Organoid Intelligence (OI), an emerging field that applies neural interfacing to human brain organoids, is beginning to demonstrate measurable features of cognition in vitro.

28bio recently shared that three-dimensional neural models can display adaptive responses, learning-associated changes, and transcriptional profiles that reflect memory-related processes. The findings contribute to ongoing efforts to establish human-relevant functional biomarkers that supplement or replace traditional structural or molecular measurements.

‍

Structured Neural Models Derived from Human Cells

The organoids examined in these studies are generated from human neural progenitor cells and contain maturing populations of excitatory and inhibitory neurons supported by astrocytes. Their architecture gives rise to spontaneous electrical oscillations and coordinated bursts, reflecting the basic organization of developing neural networks. Each organoid is cultured around an embedded electrode array that allows continuous multichannel recording and targeted stimulation throughout growth.

This integrated format provides access to electrical behavior at multiple sites, enabling controlled induction of activity and systematic evaluation of how the network responds over time.

‍

Evidence of Learning and Long-Term Potentiation

A central component of the research examined whether these organoids could express long-term potentiation1 (LTP)-like changes in response to repeated, structured input. When subjected to patterned stimulation, trained organoids showed progressive strengthening of evoked responses, including increased spike count, more consistent spike bursts, and recruitment of additional spiking units. These changes developed gradually over repeated trials and were not observed in unstimulated controls.

The organoids’ responses proved highly sensitive to pharmacological modulation. Inhibiting glutamatergic signaling reduced potentiation and weakened evoked activity, confirming that these behaviors depend on canonical excitatory pathways. Complementary experiments tested the effects of brain-derived neurotrophic factor (BDNF), a molecule central to synaptic strengthening and learning. BDNF exposure increased bursting activity, enhanced responsiveness to stimulation, and shifted firing patterns in ways consistent with elevated plasticity.

‍

‍

‍

Together, these pharmacological results mirror the LTP-like changes observed during patterned training and reinforce that the organoids retain the molecular machinery necessary for activity-dependent synaptic modification. Rather than reflecting nonspecific electrical fluctuations, their behavior appears to rely on established biological pathways associated with long-term potentiation.

‍

Transcriptional Profiles Reflect Learning-Related Pathways

To complement the electrophysiological findings, researchers performed RNA sequencing on trained and untrained organoids. Trained organoids exhibited upregulation of genes linked to synaptic organization, neurotransmission, calcium signaling, and other pathways associated with plasticity and memory formation. Genes related to ion transport, vesicle cycling, and excitatory-inhibitory balance also showed increased expression.

In contrast, untrained organoids retained genetic profiles consistent with baseline developmental processes. The contrast between the two conditions provides additional evidence that structured stimulation induces systematic, biologically grounded adaptation rather than short-term electrical fluctuations.

‍

Adaptive Behavior in a Closed-Loop Environment

A closed-loop framework provided further confirmation of functional adaptability. In this system, neural activity recorded from selected electrode sites controlled movement within a simplified virtual environment modeled after a game of Pac-Man. The organoids received structured sensory-like feedback through patterned electrical stimulation, giving them a rudimentary “sense” of what surrounded them in the virtual maze—such as nearby food or danger. In turn, the system translated their evoked activity into directional movements: Pac-Man advanced toward whichever direction elicited the strongest neural response. Through this continuous cycle of sensing and acting, the closed-loop system enabled the organoids not only to perceive aspects of the virtual environment but also to modify their activity across successive attempts, demonstrating adaptive behavior driven by ongoing interaction.

‍

Trained organoids improved their task performance over time, showing more consistent control signals and better avoidance of danger. Untrained organoids did not show these improvements. Although the task is not directly comparable to human learning, the results suggest that organoid networks can adjust activity patterns in ways that reflect experience-dependent adaptation.

‍

‍

Potential Applications in Neuroscience and Drug Development

The platform is compatible with a 24-well experimental format, allowing multiple conditions to be evaluated simultaneously. This structure may enable comparative studies of disease-associated variants, toxicity assessments involving cognitive endpoints, and early-stage screening of compounds that influence synaptic plasticity.

These findings indicate that human-derived neural networks can exhibit measurable markers of learning and memory processes. As methods advance, OI-based models may offer a means of studying cognitive biology in a controlled laboratory setting, complementing molecular biomarkers and supporting efforts to investigate neurodegenerative diseases, therapeutic mechanisms, and cognition-related toxicity.

‍

To learn more, please view “Predicting Human Cognitive Outcomes with Organoid Intelligence.”

Rountree, Corey, et al. “Long-Term Potentiation and Closed-Loop Learning in Paired Brain Organoids for CNS Drug Discovery.” bioRxiv, 7 July 2025, https://doi.org/10.1101/2025.07.03.663054v1.

‍