template, QC checklist, and acquisition pipeline; coordinate pilot workflow logistics.
Months 9 through 12: Execute feasibility demonstration, analyze pilot dataset with benchmarked code, and release workflows, documentation, and example outputs.

Ethics

This project does not involve human participants or collection of new human-subject data by the applicant. Aim 1 analyzes public datasets. Aim 2 focuses on benchmarking and standardized acquisition workflows for neural organoid electrophysiology and, where applicable, on de-identified pilot data generated under collaborating institutions’ existing oversight, donor-consent frameworks, and applicable policies. The applicant will not access identifiable private information.

PI Bio

I am a neuroscientist and open-science developer with over two decades of experience in cellular, synaptic, and extracellular electrophysiology, including neural organoid recordings and quantitative neural data analysis. My work includes open-source analysis using SpikeInterface, Kilosort4, and DANDI-based neural datasets. This combination of experimental and computational experience positions me well to lead a reproducible, open-source project on benchmarked criticality analysis in neural organoids

Budget

Total: $ 25,000

Compute resources and cloud services: $4,000
Open-source software development, documentation, and public release: $5,000
Project operations and coordination: $4,000
Pilot acquisition workflow; consumables and recording costs: $10,000
Final release and dissemination: $2,000

References

Beggs, John M., and Dietmar Plenz. “Neuronal Avalanches in Neocortical Circuits.” Journal of Neuroscience 23, no. 35 (2003): 11167–77.
Shew, Woodrow L., Hongdian Yang, Shan Yu, Rajarshi Roy, and Dietmar Plenz. “Information Capacity and Transmission Are Maximized in Balanced Cortical Networks with Neuronal Avalanches.” Articles. Journal of Neuroscience 31, no. 1 (2011): 55–63.
Levy, Rebecca J., and Sergiu P. Paşca. “From Organoids to Assembloids: Experimental Approaches to Study Human Neuropsychiatric Disorders.” Annual Review of Neuroscience 48 (2025): 363–79.
Osaki, Tatsuya, Tomoya Duenki, Siu Yu A. Chow, et al. “Complex Activity and Short-Term Plasticity of Human Cerebral Organoids Reciprocally Connected with Axons.” Nature Communications 15, no. 1 (2024): 2945.
Molen, Tjitse van der, Alex Spaeth, Mattia Chini, et al. “Preconfigured Neuronal Firing Sequences in Human Brain Organoids.” Nature Neuroscience (2025): 1–13.
O’Byrne, Jordan, and Karim Jerbi. “How Critical Is Brain Criticality?” Trends in Neurosciences 45, no. 11 (2022): 820–37.
Hengen, Keith B., and Woodrow L. Shew. “Is Criticality a Unified Setpoint of Brain Function?” Neuron 113, no. 16 (2025): 2582-2598.e2.
Smirnova, Lena, Brian S. Caffo, David H. Gracias, et al. “Organoid Intelligence (OI): The New Frontier in Biocomputing and Intelligence-in-a-Dish.” Frontiers in Science 1 (February 2023).
Ma, Zhengyu, Gina G. Turrigiano, Ralf Wessel, and Keith B. Hengen. “Cortical Circuit Dynamics Are Homeostatically Tuned to Criticality In Vivo.” Neuron 104, no. 4 (2019): 655-664.e4.
Wu, Fan, Cuthbert Steadman, Liam Argent, et al. “High-Density Extracellular Recordings from the Interior of Intact Brain Organoids Enable Automated High-Throughput Functional Assay.” Preprint, bioRxiv, September 13, 2025.

Screening brain organoids for their information-processing potential

Screening brain organoids for their information-processing potential

Funding Progress

Peer Reviews

Related Funding

Request for Proposals - Project Gigabrain (Brain Organoids)

Nonprofit Recipient

Topics

DOI